Zienkiewicz 1964

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Paper No.

6726

THE €?INITE ELEMENT METHOD FORANALYSIS OF


ELASTIC’ISOTROPIC AND ORTHOTROPIC SLABS
by
Olgierd Cecil Zienkiewicz, B.Sc.(Eng.), PhD., M.I.C.E.
Professor of Civil Engineering. University of Wales
and
Yau Kai Cheung
Research Fellow, Universityof Wales
For written discussion

SYNOPSIS
The method of finite elements is extended to problems of bending of slabs.
The simplicity of formulation and of dealing with boundary conditionsallows
hitherto difficult problems of slabs with variable thickness and of orthotropic
slabs to be dealt with. The slab can be included in the analysis of a more
complex structure involving other types of members (such as beams, columns,
etc.) without difficulty. Several illustrative examples are given to show the
accuracy attainable and the range of problems that can be tackled by the
method,

INTRODUCTION
FORthe solution of slab problems even of moderate complexity, numerical
procedures have to be adopted. Of these, the finite difference approach has
to date been the most versatile and its well-known success amply justified.
In recent years another alternative procedure has been suggested. This is the
so-called ‘finite element’ approach pioneered by Clough and others.lSa
2. The philosophy of the ‘ h i t e element’ formulation is considerably dif-
ferent from that of the ‘finite difference’. While in the latter a numerical
approximation is made to the exact mathematical differential equation govern-
ing the problem by concentrating on a number ofselected values of the unknown
function at specified mesh points, in the former, a more physically obvious
approach is made. The structure, in the present case the slab, is divided into
a series of small elements. These elements are then ‘joined’at specified nodal
points only and continuity, together with equilibrium, is established at these
points. Now, not only the physical picture of theapproximationcan be

The references are listed on pp. 487-488.


47 1
472 ZIENKIEWICZ
CHEUNG
AND ON THE FINITE ELEMENT METHOD

visualized but, what is more important, standard structural procedures (stiff-


ness analysis) can be used for the solution of the composite problem. With
digital computers generally available, this standard formulation is important
and, inaddition, it will be found in general that the resulting system of
equations is better ‘conditioned’ than that corresponding to a finite difference
system.
3. The insertion of prescribed boundaryconditions is a trivial matter
irrespective of whether these are of the displacement or force type. Re-
entrant corners, openings, etc., notoriously difficult to deal with by finite
differen~e,~ do not present any problem here.If the slab is combined with
otherstructural elements such as beams, columns or walls, ‘the general
formulation procedure remains unchanged provided the stiffness character-
istics of such elements are known. This fact is of great value when dealing
with complex structures.
4. While for homogeneous slabs the relative merits of the two approaches
can be debated, the advantages of the finite element procedure fornon-
homogeneous situations willbe evident. Once an elementary matrix fora
typical element is established, this can be combined with others of the same
form but of different properties without difficulty.
5. The most obvious shape of a finite element of a slab appears to be a
triangle. However, certain difficulties arise with such elements, and, in the
present Paper, rectangular and quadrilateral elements only will be discussed.
6. With the help of quadrilateral elements, it is possible to improve
accuracy in areas of special interest by using a localized finer subdivision into
elements. This procedure, equivalent to the use of ‘graded mesh’ in finite
difference, is again easy to apply, the only extra labour involved being the
calculation of some special element matrices.
7. The problem of slabs formulated by the finite element procedure has
received scantattention.Melosh4obtainedsome solutions for isotropic
materials using a matrix derived for a rectangular element. In this Paper a
more general stiffness matrix applicable toorthotropic rectangular slab
elements is formulated, and a general procedure for deriving such matrices for
elements of non-rectangular shapes is presented.

GENERAL OUTLINE OF THE PROCEDURE


8. In the general procedure of the finite element method, theslab is
imagined to be separated into a small number of quadrilateral elements (Fig. 1).
These elements are connected to each other only at their corners or ‘nodal
points’, and at these corners the continuity of deformation and equilibrium
must be established.
9. Considering a typical element A with nodal points i, j , k , l, it is apparent
that three degrees of freedom (or three independent displacement components)
will have to be considered at each point in the problem of transverse bending.
These are two rotations about perpendicular axes X and y within the plane of
the slab
FOR ANALYSIS OF ELASTIC ISOTROPIC AND ORTHOTROPIC SLABS

and the lateral deflexion W. Denoting these displacements by*

U1 = [,l:
0x1
. . . . . .
we can refer to the displacements of the finite element as

1:
Co-ordinates I
Nodal dirplacements

FIG.1 : A TYPICAL PROBLEM FORMULATED IN QUADRILATERAL ELEMENTS

10. Similarly at every nodal point, ‘forces’ FIwill be assunled to exist, each
‘force’ consisting of three ‘components’, in thedirection of the displacements.
These components are simplytwo concentrated moments and a downward
force

. . . . .

and again the forces at the nodes of the finite element may be referred to
simply as

. . . (4)

* Standard matrix notation is used throughout. Since submission of this Paper, a


further publication by Melosh (Basis for derivation of matrices for the direct stiffness
method. J . Amer. Inst. Aero. Astro., vol. 1, July 1963, p. 1631) quotes a stiffness matrix
similar but not identical to the one developed here.
474 ZIENKIEWICZ
CHEUNG
AND ON THE FINITE ELEMENT METHOD

11. Obviously, the nodal forces are a fictitious concept and represent in
some way a static equivalent of the shear forces and moments distributed along
the edges of the element. However, accepting for the moment this simplifica-
tion, it is clearly possible to write generally
FA = K A UA . . . . . . . (5)

where K A representsa stiffness matrixofthe element A . Thismatrix is


composed of 12 X 12 elements or 4 X 4 submatrices kii, kij, etc.
12. Once the stiffness matrices of the elements areestablished,it is an
easy matter to complete the formulation of the whole problem. If, at the
nodes, external forces (either of loading or reaction) Ni are acting (with Ni
having in general three components as before), equilibrium equations at the
jointscan be written. With

representing all the external nodalforces and

giving all the nodal displacements, we have

N = S u . . . . . .
in which S is the stiffness matrix for the whole structure. The elements of
this matrix are built by adding the stiffness of the elements adjacent to a node.
Thus, the elements of S are built up as

si, =A =2 k:...:
1.2.
. . . . . . . (9)

13. The final relation (8) gives now a sufficient number of equations to
solve thecompleteproblem. If no displacements are specified, thematrix
willbe singular. However, in general, some displacement will be known at
the various supports of the plate, or can be arbitrarily specified to prevent
rigid body movements. The matrix is thenreduced by deleting the rows
corresponding to these known displacements and substituting their values in
the appropriate columns, with a resulting system of equations permitting all
the unknown displacements to be evaluated. This procedure will be recog-
nized asthegeneral stiffness analysis procedure used frequentlyin many
structural problems.
FOR ANALYSIS OF ELASTIC ISOTROPIC AND ORTHOTROPIC
SLABS 475
14. As the primary interest is usually in the internal distributed moments
of the elements, these have to be related with the displacements. In general
these will be given as *

[2IA=
a, i
E t uA . . . . . . (10)

where E? is a known matrix (3 X 12). The solution of the problem is thus


shown to hinge on the determination of the matrices K and E which will be
discussed in the following section.
15. As presented above, the procedure follows a series of systematic steps
well suited to programmingona digital computer. In many cases the
matrix S can be simply assembled by hand and the computerused only for the
final solution.

PROPERTIES OF THE ELEMENT

16. Consideration will now be given to a typical finite element as shown


in Fig. 1.
17. If the lateral deflexion W throughout the element is represented by a
polynomial in X and y , 12 undetermined constants can be used since three
degrees of freedom exist at each node. It is immediately apparent that
W = A1 +AZX+ A3y +A& + A ~ X +J JA& + A7x3 + ABxZy
+ A s x y ~ + A l o y 3 + A l l X ~ Y + A 1 2 X Y 3 (11)
satisfies all the above conditions.f
18. Substituting the co-ordinates of each point in turn, we have

in which C is a 12 X 12 matrix simply written in terms of the co-ordinates of


the nodes and A a column matrix of the constants A . Thus,
A = C-l uA . . . . . . . (14)

* M,, My, M,, are internal moments defined in the conventional manner for slabs.
f As the even 4th order derivatives are zero,
it is clearthat the deflexion formassumed
also satisfies the well-known equation governing the deflexion at
an unloaded plate.
476 ZIENKIEWICZ
CHEUNG
ANDON THE FINITE ELEMENT
METHOD

[-Tw
19. The curvatures and twists at any point of the plate can now be deter-
mined in terms of the constants, and therefore
--a 2 W
X = 3x2 ]=BA=BC-IuA . . . . (15)

2-
axay
20. The internal moments are related to the curvatures by known expres-
sions from the theory of plates. For orthotropic materials this relationship
is determined in terms of four constants5

or simply M = Dx . . . . . (17)
where D stands for the appropriate matrix of coefficients.
21. At this stage itis possible to establish the equivalentvalues of the nodal
forces. By the principle of virtual work, if these forces are statically com-
patible withthe internal forces ( M ) ,then, during any virtual displacement,the
external work done must be equal to the internal work. In particular, if the
displacement i s such that it is unity in the directionof a selected external force
and zero in the directionof all other forces,the internal work will be the same
as the value of this selected force. Taking the virtual displacements 8uA as
equal to Z (the identity matrix) and writing out the corresponding external
work in matrix form, we have
W, = (FuA)'FA = IFA = F A . . . . . (18)
To each of these displacements corresponds an internal work done by the
moments equal to
W, = jJ ( 8 ~M) d~x d ~ . . . . (19)
where Sx = B C 1 ( 8 ~ =) ~( B C - l ) I = BC-' . . . (20)
Substituting for Sx and M and equating internal and external work results in
F A = JJ ( B C - l ) TD B C - l uAdxdy = [(C-l)T{J,fBTDBdxdy}C-']uA (21)
The only matrix dependent on the co-ordinates X and y in equation (21) is B
and the integrationof the central part has to be carried outover the entire area
of the element.
22. The whole expression in thesquarebrackets is simply therequired
stiffness matrix K* of the element, while obviously relations (15) and (17)
define the internal moments in terms of the nodaldisplacernents giving
M = (DBC-')uA . . . . . . (22)
Equation (22) is actually a definition of the matrix E t in equation (10).
FOR ANALYSIS OF ELASTIC ISOTROPIC AND ORTHOTROPIC SLABS 477
23. For a rectangular element, matrices K A and EA can be worked out
explicitly and are given in Appendix 1. When a completely general quadri-
lateral is used for anelement, it is convenient to perform thevarious operations
numerically. In fact, the whole process of inverting the matrix C, etc. can
be programmed fora computer and treated as a part of the complete solution.
It is convenient then to derive the integral in an explicit form.
24. The boundary conditions which have to be inserted in equation (6)
are either prescribed displacements or prescribed forces (along a free edge,
these are simply zero). No ambiguity or difficulty is experienced with corners
whether these are convex or concave.

DISTRIBUTED LOADING OR MASS


25. In the previous discussion it was assumedthat theexternal loads applied
to the plate were concentrated at the nodes. With distributed loading over
the area of the element it is possible to proceed in one of two ways. In the
first most obvious approach the totaldistributed lateral load on an element is
assigned as four equal concentrated loads actingat thenodes. This procedure
is reasonably acceptable if the size of elements is small, and has .in fact been
used throughout the examplesgiven in this Paper. To achieveconsistency
and also to improve the accuracy, an alternative method of assigning nodal
loads must be followed.
26. In this alternative procedurethenodalloads are so assigned that,
duringanyvirtual displacement, theworkdone by these is equal to the
corresponding work done by the actual distributed loading. This procedure
is clearly consistent with that used in deriving the stiffness matrix.
27. Consider a set of nodal forces NAequivalent to the distributed loading
acting on a particular element. During the virtual displacement 6uA= 1, the
work done by the nodal forces is
W, = = INA = N A . . . . (23)
and the work done by the distributed lateral loading q,
W , = ( s W ) T q dxdy . . . . . . (24)
Rewriting equation (1 1) as
W = { 1 , ~ , ~ , ~ ~ , ~ ~ , ~ ~ , x ~ , x ~ y , x3y ~LA,
,y~,x~y,xy~}A
we have by (14)
SW = LC-'SuA = LC-' . . . . . . (26)
and equating the two expressions for work done yields
Jr
N A = JJ (LC- 1)Tq dxdv = (c- 1 ) ~ L T q dxdy . . . (27)
28. In general ( L C -l )Twillbe a vector containing twelve quantities, and
it appears, therefore, that to represent completely a distributed load system,
couples as well as lateralforces have to be substituted at nodes. For a uniform
loading and a rectangular element, expression (27) can be evaluated explicitly.
29. Procedures identical to the above can be used when considering the
vibrational behaviour of a plate. Clearly in the first approximation the mass
of the whole plate element can be considered divided into four equal masses
lumped at the nodes. Following the second line of reasoning, however, we
478 ZIENKIEWICZ
CHEUNG
AND ON THE FINITE. ELEMENT METHOD

have to replace the distributeddynamic forces m W* by equivalent nodal forces,


say VA. These will be given by equation (27) by
V A = jJ (LC-l)Tm W dxdy . . . . . . (28)
Using (25) and (14)
i i = LC-'(iiA) . . . . . . . (29)
we have
V A = [(C-l)T(JjLTmL dxdy) C-']iiA . . . . (30)
In equation (30) the expression in the square brackets represents the equivalent
system of lumped masses which has to be added at the nodes when dynamic
problems are considered.

THERMAL EFFECTS
30. Ifitis desired to includeinthe analysis theeffect of temperature
changes, this can be done by rewriting equation (17) as
M = Dk-xT) . . . . . . . (31)
with ,yT denoting the curvature changes occurring in an unrestrained element
due to temperaturealone.Repeatingthesteps (18) to (21) with this new
expression for the moment matrix, it will be found that
FA = KA~A-(C-l)T JJ BD& dxdy . . . . (32)
in which the last term represents thermal forces which have to be added to
obtain the thermal deformations.

SOME TYPICAL EXAMPLES


31. To establish the accuracy of the method several problems, for which
exact solutions are known,have been computed. At the same time, an attempt
was made to compare the accuracy with that attainable by finite difference
procedure. In addition some furtherexamples designed to show theversatility
of the procedure are included.
32. It will be noted that due to the particular formulation of the displace-
ments,thebendingmoments within each element vary linearly. This will
result in a certain approximation in regions where rapid variations occur.
However, as can be seen from the diagrams, this linear variation is of a type
that straddles the true diagram as a best approximation.
33. In addition,adiscontinuity of momentscalculatedfromadjacent
elements is liable to occur at their junctions. However, such discontinuities
are seen to be small, and if a best average curve is drawn, itwill be found in all
the examples to run exceedingly closely to the exact solutions.

Isotropic square plate clamped around the boundaries under uniform load
34. Fig. 2 and Table 1 summarize the results obtained.
35. Three sizesof elements wereused representingthe division of the
whole plateinto 4, 16, 36, and 64 square elements. For comparisonare

* m being the mass per unit area of the plate, and W = da(w)/dta.
FOR ANALYSIS OF ELMTIC ISOTROPIC AND ORTHOTROPIC SLABS 479

DEFLEXION CURVE

16 X 16 finite difference
0 6 X 6 finite element
.. - ..-. .- h 4 X 4 finite element

BENDING MOMENT

FIG.2: CLAMPED
ISOTROPIC SQUARE PLATE UNDER UNIFORM LOAD q

TABLE
1

v = 0.3 Central Maximum,


Maximum,
negative M 11 positive M
Finite
element 2X2
46 x 46
8x8
1 0.00148
0.00140
0.00133
0.001 30
1 -0.0355
- 0.0476
-0.0496
- 0'0503
1
0.0460
00278
0.0249
0.0240
Finite difference E 8 X 8
Exact solution5
1 6 x 16 I
,
0.00141
0.00131
0.00127
I - 0.0471
-0.0503
0.0231
-0.0513
1 0.0233
0.0230
480 ZIENKIEWICZ
AND
CHEUNG ON FINITE
THE ELEMENT METHOD

shown the results of finite differencesolutions on meshes representing a division


into 64 and 256 elements respectively, as well as the results of an 'exact' series
solution.

Isotropic cantileuered square plate under uniform load


36. Fig. 3 shows the plot of deflexions for a solution based on 3 X 3 elements,
and Table 2 summarizes some of the bending moments. It is interesting to see

Finite element
3x3 . -0.60 0.03 - 1.70 -0.57
1 -0.51
~

Point
matching
solution9 . 0.00 ~ -1.26 -0.37

Multiplier qL2/4

(For location of points, see Fig. 3)

TABLE
3
l
~ Point 1 Point 2
W M,
Finite
element
- ~~ - ~~ .-

2x2
4x4
-1
1 0.0126
0.0165
~~

~ 0.0232
0.095
. _,.
0.108
Marcus
Lee and
Bailerjeros'l
. 6. x 6.
. 1
1 0.0173
0.0180
0.0170
'

1
1
0.0244
0.0281
0.0265
0.109
0.110
0.109
Multiplier . . . . .
Point I, centre of side: point 2, centre of plate

TABLE
4

E I = LD Point 1 Point 2

Finite element6 X 6 (~=0.3) 0.0037 0.0332 0.0083 0.061 1


Timoshenko ( V = 0.25) . ~ 0,0087 0.0601
Multiplier . . . . .
Point 1, centre of side: point 2, centre of plate
FOR ANALYSIS OF ELASTIC ISOTROPIC AND ORTHOTROPIC
SLABS 48 1
that even with the very coarse finite element solution, the anticlastic curvature
is observed.
37. The results are compared with others obtained by finite differences,7
experimentally8 and by a point matching s o l u t i ~ n . ~

Isotropic square plate supported at corners (uniform load)


38. Problems of this type are very difficult to deal with by finite differences
and this particular one has commanded an appreciable amount of analytical
work. In finite element presentation, no difficulties are encountered and the
results of the solution with various divisions into elements are summarized in
Table 3, together with solutions obtained by other approximate procedures.

Point2
Polnt I

DEFLEXIONS ALONG CENTRELINE A-A


Point 3
A Finiteelement. (3 X 3)
0 Finitedifference’ (5 X S)
0 Expermental VIIY~S’

DEFLEXIONSALONG B-B
(SHOWING ANTICLASTIC CURVATURE)

FIG.3 : CANTILEVERED
SQUARE PLATE UNDER UNIFORM LOAD q

Isotropic square plate with edge beams supported at corners (yniform load)
39. This example is included to show how the stiffness matrix of the plate
can be combined with that of other elements such as beams. A general stiff-
ness matrix of a beam element (Appendix 2 ) is combined simply with the plate
elements. In this particular case, the results of which are shown in Table 4
where they are compared with those obtained by T i m ~ s h e n k oa, ~6 X 6 division
of the plate was used and torsional rigidity of the beam ignored (though its
inclusion presents no additional difficulty). ’

40. To check on the solution of orthotropic situations a simple example


was worked out.

Simply supporfed orthotropic square plates (ungorrn load)


41. The results are summarized inTable 5 and compared with a series
solution of reference 5. Again the good accuracy is demonstrated.
13
482 ZIENKIEWICZ AND CHEUNG ON THE FINITE ELEMENT METHOD
TABLE5

D, = 6.025 D,,
D1 = 0.375 D,, deflexion
D,, = 0.9375 D,
Finite
element 2x2 1 0.00131 ,
4x4 l 0.00148 0.0767
-
6x6 I 0.00151 1
' 0.0181
0.0222 0.0798
Series
solution
___-__________-
. . . 1 0.00152 1 0,0222 0.0833
Multiplier . . . . 1 qL4/Dx

An isotropic plate with a hole


42. This, and the following examples, are designed to show the range of
applicability and thepower of the method rather than to test its accuracy. In
this case a non-rectangular shapeof elements is needed to approximate to the
shape of the circular opening. Results and the geometry of the problem are
shown in Fig. 4. The slight positive moment at the edge of the hole is due to
the shape approximation, and the probable moment distribution is indicated.

An orthotropic variable thickness slab


43. The situation dealt with here and the results indicated in Fig. 5 are
typical of the range of problems encounteredin bridge deck design. Although

SECTION A-6 ; MxDIAGRAfl L ,02

J
SECTION A-6 : W,DIAGRAfl

FIG.4: SQUARE
PLATE WITH A CIRCULAR HOLE (V =0.3)
FOR ANALYSIS OF ELASTIC ISOTROPIC AND ORTHOTROPIC
SLABS 483
a rectangular shape was used here, a skew situation can easily be worked out
allowing forthe special shape of elements required. The slab issimply
supported along two edges and free along the others.

CONCLUSIONS
44. The few examples included in the Paper which show the comparison
of both deflexions and moments obtained by the finite element method and
available exact solutions, indicate remarkable agreement even for relatively

G,

SECTION A-A : .l/, DIAGRAM

SECTION B-B : .Mx DIAGRAM

FIG.5 A : VARIABLETHICKNESSORTHOTROPICPLATE

FIG.5B: DEFLEXION
CURVESOFVARIABLETHICKNESSORTHOTROPICPLATE
484 ZIENKIEWICZ AND CHEUNG ON FINITE
THE ELEMENT METHOD

coarse division of plate. In particular, the moment distributionfor all meshes


‘straddle’thecorrectresultsindicatingthatapproximationerrorsare not
accumulative.
45. Theratherarbitrary examples of avariable-thicknessorthotropic
plate and of the perforated plate show the possibility of applying the method
to problems which present considerable difficulties when dealt with by other
procedures.
46. Introduction of coupling of the plate to other structural elements, for
which stiffness is known, again presents no difficulty as was demonstrated by
the example of beam support. A slab supportedby columns or elastic founda-
tion presents no additional difficulty as these are simply now elastic com-
ponents of the whole structure.Thus,problems of foundationslabs or
grillages are easily solved by the methods described.
47. The expression assumed for the displacement W is such that continuity
of slope and displacements at nodes ensures that W is also continuous along the
line joining the elements, provided these are rectangular in shape. With non-
rectangular elements, however, their continuity is violated, which may account
for the inferior accuracy achieved with such elements.

1
APPENDIX
48. The matrix K A for a rectangular element shown in Fig. 6 is iven below. The
axes of orthotropy areassumed parallel and perpendicular tothe sifes of the element.

Directions of
prthotropy

FIG.6 : A RECTANGULAR amno-


TROPIC ELEMENT

Writing

* Here F A denotes {T,,/a, TJb, P, . . . etc.}T, U* denotes {aSxl,bey1,W, . . . etc. }T


FOR ANALYSIS OF ELASTIC ISOTROPIC AND ORTHOTROPIC SLARS 485
w e have
K1 =

I-0
0 20
0 3Op-I 60p - Symmetricalp =
0 0 0 0 b
0 10 15p-1 0 20
0 15p-I 30p-' 0 3Op-1 6 0 ~ - ~
0 0 0 0 0
0 10 3Op-I
0 - 3 0 ~ - 1 - 60p-=
0 0 n 0 0- 0- 0 0- 0 0
0 5 15p-1 0 10 3Op-: 0 10 -15p-I 0 20
0 - 1 5 ~ - l - 30p-' 0 - 3 0 ~ - ' - 6 0 ~ - 0 - 15p-I 30p-' 0 - 30p-- l 6 0 ~ - ~

20
0 0
-3Op 0 60p2 Symmetrical
10 0 -3Op 20
0 0 0 0 0
30p 0 -6Op' 30p 0 60p2
10 0 -15p 5 0 2015p
0 0 0 0 0 0 0 0
- 15p 0 30p' - 15p 0 - 3 0 ~ ' -3Op 0 60p2
5 0 -15p 10 0 15p 10 0 -3Op 20
0 0 0 0 0 0 0 0 0 0 0
-
15p 0 - 3 0 ~ ' 15p 0 30p" 30p 0 - 60p2 30p 0 60p'
J
0
15p 30
D O 0 Svmmetrical
0 -15p
- 1 5 p -30
0 15p --l

1
O n 0 0 0 li: n
15p-1, 0 -30

!
0 0
0 0
0 3 0
L _I
K4 =
l- -1
486 ZIENKIEWICZ
CHEUNG
ANDON THE FINITE ELEMENT METHOD

I l

l
FOR ANALYSIS OF ELASTIC ISOTROPIC AND ORTHOTROPIC SLABS 487
The internal moments are given by the equation
1
M = - E A uA,
ab
and E A is defined on p. 486.

APPENDIX
2
49. The stiffness matrix of a beam element (Fig. 7) is as follows:

2 -6
6 -12
4 -6
6 12
Y -Y
-Y Y

y =
GJ
-.
where
EI

REFERENCES
1. M. J. TURNER, er al. Stiffnessanddeflexionanalysis ofcomplex structures.
J . Aero. Sri., vol. 23, no. 9, 1956, pp. 805-823.
2. R. W.CLOUGH.Thefiniteelementmethod in planestressanalysis.Proc. 2nd
Conf. on Electronic Computation, Amer. Soc. civ. Engrs,8-9 September, 1960.
3. D. N. DE G. ALLEN and R. T. SEVERN. The stresses in foundation rafts. Proc.
Instn ciu. Engrs. Part I: vol. 15, January 1960, pp. 35-48. Part IT: vol. 20,
October 1961, pp. 293-304.
4. R. J. MELOSH.A stiffnessmatrix for theanalysisofthinplatesinbending.
J. Aero. Sri., vol. 28, 1961, pp. 34-42.
5. S. TIMOSHENKO and S . WOINOWSKY-KRIEGER.‘Theory of platesandshells.’
2nd edition, McGraw-Hill, New York, 1959.
6. R. V. SOUTHWELL. ‘Relaxation methods in theoretical physics’, vol.11. Univer-
sity Press, Oxford, 1956.
7. R. K. LIVESLEY and P. C. BIRCHALL.Analysisof a loadedcantileverplateby
finitedifferencemethods.R.A.E.,Tech.NoteM.S. 26, June 1956.
S. J. W. DALLEY.Experimentalvalues of deflexion,stresses,andinfluenceco-
efficients for a thin square plate fixed along one edge. Defense Res. Lab. Rept,
no. 189, Univ. Texas, November 1948.
488 ZIENKIEWICZ AND CHEUNG ON THE FINITE ELEMENT METHOD
FOR ANALYSIS OF ELASTIC ISOTROPIC AND ORTHOTROPIC SLABS

9. A. W. LE~SSA and F. W. NIEDENFUHR. A study of the cantilevered plate sub-


jected to a uniform loading. J. Aero Sci., vol. 29, 1962, pp. 162-169.
10. H. MARCUS. ‘DieTheorieelastischerGewebe und ihre Anwendungaufdie
BerechnungbiegsamerPlatten.’Springer,Berlin,1932,p.173.
11. S. L. LEEand P. BALLESTEROS. Uniformlyloadedrectangularplatesupported
at the corners. Int. J. Mech. Sci., vol. 2, no. 3, 1960, pp. 206-211.

The Paper, which was first received on 26 March, 1963, and in final form
on 9 August, 1963, is accompanied by 7 sheets of drawings, from which the
Figures in the text have been prepared.

Written discussion on this Paper shouldbe forwarded to reach the Institution


before 15 October, 1964, and will be published after February 1965.
Contributions should not exceed 1200 words.-sEc.

You might also like