VOL 2, NO 5 MAY 1964 AIAA JOURNAL 821

A Generalization of the Direct-Stiffness Method

of Structural Analysis
ROBERT E JONES*
The Boeing Company Seattle, Wash

This paper piesents a generalization of the direct-stiffness method of structural anal-

ysis which is expected to be of particular value in the analysis of plate and shell bend-
ing problems The methods given eliminate the difficulties encountered in obtaining con-
tinuity of deflection and slope between neighboring elements in the structure idealized by a
finite element representation This is accomplished by formulating the method as a vari-
ational problem, thereby gaining for it some of the advantages of a stationary potential energy
formulation, and by introducing Lagrange multiplier functions after the manner of E Reiss-
Downloaded by Universitats- und Landesbibliothek Dusseldorf on May 21, 2013 | http://arc.aiaa.org | DOI: 10.2514/3.2437

ner and K Washizu The plane stress and Kirchhoff plate bending problems are considered

Nomenclature stress or deformation forms assumed for the elements of the

idealized structure Although this latter problem involves
€xx,€yy,6xy = strains referred to Cartesian coordinates x, y much numerical calculation, the work consists mainly of
<Txx,ffyy,crXy = stresses referred to Cartesian coordinates x} y matrix manipulation, so that it is conveniently carried out
a,/3 = Lagrange multiplier functions by digital computer
u,v = displacements referred to Cartesian coordinates
The two requirements on which are based the finite element
t = plate thickness developments are that 1) the elements are connected to-
v = Poisson's ratio gether in such a way that no discontinuities of deformation
A = area of plate occur, and 2) the elements are in equilibrium subject to the
Si = internal boundary lines of elements in idealized external loads and the forces they exert on each other Satis-
structures faction of these conditions is generally approximate Cus-
SE = external boundary lines of elements in idealized tomarily, one of them is required explicitly (though not
structures necessarily exactly), whereas the other follows implicitly
E = Young's modulus according to the mechanics of the particular method being
u,v,w = prescribed boundary displacements
Xn,Yn = resultant x, y direction stresses on an element used In the "direct-stiffness method,"1 2 continuity of
_ boundary deformation is an explicit requirement, whereas the equilib-
Xn, Yn = prescribed boundary values of X, Y rium requirement is implicit in the procedures of the method
n = outward normal of a side of an element Attempts to extend the direct-stiffness method have occa-
X, Y = body forces sionally met with perplexing difficulties A basic investiga-
Xn ,Yn = line loads on internal element boundaries tion aimed at explaining these difficulties has yielded two
6n = bending slope of outward normal of element principal conclusions First, the method is actually based
side having normal n, positive upward on the stationary total potential energy principle and is an
w = upward plate deflection normal to its surface application of the Ritz procedure for solving such minimal
MX,My,MXy,

_ QxjQy = plate moments and shears problems Second, a generalization of the method can be
Qn = applied upward line load on element boundary derived through the use of a broader variational principle,
having normal n along the lines of the work of Reissner3 and Washizu 4 The
M = applied line moment on element boundary line first item has since been studied by several researchers; its
having normal n, positive when compressing implications concerning the theory and practice of the direct-
upper surface stiffness method are very important The second item
p = upward surface load provides a means of extending the direct-stiffness method
Pm = upward concentrated loads at nodes to problems in which the deflection continuity requirements
W XX,W yy,W Xy =
of the Ritz procedure appear difficult or impossible to satisfy
with the finite element representation It forms the main
Introduction subject of this paper
The variational formulations offer nearly foolproof pro-
I N the analysis of complex structures, the use of finite ele-
ment representations has proved very useful in recent
years Such procedures involve the breakdown of the struc-
cedures for deriving finite element analysis techniques
They avoid a number of pitfalls inherent in the less rigorous
formulations and have the added advantages of permitting
ture into large numbers of small, interconnected elements and statements to be made regarding bounds on solution quanti-
the representation of the state of deformation and stress ties and convergence to exact solutions with vanishing ele-
within the elements by forms that are simple compared to ment size 5~8
the state of the structure as a whole The analysis problem
is transformed from one of determining the complicated
mathematical functions pertinent for the continuum problem General Discussion
to one of determining the amplitudes of the various simple The contrast between the usual Ritz procedure and the
direct-stiffness method lies in the choice of displacement
Received August 26, 1963; revision received February 6, shapes of the latter Rather than the smooth deflection forms
1964 The assistance of a number of associates is acknowledged of the Ritz method, each extending over the entire structure,
The author is particularly grateful to M J Turner of The Boeing the direct stiffness method uses many localized displacement
Company for encouragement and advice states, each restricted to a small part of the structure Over-
* Research Specialist lapping of adjacent localized shapes is provided to insure
 822 R E JONES AIAA JOURNAL

sufficient generality of their total linear combination The

convenient handling of such deflection shapes is accomplished
through the small element structural representation The
localized deflections are formed by the mutually compatible
deformations of small groups of adjacent elements
Consider Fig 1, representing an array of plane elements
Suppose the shaded region is to be deformed This can be
accomplished by a movement of node 3, with nodes 1, 2, 4,
5, 6 held fixed Sides 1-4, 4-5, and 5-6 must also be held
fixed, but the interior sides must deform to accommodate
the movement of node 3 Further deformations can be
accomplished by deflections of the interior sides For the
sake of independence, these are accomplished one at a time,
and node 3 and all nodes and sides exterior to the group are
held fixed It is noted in passing that the overlap of localized
shapes previously referred to is accomplished by subjecting
the element groups surrounding nodes 1, 4, 5, 6, and 2 to Fig 2 Plane region and partial elemental breakdown
similar deformations, and so on for every node throughout
Downloaded by Universitats- und Landesbibliothek Dusseldorf on May 21, 2013 | http://arc.aiaa.org | DOI: 10.2514/3.2437

the structure to replace them with an implicit status similar to that of the
For convenience in digital computation, the deformations equilibrium requirements themselves The solution of prob
of single elements rather than those of element groups are lems by such a method will in the end provide deflection con-
prescribed The digital machine is instructed to combine tinuity In some cases it will be exact, whereas in others, at
properly the element deformations in order to form those of the discretion of the analyst, the deflections will be con-
the groups This procedure implies certain requirements on tinuous in a weighted average sense
the deflection shapes attributed to each element Consider,
for example, element 1-2-3 in the figure Its available de-
flection patterns must permit the movement of one node at Variational Formulation of Plane Stress Problem
a time and one side at a time Moreover, the precise shapes The plane region shown in Fig 2 is cut by various lines
of its side motions taken individually, as well as those due into a large number of small elements The lines need not
to node movement, must be identical to those for all other be straight for present purposes, and the elements may be
elements, regardless of their form or orientation Without of any shape The purpose of this section is to formulate a
such element deformation characteristics, deflection con- general variational procedure applicable to the structural
tinuity could not be had This discussion is not limited to analysis of this idealized structure Displacement discon-
simple arrays of triangular elements For example, elements tinuities and nonequilibrated stress discontinuities will be
3-4-5 and 3-5-6 could be replaced by a single quadrilateral permitted to occur across common element boundaries
element Much more complicated array and element forms The variational principle of Reissner3 is adopted as a starting
could be considered also, and the same independence require- point with various requirements added to it in order to cope
ments on the individual element deformations would hold with the present problem It will be shown that the variation
true of the functional given in Eq (1) yields all the conditions
There are situations in which the element deformation needed to qualify this small element representation as an ap-
shapes cannot be chosen so as to provide the independence proximate solution method consistent with the laws of the
requirements described This occurs in the Kirchhoff plate plane stress theory of elasticity The quantities subjected
bending problem, where both deflection and bending slope to independent variation are u, v, o-xx, ayy, o-xyj a, ft a and
continuity are required There are other situations in which /? are the Lagrange multiplier functions which serve to con-
it is very difficult, though not impossible, to provide the strain the deformations such that displacement discontinuities
independencies To handle such cases while adhering to the across element boundaries do not occur The quantities
localized deformation concept of the direct-stiffness method, Xn and Yn are constrained to be consistent on element bound-
a basic extension of the method is needed It is necessary aries with the stresses It is noted that the stresses need not
to remove certain deflection continuity requirements from satisfy the conditions of equilibrium, the statical boundary
their explicit status in the digital machine instructions and conditions, nor the conditions of compatibility The dis-
placements need not be continuous across element boundaries,
nor satisfy the geometrical boundary conditions or the condi-
tions of equilibrium The stress-displacement elastic law
need not be satisfied Subscripts 1 and 2 refer to elements
on either side of a common element boundary The func-
tional is

( du&2/ dA —

[a(ui — u%) + fl(vi -

/ bjr

f a(u - u)dSE - f m@(v

J ^Eti J ®Ev

f (Xu + Yv)dA - f _ Xnu dSE - _ Ynv

JA^ J&EX
Vi
ns ^-7T-~ + 1
Fig 1 Array of tiiangular elements in plane deformation
 MAY 1964 DIRECT-STIFFNESS METHOD OF STRUCTURAL ANALYSIS 823

The external boundary SE has been divided into the portions boundary conditions identically, then there is obtained the
over which specific boundary conditions have been required; stationary potential energy formulation, which consists of
e g , SEU is that portion of the external boundary over which only the first, third, fourth, seventh, and eighth integrals
the displacement u is prescribed The integrals over SE ex- Recognition of this fact is aided by reversing the integration
tend over the entire exterior boundary of the region, while by parts process and introducing the strain energy function
those over Si extend over all internal element boundaries expressed in terms of the strains It is worth noting that, by
The integrals over A extend over the entire surface area of formulating the direct-stiffness method by the variational
the region including the element boundaries Performing principle, it is found that it is unnecessary to require that the
the variations, integrating terms like <rxx(d8u/dx) by parts, assumed element deformation states satisfy any equilibrium
and taking the resulting line integrals to the element bound- conditions at all, including, in particular, the differential
aries, it is found that equations of equilibrium inside an element
To reduce the general variational equation to the comple-
8ft mentary energy principle, let the stresses satisfy the differ-
t +
ential equations of equilibrium in each element, the stress-
jump condition across element boundaries and the statical
boundaiy conditions Further, integrate terms like (du/
dx)8crxx by parts, and define the complementary potential

JJlfe-—E—] 5ff ** + = 2E]_-

j-fr
Downloaded by Universitats- und Landesbibliothek Dusseldorf on May 21, 2013 | http://arc.aiaa.org | DOI: 10.2514/3.2437

C CY foU &XX ~ V(Tyy\

B 2(1

The result will be the complementary energy principle, which

"•*
<Txx\ *

here contains implicitly, in addition to the usual compatibility

and geometrical boundary condition requirements, the
2(1 + v) presently pertinent requirement of deflection continuity across
dA + (Xn - Xn)8u dSE
E common element boundaries

f (Yn - Yn)8v dSE - fs (u- u)8a dSE - Functional for Direct- Stiffness Method
J SEY J Eu
The present work is aimed primarily at a needed generaliza-
f (v - v)5p dSE + f Sf 2 tion of the direct-stiffness method of finite element analysis
J SE* J* The formulation given here is a specialization of the equa-
Ynl -f Fn2 — ^ns tions of the previous section and satisfies the need for
v2)dSI - generalization
Suppose that the stresses are defined in terms of the dis-
f (u, - u*)8a dSi - f fa - VJ8/3 dSi - placements by the elastic laws Then the functional
J Si J Sx
[a(Ul V2)]dSl
7 " ~ "
f a(u - - fs
JS J Ev
S(Vi -
f f (Xu
J AJ
Yv}dA - f _ - fS Ynv dSE -
J SEX J EY
f (a - Xn)Su dSE - f 03 - Yn)Sv dSE (2)
J SEU J ^EV
The completeness of the functional can be demonstrated
The first integral requires that the stresses satisfy the differ- where
ential equations of equilibrium The second requires that
the elastic relationship between stress and displacement be -____V- €
€
2
satisfied The third through sixth integrals require satis- 2(1- 2 *"J
faction of boundary conditions on stress and displacement
The seventh through tenth integrals require satisfaction of yields the variational equation
stress jump vs line-load equilibrium and deflection con- 8ft . ,
tinuity, respectively, across common element boundaries ^— oexx + -— , J dA -
t
The last three integrals serve to define a and /3 The func-
tional is seen to provide all the requirements of the elasticity
problem and, therefore, will serve as an acceptable formulation
of the finite element analysis method
The retention in the functional of stresses which are inde-
pendent of the displacements is a degree of generality which OJ

might have been omitted for present purposes It has been

done in order to suggest a means of stress calculation (re- f (u- u)da dSE - fs (t; - dSE -
tention of independent stresses) which has not received con- J SEU ^ Ev
sideration in application of finite element methods, and in
order to permit identification of the present approach as one f a5^ d^ - f /38v dSE - C f (Z5^ + 75^)dA -
J SEH * SEU JAJ
in between the potential and complementary energy formu-
lations It is of some interest to demonstrate the latter fact f _ Xn8u dSE - f _ 7n5f; ^& -
In the first case, if the stresses are defined by the stress- J SEX J SEY
displacement equations in terms of the displacements (the
r * "i + W2 2\ . , Wi + ^2\ ,„ ,.^
second integral becomes identically satisfied) and if the dis- tn 8 —-— + T7
Yn 8 —-— 1J dSi (4)
placements are made to satisfy all continuity and geometrical
 824 R E JONES AIAA JOURNAL

pendent variations Each of the forms resulting from

setting the resulting independent variations to zero will there-
fore contain terms from several elements The process of
gathering together these contributions from several elements
amounts to adding together the stiffnesses of adjacent ele-
ments against the deformations in which they participate
mutually This is termed the "merge," and is difficult in
proportion to the number of elements involved In the
present case, where the explicit continuity requirements are
largely relaxed, there is little dependence between the a* and
bi of different elements and the mathematical forms arising
from the independent variations in the first integral of Eq
(4), each contain, largely, quantities associated with a
Fig 3 Triangular element and coordinate systems single element The merge is therefore quite simple, much
more so in fact than that of the ungeneralized direct-stiffness
which may be found to supply all the conditions needed to method
qualify it as a formulation consistent with the plane stress The functions a and /3 are also represented by assumed
theory of elasticity The first integral yields the internal shapes, for example
Downloaded by Universitats- und Landesbibliothek Dusseldorf on May 21, 2013 | http://arc.aiaa.org | DOI: 10.2514/3.2437

work done in the elements due to variations of their deforma- a = S dhi(s)

tion parameters This is equivalent to the work done on the i
elements by their edge stresses, provided X and 7 are zero; (7)
18 = S dikes')
the latter viewpoint has often been used in formulations of
the direct-stiffness method The second, fourth, and fifth
integrals enforce continuity and satisfaction of geometrical s denotes distance along an element boundary, measuring
boundary conditions The last four integrals account for from one end Figure 3 shows a triangular element in which
the work done by the external loads The third, sixth, and such quantities are designated for all sides hi and fo are
seventh integrals supply the extra equations needed for the assumed functions; c,- and di are unknown constants to be
determination of a, ft, which can be found to be the element subjected to independent variation The choice of the func-
edge stresses on SE and the average of the element edge tions hi and fo is based on the shapes of the deflection dis-
stresses at interior element boundaries In its implicit com- continuity between elements as functions of the side-length
patibility requirements and its retention of the stress un- variables s This is predictable from the forms of the func-
knowns a and /3, the formulation is seen to have the char- tions fi and Qi in Eq (5)
acter of both the potential and complementary energy The functions hi and ki may be chosen such that the vari-
methods The preceding equation is the one needed for the ational formulation yields complete continuity in the final
formulation of the algebraic equations of the generalized solution, or a less rigid requirement of weighted average con-
direct-stiffness method, which is the subject of the next tinuity may be had by reducing the number of functions kept
section in Eq (7) The complete matrix statement of the algebraic
equations arising from Eq (4) is given in Eq (8):
Formulation of Algebraic Equations - -~ r
a, ' ' Ft '
equations bi
Equation (4) yields, as a consequence of setting each of from Stiffness
its independent variations to zero, the set of algebraic equa- da* = 0 matrix
tions governing the structural analysis problem The choice dbi = 0
of the freedoms of a single element provides the substance hi
of these equations Suppose the deformation of a single
representative element is given by •= 1 • (8)
equations 0
u = 2 a,ifi(x,y)
i from
da = 0 Constraint
All zero
in this
I 0
(5)
v = I, bigt(x,y) ddi = 0 matrix submatrix

where /», gi are assumed functions, and the a* and bi are un-
known constants which are subjected to independent vari- . . . .
ation
Consider the first integral of Eq (4) for one element only Integrals two, four, and five provide the lower left sub-
On dropping the ddi and 5bt, the following form results matrix The calculation of the coefficients in this submatrix
requires the determination of the side motions of an element
associated with its a» and bi This may be done, and the
integrations performed, for a single element at a time
(6) However, since the c* and di belong to particular network lines
(between nodes) rather than to the element themselves, a
merging process is needed The fourth and fifth integrals,
which contain u — u and v — v, involve only the a» and bi of
kij is a symmetric matrix which is called the stiffness matrix a single element The second integral contains u\ — u% and
of the element; it is the central quantity of the direct-stiff- Vi — t>2, so that each of the rows of the submatrix pertaining
ness method Its derivation here is based on the strain to Ci, di on an interior network line contain the a; and bi from
energy function <£ The extension of the first integral to two elements—those common to the line The merge of the
the entire area involves the constraints between the a* and lower left submatrix, then, is very simple, usually considerably
bi of neighboring elements due to explicit continuity require more so than that of the upper left submatrix The deter-
ments If such explicit requirements are made, then the a; mination of the upper right submatrix follows the same lines
and bi of an element must be grouped in the required way discussed previously However, the entire matrix is sym-
with those of adjacent elements in order to obtain truly inde metrical, so that this submatrix need not be calculated
 MAY 1964 DIRECT-STIFFNESS METHOD OF STRUCTURAL ANALYSIS 825

The remaining integrals provide the external work terms, a, ft, and w Note that p, Pm, and Qn are upward directed
which furnish the right-hand column of the matrix equation loads and that Mn, the applied line moment, is plus when it
Determination of these terms by the work integrals, as is done produces compression in the upper surface of the no 1 ele-
here, is an improvement over the method of taking all loads ment joining a cut (choice of no__l and no 2 elements is
to the nodes as point loads The generality of the direct- arbitrary) On an external edge, Mn is plus when it produces
stiffness method is thus seen to include point, line, and dis- compression in the upper surface of the element joining the
tributed loads edge On an internal edge, Qn and Mn are divided equally
The equations arising from equating to zero the coefficients
between the elements common to the edge
of dci and ddi on all network lines are the ones which provide
Upon carrying out the variation, there is obtained
continuity of deflection between elements It is their pres-
ence that permits the independence of the a; and bi which
makes the merge of the present method so simple The 8W vy \dA —
OW OW
equations arising from equating to zero the coefficients of 5di
and dbi contain as unknowns the c» and di as well as a; and bi
The total of the unknowns in the present method is larger s

than that of the direct-stiffness method, in some cases much

larger On the other hand, the merge procedure, which is [ad(w, - w,)
the core of the direct-stiffness method, is made much simpler
Downloaded by Universitats- und Landesbibliothek Dusseldorf on May 21, 2013 | http://arc.aiaa.org | DOI: 10.2514/3.2437

by the generalizations discussed here In addition, the

present formulation is particularly susceptible to solution f (w - w)ba dSE - f (On - dn)bft dSE -
by matrix partitioning, which simplifies the calculations of J ^Ew ^ ^EBn
the solution
Wl
f adwn dSE - f S (!50n dS* - f QJ ~ W*dS,
J sEw •> E6n J SI 2
Variational Formulation of Plate Bending
Problem 6nl 6n
f MnS ~ *dS, - f fpSw dA - f Qn
J SI * J AJ J SEQH
The plane region and elemental breakdown shown in Fig 2
is representative also of the plate bending problem; the
loads are now considered to be applied out of plane Again,
the lines need not be straight, and the elements may be of
any shape Each element is presumed to have a simple state <f> is the strain energy function of plate bending expressed
of curvature and twist The deformations are taken to be in terms of the curvatures w xx, w yy, and the twist w xy
consistent with the Kirchhoff hypothesis In-plane deforma- The preceding equation is the one which is used in formulating
tions may be considered separately and are ignored here the simultaneous equations of the generalized direct-stiffness
The concepts explained in previous sections concerning the method
assumed shapes and the variational formulation carry over After two integrations by parts of the first integral, collect-
to the present case ing terms and defining Mn and Q (the plate moment and
A variational formulation will be given for the case in which shear on an edge with normal n) in terms of Mx, My, Mxy,
there may be displacement and slope discontinuities across Qx, Qy, and performing various other manipulations, it is
common element boundaries Rather than start from a found that the preceding functional provides the following
variational principle of the type of Reissner, it will be pre- requirements:
sumed to start that the bending and twisting moments are 1) Satisfaction of equilibrium conditions in the interior
defined in terms of the curvatures and the twist by the usual of each element
equations of plate theory The Lagrange multipliers a and 2) Balance of line load Qn and line moment Mn by the
ft are introduced in order to require, respectively, displace- work equivalents of jumps in transverse shear and moment
ment and normal slope continuity across common element across common element boundaries and by the work equiva-
boundaries By normal slope 6n, will be meant the derivative lents of the shears and moments themselves on external
dw/dft, where n is the outward directed normal from an ele- boundaries (The conditions on transverse shear include
ment boundary Similarly, Mn and Qn are the plate bending the twisting moment, as usual)
moment and transverse shear on an element boundary with 3) The finite transverse loads at element corners associated
normal n The needed functional is with the jump in twisting moment around the corners are
balanced by the work equivalents of applied concentrated
loads at the node points
4) Displacement and slope are continuous across common
element boundaries and satisfy geometric boundary condi-
f a(Wl - wJdSi - f tions, in each case, either exactly or approximately
J £>/ J &x 5) The quantities a and ft are found to be, respectively,
the average of Qn and the average of Mn at common element
f a(w - w)dSE - f boundaries and the values themselves at external boundaries
J &EW J °E where geometric boundary conditions are specified
- Wl + W2
Numbers 1-4 are all the requirements of the plate bending
dSl problem Their representation, by the functional given,
*& —— 2~ establishes its completeness and makes unnecessary any
additional requirements In particular, the satisfaction of
the differential equation of equilibrium in the interior of the
element is unnecessary in the choice of element deformation.
shapes

Formulation of the Algebraic Equations

where 2 extends over all nodes at which concentrated loads
m The use of Eq (10) to generate the algebraic equations of
act The quantities subjected to independent variation are the plate bending problem follows along the lines discussed
Downloaded by Universitats- und Landesbibliothek Dusseldorf on May 21, 2013 | http://arc.aiaa.org | DOI: 10.2514/3.2437

826 R E JONES AIAA JOURNAL

previously in the section on the plane stress problem The References

deflection shape of an element is assumed as is Eq (5): 1
Turner, M J , Clough, R W , Martin, H C , and Topp, L J ,
w = 2 Oifi(x,y) (11) "Stiffness and deflection analysis of complex structures " J Aero-
naut Sci 23,805-823(1956)
2
where the a* are constants to be subjected to variation Turner, M J , "The direct stiffness method of structural
The first integral of Eq (10), taken for a single element, analysis," AGARD Structures and Materials Panel Meeting,
yields the symmetrical-stiffness matrix of the element in the Aachen, Germany (September 17,1959)
3
form Reissner, E , "On a variational theorem in elasticity, ' J
Math Phys 29,90-95(1950)
[fcy] (12) 4
Washizu, K , "On the variational principles of elasticity and
As in the plane problem, extension of the integral to the plasticity/' TR 25-18, Aeroelastic and Structures Research Lab ,
entire area involves constraints between the ai of neigh- Massachusetts Institute of Technology (March 1955)
boring elements due to continuities which may be explicitly 5
Melosh, R , "Basis for derivation of matrices for the direct
required In the present case, the resulting merge turns out stiffness method," AIAA J 1, 1631-1637 (1963)
to be very simple due to the partial relaxation of the explicit 6
de Veubeke, B F , "Duality between displacement and
continuity conditions equilibrium methods with a view to obtaining upper and lower
The functions a. and /3 are again represented by Eq (7) bounds to static influence coefficients," AGARD Structures and
The hi and fo are chosen with regard to the shapes of the Materials Panel Meeting (Pergamon Press, London), to be pub-
deflection and normal slope discontinuities between elements, lished
respectively, based on the functions /»in Eq (11) As in the 7
Melosh, R , "Development of the stiffness method to define
plane stress case, depending on the number of functions re- bounds on elastic behavior of structures," Ph D Thesis, Depart-
tained in (7), the final result may have complete or approxi- ment of Civil Engineering, Univ of Washington, Seattle, Wash
mate continuity The merge process associated with the (June 1962)
Ci and dt is quite simple The complete matrix equation is 8
McLay, R W , "An investigation into the theory of the dis-
in the form of Eq (8) with references to 6* and dbi deleted placement method of analysis for linear elasticity," PhD Thesis,
The discussion of the various parts of the coefficient matrix Dept of Engineering Mechanics, Univ of Wisconsin, Madison,.
given previously applies again to the present case Wis (August 1963)