Sie sind auf Seite 1von 16

An Exact Three-Dimensional Solution for Normal Loading of Inhomogeneous and Laminated

Anisotropic Elastic Plates of Moderate Thickness


Author(s): T. G. Rogers, P. Watson and A. J. M. Spencer
Source: Proceedings: Mathematical and Physical Sciences, Vol. 437, No. 1899 (Apr. 8, 1992),
pp. 199-213
Published by: Royal Society
Stable URL: http://www.jstor.org/stable/52063
Accessed: 13-10-2016 10:43 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

Royal Society is collaborating with JSTOR to digitize, preserve and extend access to Proceedings:
Mathematical and Physical Sciences

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
An exact three-dimensional solution for normal
loading of inhomogeneous and laminated
anisotropic elastic plates of moderate thicknes
BY T. G. ROGERS, P. WATSON AND A. J. M. SPENCER
Department of Theoretical Mechanics, University of Nottingham,
Nottingham NG7 2RD, U.K.

An exact three-dimensional solution is presented for the deformation and str


distribution in an elliptical plate under uniform normal loading of the later
surfaces, and clamped along its edge. The plate is assumed to be of constant
moderate thickness and composed of anisotropic elastic material which is
inhomogeneous in the through-thickness direction but symmetric about the mid-
plane. The only material symmetry assumed is that of reflectional symmetry in
planes parallel to the mid-plane.
A transfer matrix method is used which, without making any further assumptions,
gives the exact solution at each point in the plate in terms of the stress and
displacement at the mid-plane. The two-dimensional differential equations governing
these mid-plane values are found to be the same as those for an equivalent
homogeneous plate whose constant elastic moduli are determined by appropriate
through-thickness weighted averages of the inhomogeneous moduli. The solution of
the two-dimensional problem is known for such a plate when subject to the specified
surface and edge conditions, and yields a closed form analytical solution that satisfies
all the governing equations and surface conditions of the full three-dimensional
elasticity problem, with edge displacement conditions satisfied on the mid-plane.
The important special case of an anisotropic laminated plate is given by assuming
piecewise constant properties through the thickness.

1. Introduction

The particular problem treated in this paper is of a flat, elliptical elast


constant moderate thickness which is clamped along its edge and sub
uniform normal loading of the two lateral surfaces. The plate material is an
and inhomogeneous in the through-thickness direction. For convenience th
assumed to be symmetrical about its mid-surface, but apart from this the
inhomogeneity is completely arbitrary and can be discontinuous. The only restriction
on the material response is that it has reflectional symmetry in planes parallel to the
mid-surface, i.e. it is monoclinic. Accordingly, the theory includes the important
special case of a laminated plate in which each layer is homogeneous and orthotropic,
but distinct from the other layers either by being of different material or by having
a different orientation.
The paper has two principal aims. The first is to present a new exact closed-form
solution to a non-trivial boundary-value problem in three-dimensional anisotro
t This paper was accepted as a rapid communication.

Proc. R. Soc. Lond. A (1992) 437, 199-213 ? The Royal Society, 1992
Printed in Great Britain 199

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
200 T. G. Rogers, P. Watson and A. J. M. Spencer
elastic plate theory. The other aim is to demonstrate a novel general method of
obtaining accurate solutions for three-dimensional elastic problems involving
inhomogeneous plates.
There are relatively few exact solutions to the full equations of three-dimensional
elasticity theory as applied to the deformation of anisotropic plates (Green & Zerna
1960). This deficiency is particularly marked in the context of inhomogeneous elastic
plates, and of those solutions that do exist almost all are restricted to laminates
(Christensen 1979). Apart from simple problems involving homogeneous defor-
mations and uniform shear stresses, the only exact solutions hitherto available are
those of Pagano (1969, 1970) on laminated strips and rectangular plates under
sinusoidal normal pressure, by Basi et al. (1991) on the bending and stretching of an
inhomogeneous, thermoelastic plate subjected to a uniform applied temperature
gradient and edge tractions, and by Fan & Ye (1990a) for a simply supported,
uniformly loaded square plate. To the authors' knowledge, all the other solutions
obtained for anisotropic laminated plates are approximate.
Most of the early work is based on the anisotropic equivalent (Lekhnitskii 1963)
of the conventional, but approximate, isotropic 'thin-plate' theory (Love 1927). This
is termed the 'classical laminate theory' when applied to the special case of
laminated plates (Calcote 1969; Christensen 1979), and assumes the Kirchhoff-Love
hypothesis of straight inextensional normals. It provides reasonably accurate
predictions of the displacement field in thin plates (when compared with experimental
observations) and hence forms the basis of many, if not most, software packages.
However, it yields only average through-thickness values for the in-plane stresses
and gives no information at all about the important interlaminar shear tractions.
Generalizations of this theory to include nonlinear distributions of the displacements
or strains in the thickness coordinate fail to satisfy conditions of continuity of
traction across the interfaces, and alternative assumptions for the stress distributions
in each layer lead to equilibrium being satisfied only in an average sense (see
Christensen 1979; Noor & Burton 1989). All such theories also suffer the disadvantage
of being unable to assess the order of error involved in the approximations.
By considering expansion in terms of an aspect ratio e of thickness divided by a
typical in-plane length, Kaprielian (1985) has shown that, for plates of arbitrary
shape but subjected to edge loading only, the classical laminate solution provides the
leading-order term of an asymptotic expansion of the full three-dimensional solution.
In the special case of isotropic laminates (Kaprielian et al. 1988) the expansion
terminates after four terms, thus giving an exact solution that holds for any value
of e. In a recent separate development, by adapting work by Iyengar & Pandya
(1983, 1986) that used a transfer matrix method (Bufler 1971), Fan & Ye (1990a, b)
have derived solutions in the form of double Fourier series, appropriate for
rectangular plates. The method used in the present paper contains elements of both
of these approaches, being formulated in terms of a transfer matrix and obtaining
solutions in the form of asymptotic expansions; as in Kaprielian et al. (1988), the
exact solution is derived as a consequence of the expansion's termination after a few
terms (three, in fact).
In the next section we present all the governing equations of the problem, together
with the relevant boundary conditions. In common with all plate theories, the edge
conditions are not specified at each point of the edge surface; in our case the zero-
slope condition at the clamped edge is imposed on the mid-surface there. On the
lateral surfaces of the plate, the traction is specified everywhere. In ? 3, the equations
Proc. R. Soc. Lond. A (1992)

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
Inhomogeneous and laminated anisotropic elastic plates 201

are reformulated in a way that is convenient and necessary for the introduction of
the transfer matrix described in ?4. This reduces the problem to solving a system of
partial differential equations in two independent variables only, and it is shown in ?5
how this is done by expansion in terms of e again. The leading order solution is again
found to be that of the classical theory, now expressed so as to incorporate the
general inhomogeneity assumed. Also, as in Kaprielian et al. (1988) for isotropic
laminates, the subsequent successive terms of the solution are expressed in terms of
this classical solution. Hence it shows how solutions can, in general, be obtained to
any desired degree of accuracy, subject only to the limits imposed by the accuracy
of any numerical methods (finite differences, finite elements, etc.) that may be
required to obtain the two-dimensional thin-plate solution. Apart from generating
the present exact solution, described in ?6, the development of the theory also
suggests how exact closed-form solutions may be obtained for other non-trivial
boundary-value problems. The solution obtained in ?6 is true for all monoclinic
inhomogeneous elliptical plates; in the final ?7, we examine its implications for
laminated plates.

2. Notation and governing equations


We consider a plate of uniform thickness 2h of anisotropic, inhomogeneous,
linearly elastic material. We use rectangular cartesian coordinates x1, x2, x3 such that
the lower and upper faces of the undeformed plate lie in the planes X3 = +h. All
vector and tensor components are referred to these coordinates. We denote the
displacement at any point x by u, with components ui, and the symmetric stress
tensor by a, with components -j. Here, and in the remainder of the paper, Latin
subscripts imply values 1, 2 or 3, unless otherwise specified, whilst Greek subscripts
refer to the values 1 and 2; we also adopt the usual summation convention that a
repeated suffix implies summation over its entire range.
The anisotropic response of the material is described by the stress-strain relations
for monoclinic linear elasticity, which can be written as

11- C11 C12 C13 0 0 C16 1,1


'22 C12 C22 C23 0 0 C26 U2,2
33 _ C13 C23 C33 0 0 C36 U3,3 (2.1)
023 0 0 0 C44 C45 0 U2,3+ 3,2
013 0 0 0 C45 C55 0 U1, 3 +U3,1
012 _ _C16 C26 C36 0 0 C66 U1, 2 + U2, 1

where commas denote partial differentiation with respect to the relevant


coordinates x1, x2, x3 (so u2,3 - u2/ax3, etc.). Thus in general the material b
is described by thirteen independent moduli. This number is reduced to nine
behaviour is orthotropic with respect to the x1, x2, x3 axes, with
C16 = C26 = C36 = C45 = , (2.2)
and to five if the material is transversely isotropic with r
example), with the further restrictions:
C12 = C13, C22 = C33, C44 = 2(C22-C23
We also note that if the material is orthotropic or tran
symmetry axis which makes an angle 0 with the xl-ax
Proc. R. Soc. Lond. A (1992)

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
202 T. G. Rogers, P. Watson and A. J. M. Spencer
representation takes the form (2.1), but the 13 moduli are now all related to 95 and
to the nine or five independent moduli of the respective anisotropy. The case of
isotropic behaviour has, in addition to (2.2) and (2.3), the properties
c11 = C22 = A + 2/,, C12 = C23 = A, C44 55 = , , (2.4)
where A and It are the usual Lame moduli.
The plate is also inhomogeneous, with the elastic moduli cll, c12, ... all functions
of X3, but independent of x, and x2. Thus the elastic properties of the plate vary
through the thickness of the plate. In particular, if the plate is a layered or laminated
medium, comprised of layers of different anistropic elastic materials, then the moduli
are piecewise constant functions of X3. The only restriction that we impose on the
inhomogeneity, mainly for the sake of brevity, is that the plate is symmetric with
respect to its midplane, so that all the elastic moduli are even functions of x3:
Cl1(X3) = Cll(-X3), C12(X3) = C12(-X3) ..."
The equations of equilibrium complete the set of governing equations. Assuming
body forces to be zero, these are
O11, 1 + 12, 2 + '13,3 = 0,
012,1+ -22,2+-'23,3 = 0 (2.5)
013,1 + 23,2 + O33,3 =- 0.

The equations (2.1) and (2.5) comprise nine first-order partial differential equations
for the three displacement and six independent stress components. In this paper this
system is complemented by the particular boundary conditions of uniform normal
loading of the lateral surfaces of the plate:
o13, = 0, o23 = 0, o33 =-P+ on X3 = h, (2.6)
together with specified edge conditions.
For our specific example of an elliptic plate with semi-major axis a an
axis b, clamped along its edge, these edge conditions take the form that
require
u = , u3/an = 0 on xl2/a2+x/b2 =, (2.7)
where the xl-axis is now chosen to coincide with the major-axis
c = a/b > 1), and /8an denotes the normal derivative on the edge of i
Conditions (2.7) are the relevant edge conditions that are usually
plate theories. It should be noted that they are not pointwise conditi
that might mean in the context of a 'clamped' edge. If more detailed
specified for the edge of the plate, they would have to be incor
appropriate boundary layer solutions in the regions adjacent to th
Finally we point out that although ul, u2, u3, c13, r23 and c33 must
with respect to x3, even for layered materials, the derivatives ca
ous - at an interface between two different layers, for example - so
the equations so as to avoid any further differentiation with respect

3. Reformulation of the problem


It is convenient first to rewrite the equations in non-dimensional f
the dimensionless variables

U, = ui/a, Tij = cij/c*, (3.1)


Proc. R. Soc. Lond. A (1992)

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
Inhomogeneous and laminated anisotropic elastic plates 203

where c* is a typical stiffness modulus. We also let

X1 = x1/a, X2 = x2/a, X3 = x3/h. (3.2)

Then the configuration of the plate is given by

X + c2X , IX3l < 1, (3.3)

and a dimensionless thickness parameter

e = h/a (3.4)
is introduced.
By substituting (3.1) and (3.2) into the governing equations (2.1) and (2.5), we can
then rearrange the equations in the convenient matrix form

^B oJl J ^(3.5)
and H= CF. (3.6)

Here F, G and H are defined as

PI- [^ p[13] [i]


F= U2, G= TN23, H= T22 (3.7)
33 U3J [T12
and A, B and C are matrix operators which crucially inclu
with respect to X1 and X2 only, and depend on X3 only t

Q44 Q45 -al


A = -Q45 Q55 -2 (38)
-al -a2 ?

la aaa # a'8 a, a,B ra aa


B=- ma#pa a# na,p sa a aa (3.9)
r- a s a -Q33
Ila aa Mica aa ri
C= ma2 a n2a a 2 . (3.10)
la2 a m2a a r2

Here

a _ a/aX1, a2 - a/X2,

Qll Q161 Q16 Q1


I =(l) Q16 Q66 m = (m) [Q66 Q 26

n = (n) =[Q66 Q26 r [Q= 13 [Q36] (3.12)


n=(c)=RQ26 Q22J LQ36J LQ23J
Proc. R. Soc. Lond. A (1992)

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
204 T. G. Rogers, P. Watson and A. J. M. Spencer
with the dimensionless material moduli Qll, Q12 ... defined by

Q ((/c*) (ij--Ci3C3/C33), i,j = 1,2,6


ccij/(C44 55-C45), ,j 4,
Qi3 = C3/33, i = 1,2,6 (3.14)
Q33 = C*/C33 (3.15)
The crucial equation is equation (3.5). Th
from a straightforward rearrangement o
(2.5)3; the second part is determined from
by eliminating u3,3 from equations (2.1)1,2

4. Transfer matrix form

Equation (3.5) has the formal 'solution' (Gantmacher 1960)

[FG] 1= F(4.1)

where the initial functions (Vlasov


specific value of X3; in this problem
so that
F*(X1,X2) =F(X1,X2,0), G*(X,,X2) = G(X1,X2, 0). (4.2)

The transfer matrix P is given by

2 0 A(2n) eA(2n+1)
P=- 2n A A (4.3
n-O Eg\B(2n+l) g(2n) I )

where A 1(k+l)(X_)= -f A() )B" ()(3)d, (4.4)


fo

B(k+l)(X3) - B(6) A(k)(6) d, (4.5)

for k = 0, 1,..., with A() = B(O) = I, the identity matrix.


components of A(k) and B(k) that are required later are giv
Since the plate is symmetric, the analysis can be simplified
as the sum of a 'bending' and a 'stretching' mode. The f
produced by imposing normal pressures of ?+ (p+-p-) o
surfaces respectively, whilst the 'stretching' mode is that
pressures }(p+ +p-) exerted on each surface of the plate. He
plane displacement components u1 and u2 are odd funct
component a33; hence
F* = 0 (4.6)

and it then remains to determine G* such that the stress conditions at the outer
surfaces are satisfied, namely

T13 = 723 = 0, 733 =- on X3 = 1 (4.7)


Proc. R. Soc. Lond. A (1992)

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
Inhomogeneous and laminated anisotropic elastic plates 205

with p = 2(p+-p-)/c*. (4.8)


Similarly, for the stretching mode

G* = 0 (4.9)
and F* is to be determined such that

713 = 723 =0, T33 =-q on X3= 1 (4.10)


with q = (p+ +p-)/c*. (4.11)

5. Expansion solution
In fact, substituting X3 = 1 into (4.1) and the associated quantities of (4.3)-(4.5),
and then using the boundary conditions, does not immediately give G* for bending,
nor F* for the stretching mode. The substitution produces three scalar differential
equations of infinite order for the three components of these quantities, though only
X1 and X2 are now involved; furthermore, the solution must also satisfy the edge
conditions, given by (2.7) for the present case. However, by expanding the unknown
reference values (G* or F*) as power series in e, the problem can be reduced (Watson
1991) to finding a sequence of particular integrals for the classical plate equations.
Thus for the bending mode, equations (4.1), (4.6) and (4.7) yield

=G no [ B 2n (l) ] *, (5.1)
subject to
G1 = G2 , F =-p on X3= . (5.2)
The appropriate expansion for G* is
00

G* = e2n G(2n)(X1,X2), (5.3)


n=O

which, when substituted into (5.1), results in

F = eA(1)G(0) + e3(A(1)G(2) + A(3)G(O)) + e5(A(1)G(4) + A(3)G(2) + A(5)G() +


G = G(?) + e2(G(2) + B(2)G() + e4(G(4 + B(2)G(2) + B(4)G() +... (5
Hence the boundary conditions (5.2) now give, with a = 1, 2 and i = 1, 2,
0 =G(O) + e2{G2) +B(2)G(0} + e4{G(4) +B)G() +B)G(0) + o(e6) (5
and

-p = eA(1)Go) + e3{AA 2 + e{A1)(2) G)+ 6e5{A(1)G(4) +A(A)Gi2) +?A(5)G0)} + 0(67), (5.7)

where the differential operators A(), A(1), B(2) etc., are determined from (4.4) and
(4.5), and evaluated at X3 = 1.
By equating coefficients of powers of e in (5.6), we obtain
G() = 0, (5.8)
G2) =-B 2 G(o) = -B(2) GO) (5.9)
G(4) = -(2G(2) -B(4) G(O) (5.10)
Proc. R. Soc. Lond. A (1992)

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
206 T. G. Rogers, P. Watson and A. J. M. Spencer
and so on. It is easy to show that A() is zero so, from (5.8), the e term in (5.7) is
identically zero. Hence p must be of order 63 for such an expansion to be valid, with

p = 3p (say), (5.11)
and the e3 terms in (5.7) give

A()
31 G()
1 32+A () G2)
2 33 3 +A (3)G -p (5.12)
Substituting for G(2) and G

A (3) -A(1)B(2-A(1) B G( -p' (5.13)


Straightforward evaluation of the operators shows that (5.
qG(?) =-p (5.14)
with

S = {2 [/l(,) 2 +2mfi2m(5) a1 a2

_= {Q1 Q61 2 + 2(Q12+ 2Q66) 2+ 4Q26 1 3+224}, (5.15)

where Qij =3J jd (ij = 1,2,6). (5.16)


Hence the differential operator 9 can be recognized as being
classical plate theory (Lekhnitskii 1963; Christensen 1979) for t
monoclinic plate or laminate, and (5.14) is the classical equ
displacement of such a plate under normal pressure. Since G() i
the expansion of G* = U3(X1,X2, 0), we see that the present m
only to assess the accuracy of the classical theory but also to e
the accurate three-dimensional bending solution in terms of th
function given by the two-dimensional classical equation for p
Continuing the procedure to the e5 terms in equation (5.7) sho
e2G(2 in the G* expansion is given by solving

G2) = G() (5.17)


where X is a sixth-order differenti

=A (l)(B(4) -B-2) B32) +A (3)B(2)-A(5) (5.18)


This operator can be calculated in the same manner as for 9, evaluated at
so the right-hand side of equation (5.17) is a nn fun n pro i known function provided G is
The procedure may be continued, giving yet higher-order derivatives on the r
hand side of the equation but with the left-hand side always being the c
operator Y.
The solution for the stretching mode can be obtained in a similar manner to
described above for the bending mode. For the present problem of uniform no
loading of a clamped plate, the solution is elementary, with

F= F = (0 0 q)T (5.19)
G = eB(1)F = "y/J33
(O O -0eqB (1))T
We note that usually q is of the same order of magnitude as p - when loading
Proc. R. Soc. Lond. A (1992)

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
Inhomogeneous and laminated anisotropic elastic plates 207

applied to only one of the plate surfaces, for example - in which case q is O(e3). The
stretching then contributes only an e4 term to U3 when the full superposed solution
is constructed.
The comments immediately following (5.17) and (5.18) have two important
implications. The first is that any numerical procedure (finite differences, fini
elements, etc.) which is used to solve the classical equation (5.14) for G(3), can also b
used to solve (5.17) for G2) and, in principle, the differential equation governing G(
etc. The second implication is that for some forms of G?, and for polynomials
particular, these right-hand sides could become identically zero. In such cases th
series for G* terminates and the expressions obtained for G* and hence for G*, Fand
G become exact solutions of the governing equations, with no restriction on th
magnitude of e.

6. Rigidly clamped elliptic plate


The edge conditions (2.7) require that for the bending mode we have
G(0) = 0, G(2) = 0,...
G(3o)/-n = 0, dG2)/n =0,... on X2+
These are satisfied by choosing G, G), ..., to be
such that f(0) =f'(0) = 0. Since Y is a quartic o
obvious appropriate solution of (5.14) is

G() = k(1 -X2-c2X2)2, (6.2)


with the constant k given by
k = -p'/R, (6.3)

where R = 3J
fo
2R() d6, (6.4)
and R = {3Q11 + 2(Q + 2Q66) + 3c4Q2}. (6.5)

With this solution for G( , the right-hand side of th


so that
G2 = 0 (6.6)

and the series for G* term


that G 2) and G(2) are linear in X1 and X2 (both B( and B( are third-order differential
operators); hence with B(2) being second-order differential operators and B(43 being
fifth order differential operators, (5.10) gives

G( = G4) = 0,

and thus G(2 = 0, n 2. (6.7)

Equations (5.4) and (5.5), when


mode solution throughout the pla

1U1 eA1X3) G -e3A)X3)B2


U2 = eA13 (X3) G)-e{A21(X)B 3 +A (X3)B2-A )(X3)}G), (6.8)
T33 = - e3{A3) (1) + X(2) (2) -Ay (3) } (3 G)
3 331Proc. R. Soc. Lond. (A 3 (1992)
Proc. R. Soc. Lond. A (1992)

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
208 T. G. Rogers, P. Watson and A. J. M. Spencer
and

723 =-62{B(2) -B(2) X3 9(0)


U3-
U3 =_GB(o).
G3) 2-B(2) BX) )\ (0) e4{B3(X3)B
_4B(2) (X R(2) R(2) 2) (4)
^,-?+ee2B^(X3)
e\B- B^ ,}^ G)-e
.~A3 I CI u33 -13 B +B32j(X3)B(] -B4{(X3
A -- \3U '' "/Y 3 (32)}23
G?. J
.9)33
Here the operators B(2) and B(2) are evaluated at X3 = 1; the dependence on X3
other operators is shown explicitly.
Using (6.2) in (6.8) and (6.9), and incorporating the stretching mode so
(5.19), the exact analytical solution for the present problem is obtained ex
and eventually, as
U1 = 4ekX1X3( 1-X2-c2X2)- e63[X1 S (X3) +c2X2 T(X3), (6.10)
U2 = 4ekc2X2X3(1 -X1-c2X~) + e 3[X S2X(X3) + c2X2 T2(X3)], (6.11)
U3 k( -X2 - 2X2)2 - 4e21c[( 1- 3X2 -c2X2) {13(0)-13(X3)}
- 4c2X,X,{
-4cX1 ) -XIX,)} + c2(
X2{I36(o)-136(X3)} -XI 1-x-3c2X)
+ c2( - 3C2X2) {I
{7,() - 3(^3)}]
23(0)-I23(X3)}]
+ e4kW(X3), (6.12)
for the displacement field, where all the functions which depend on X3 are integrals
with respect to X3 and, apart from incorporating c2 in them for convenience, all
involve only the elastic moduli. The Iij(X3) are simply the integrals of the linearly
weighted Qij(X3):
Iij(X3) = Qi j(6)d (i,j = 1,2,..., 6). (6.13)
3

The remaining (so far undefined) functions of X

R* = {L(6)+c2N(6)}d6, (6.14)

Si = [Q44(6) L()-Q45(6)M(6)-2413s(6)-
+ 8X3{3I13(0) + c2I2

T1= [Q44()M(6)-Q45(6)N(6)- 1636(g)] d6+ 16X3136(0), (6.16)

S = [Q45() L(6) - Q55(6)M(6) 16C2 )] d- 16c2X3I36(0), (6.17)

T2 = [Q45() M()- Q55 (6) N(~) + 8I13(~) + 24c2I23(6)] d6


o

- 8X3 {113(0) + 3c2123(0)}, (6.18)


X3

W = [Q13() S1() ? Q36() {c2T1(6) -2(6)}


Jo

- cQ23(6) T2() + Q33(6) {R*(6) +qR/p}] d, (6.19)


where
L = 8{3111 + c2 (12 + 2166)} (6.20)
M = 24{I16 + C2126}, (6.21)
N = 8{112 +2166 + 3c212}. (6.22)
Proc. R. Soc. Lond. A (1992)

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
Inhomogeneous and laminated anisotropic elastic plates 209
It can be seen that the dependence on X1 and X2 is relatively simple. Furthermore
all the integrals are straightforward quadratures and can be computed independently
of the position coordinates X1 and X2. We also note that the solution for the mid-
surface displacements is exactly the same as the classical solution.
The stress solution is simpler, and more interesting. The out-of-plane stress
components are
T13 = 6-1p[X1L(X3) + c2X2M(X3)]/R, (6.23)
723 = e- p[XlM(X3) + c2X2N(X3)]/R, (6.24)
33 = -(p/R)R*(X3)-q. (6.25)

where the plate constant R has already been def


R* through
=R*(1) = L(6) + cN(6)} d (6.26)

The in-plane components T11, T12 and T22 are deter


a little more complicated; for example, Tll is given

T1 - 4ekX3{Q11( - 3X1-c 2)- 4Q16 2X1 X2 +


- e3k{Q11, S+ Q16(S2 + C2T1) + Q12 c2

and similar expressions hold for T12 and T22.


We note that shear stress components r13 and cr2
gives no information, are of order e-l(p+-p-), as ex
and are now linear in X1 and X2. Furthermore, th
a property which may be considered surprising in v
are the relevant shear moduli. It is also clear that
negative as well as positive (from (3.13)), then the
so as to minimise the shear traction (Cro + cr23) o
relevant Qij are all non-negative (as for an isotropi
& Spencer 1989)), then (6.13) and (6.20)-(6.24)
traction must occur on the mid-surface x3 = 0.
The analysis also shows that the normal stress component r33 is independent of
X1 and X2, but its variation through the thickness of the plate is non-trivial; the
classical laminate theory gives no accurate information about this component either.
Lastly, we note that it can be verified that by substituting the appropriate moduli
(refer (2.2)-(2.5)) into the present analysis we recover the results previously obtained
(Rogers 1990) for isotropic inhomogeneous plates of moderate thickness.

7. Laminated plates
An important special case of an inhomogeneous plate is that in which the plate is
laminated, so consisting of a number of uniform layers of homogeneous but different
anisotropic elastic materials (which includes the case of orthotropic or transversely
isotropic layers of the same material but with each having a different orientation
with respect to the axes of the elliptic plate).
We consider a plate with 2N+ 1 homogeneous layers, perfectly bonded at their
interfaces. Each of the layers has monoclinic behaviour but, in general, with different
elastic moduli or at least with different orientation. Since the plate is symmetric, the
rth-layer above the mid-surface is identical in thickness and moduli to the rth layer
Proc. R. Soc. Lond. A (1992)
Is83^~~~~~~~~~~~~~~~~~ ~~Vol. 437. A

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
210 T. G. Rogers, P. Watson and A. J. M. Spencer
below the mid-surface. Any quantity relating to the rth layer will be identified by the
index r, with r = 0 corresponding to the lamina containing the mid-surface and
r = N denoting the outermost layers. If the rth lamina has uniform non-dimensional
thickness 2Hr and elastic moduli c() (and equivalently (r), then in particular
N

Ho+2 E Hr = 1.(7.1)
r=l

We denote the position of the interface between the (r-1)th


X3 = Zr so that the rth layer is defined by Zr < X3 < Zr+i with Z1
Thus Qij(X3) = Q0), Zr < X3 < Zr+l (7.2)
From (6.13) it follows that, at the interface X3 = Zr,

- 2---2
(Zs+l--Zs)
s2 I
Q^) (7.4)
s=r

with the value of Iij within the rth layer given by


_-=(X2
Iij (X3) Iij (Zr) _ Z2)-r)
12 3-r )3
= Iij(Zr+l) + r(Z_-X) Q (7.5)
where IJ(ZN+1) = 0. Hence the value of ij at any poin
a simple quadratic interpolation between the value
-ij(X3) = ij(Zr)+ {Iij(Zr+l)-Iij(Zr)} (X3-Zr)/(Zr+Zr) (7.6)
Obviously, from (6.20)-(6.22), L(X3), M(X3) and N(X3) have the same piecewise
quadratic dependence on X3.
R*, S1, T1, 82 and T2 involve integrations of Ij terms from 0 to X3, and are therefo
piecewise cubic functions of X3. Thus noting that each is of the form

Y(X3) = y(6)d6 +xX3, (7.7)

where, from (7.6), the integrand is given in


y() = (Zr) +{y(Zr+l) - y(Zr)} (X -Z)/(Zr -), (7.8)
then the interface values of Y can be shown to satisfy the simple recurre

Y(Zr+i) = Y(Zr) +Hr2a + y(Zr) + y(Zr+l)- r {Y(Zr+l)-Y(Zr)}/(Zr +


for r = 1,2,...,N, with Z1 = Ho and

Y(Z1) = 1H0{3a + y(O) + y(Z,)}. (7.10)


At intermediate values of X3, with Zr < X3 < Zr+, we have
Y(X3) = Y(Zr) + (X - Zr) {(Z,) + }
(X3 --Z1)2(X3 ? 2Zr)
+ 4H2(2H
rH (2Hr ? + 3 (Zr) - Y(Zr)- 2Hr{y(Zr) + }]. (
3Zr)

An example of (7.7) is R*, for which a = 0 and


Y(6) = YR(6) = L() +c2N(). (7.12)
Proc. R. Soc. Lond. A (1992)

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
Inhomogeneous and laminated anisotropic elastic plates 211

Then in particular (7.9) gives

R = YR() = YR(6)d
/o
N

H0{YR(0) + yR(Zl)) + Hr{YR(Zr)+ YR(Zr+l)}


r-=

1N
- E Hr{Y(Zr+)-Y(Zr)}/(Zr+Hr). (7.13)
r=l

The remaining function W involves integrals of the p


etc., and hence is itself a piecewise quartic function of
and the integrand is of the form
Ys(6) = Ys(Zr)?+ ar+ br 6 +Cr 62+ dr3, Zr < 6 < Zr+l (7.14)
with k = 0. Then

Ys(X3) = Ys(Zr) + ar(X3 - Zr) br(X3,-Zr)2 + Cr(X3 -Zr)3 + -dr(X-,Zr)4 (7


and the interface values are related through

Ys(Zr+l) = Ys(Zr) + 2Hr ar + 2H br +H + 4H4 dr (7.1


The solution can be evaluated in a straightforward manner. An efficient metho
first to use (7.3) to evaluate and store all the interface values of Ij, starting from
outer boundary X3 = 1 and working in towards the mid-plane X3 = 0, i.e. ta
successive values of r = N, N- 1, N- 2, ... down to 0, with Z1 = Ho and Z0 defined
zero. At each value of r, the relevant interface values of L, M and N are also evalu
and stored.
The evaluation of the interface values of the other functions is then carried out by
using (7.10), and then (7.9), with y(6) taking the appropriate form for each function
at each successive value of r, increasing from 1 to N. In particular, for y = YR
(7.12), the cumulative value given by (7.13) gives R. Taken in conjunction with t
expressions in ?6 for the displacements and stresses, this completes the solution of
the problem.

One of the authors (P.W.) has been supported during this work by an ear-marked Research
Studentship and Research Grant no. GR/F21770 from the Science and Engineering Research
Council. Some of the work was undertaken under NATO Collaborative Research Grant no. 870496.
All of this support is gratefully acknowledged.

Appendix A. Components of the plate transfer matrix


From (4.4) and (4.5), we obtain

A(1)= J A()d6, B(2 = B(6)A(l)dg


where A and B are given by (3.8) and (3.9). Hence,

A(1) = Q44(6) d6, A (1) =- Q Q55() d, A


A(1)
P13 -XX3,
L AcA1) -X3a2,
(123 99)A(1)
833= 0,
Proc. R. Soc. Lond. A (1992) 8-2

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
212 T. G. Rogers, P. Watson and A. J. M. Spencer
and then, for example,

B(2) - {( ^FX3 )d iee6 a :+f ^(dca,( ^,


B(2) -- f

23= X3
r(6) r3
d6 a a ?a +{ Jn: s)(6) d 6 a2 '

31 J J {Q44(y) r,(6) a, + Q45(y) Sa(6

/2 o fo
(A3 f

32 = J J {Q45(y)r r(6a) a -Q55() s(6) a - Q33(6)2

Further application of (4.4) and (4.5) gives A(3), B(4),

References

Basi, S., Rogers, T. G. & Spencer, A. J. M. 1991 Hygrothermoelastic analysis of anisotropic


inhomogeneous and laminated plates. J. Mech. Phys. Solids 39, 1-22.
Bufler, H. 1971 Theory of elasticity of a multilayered medium. J. Elasticity 1, 125-143.
Calcote, L. R. 1969 The analysis of laminated composite structures. New York: Van Nostran
Reinhold.

Christensen, R. M. 1979 Mechanics of composite materials. New York: John Wiley & Sons.
Fan, J. & Ye, J. 1990a An exact solution for the statics and dynamics of laminated thick plat
with orthotropic layers. Int. J. Solids Structures 26, 655-662.
Fan, J. & Ye, J. 1990b A series solution of the exact equations for thick orthotropic plates. In
J. Solids Structures 26, 773-778.
Gantmacher, F. R. 1960 The theory of matrices, vol. 2. New York: Chelsea.
Green, A. E. & Zerna, W. 1960 Theoretical elasticity. Oxford: Clarendon Press.
Iyengar, K. T. S. R. & Pandya, S. K. 1983 Analysis of orthotropic rectangular thick plates. Fibr
Sci. Technol. 18, 19-36.
Iyengar, K. T. S. R. & Pandya, S. K. 1986 Application of the method of initial functions for the
analysis of composite laminated plates. Ingenieur Arch. 56, 407-416.
Kaprielian, P. V. 1985 The effects of holes and free edges on the stress in laminated plates. Ph.D.
thesis, University of Nottingham.
Kaprielian, P. V., Rogers, T. G. & Spencer, A. J. M. 1988 Exact theory for the stretching and
bending of laminated plates by edge loading. I. Isotropic plates. Phil. Trans. R. Soc. Lond. A 234,
565-594.

Lekhnitskii, S. G. 1963 Theory of elasticity in an anisotropic elastic body. San Francisco


Day.
Love, A. E. H. 1927 The mathematical theory of elasticity, 4th edn. Cambridge University Press.
Noor, A. K. & Burton, W. S. 1989 Assessment of shear deformation theories for multilayered
composite plates. Appl. Mech. Rev. 42, 1-13.
Pagano, N. J. 1969 Exact solutions for composite laminates in cylindrical bending. J. Comp. Mats.
3, 398-411.
Pagano, N. J. 1970 Exact solutions for rectangular hi-directional composites and sandwich plates.
J. Comp. Mats. 4, 20-35.
Rogers, T. G. & Spencer, A. J. M. 1989 Thermoelastic stress analysis of moderately thick
inhomogeneous and laminated plates. Int. J. Solids Structures 25, 1467-1482.
Proc. R. Soc. Lond. A (1992)

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms
Inhomogeneous and laminated anisotropic elastic plates 213
Rogers, T. G. 1990 Exact three-dimensional bending solutions for inhomogeneous and laminated
elastic plates. In Elasticity: mathematical methods and applications (eds. G. Eason & R. W. Ogden).
Ellis Horwood.

Vlasov, V. Z. 1957 The method of initial functions in problems of the theory of thick plat
shells. Proc. 9th Int. Cong. Appl. Mech. 6, 321-330.
Watson, P. 1991 The stress and deformation of laminated elastic plates and shells. Ph.D. th
University of Nottingham.

Received 20 December 1991; accepted 20 January 1992

Proc. R. Soc. Lond. A (1992)

This content downloaded from 90.147.23.92 on Thu, 13 Oct 2016 10:43:54 UTC
All use subject to http://about.jstor.org/terms

Das könnte Ihnen auch gefallen