COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING 7 (1976) 201-217

0 NORTH-HOLLAND PUBLISHING COMPANY

INCREMENTAL VARIATIONAL PRINCIPLES AND

FINITE ELEMENT MODELS FOR NONLINEAR PROBLEMS

G. HORRIGMOE and P.G. BERGAN

Institutt for Statikk, The Norwegian Institute of Technology
The University of Trondheim. 7034 Trondheim-NTH, Norway

Received 24 March 1975

The classical variational principles are formulated for nonlinear problems by considering incremental deformations
of a continuum. Associated finite element models are derived, adopting the terminology introduced by Pian for clas-
sification of linear finite element models. It is demonstrated how the classical incremental variational principles can be
modified by relaxing the continuity requirements between adjoining elements. Nonlinear hybrid finite element models
arising from such modifications are discussed.

1. Introduction

The finite element method is now widely recognized as a powerful tool in solving problems in
structural and continuum mechanics. The most common formulation in linear, elastic, small dis-
placement analysis is based on the principle of virtual work or the corresponding minimum poten-
tial energy principle. This approach implies assumption of displacement functions which maintain
compatibility between neighbouring elements, and hence it has been termed the compatible dis-
placement model.
The basic energy theorems of structural analysis were derived in the pioneering work of Argyris
[ 11 and later applied to finite element analysis by Fraeijs de Veubeke [2]. In the last decade a
number of new finite element models based on the alternative variational principles of elasticity
have been proposed. For some time, the lack of a common terminology prevented a broader under-
standing and use of the new models. However, in a series of papers, Pian [3-51 and Pian and his
associates [ 6-81 have classified the various finite element models for linear problems in accor-
dance with the associated variational principles. Similar classifications have been reported by
Gallagher [ 9 1 and Wolf [ lo].
In recent years, the need for highly accurate analyses of structures with both material and
geometrical nonlinearities has increased considerably. Following the success of the finite element
method in linear elasticity problems, attention has been drawn towards more complex, nonlinear
structural applications. Up to the present, existing finite element programs for large displacement
analysis have, almost exclusively, been based on the principle of virtual work or the associated
principle of stationary potential energy. Numerous applications have been published as indicated
in the survey papers [ 11,121. Although mixed and hybrid formulations have been employed with
considerable success in small displacement analysis, very few attempts have been made so far to
extend these models to deal with nonlinear problems. This may in part be due to the lack of incre-
mental fornulations of the associated variational principles.
202 G. Hornigmoeand P.G. Bergan, Incremental variationalprinciples

The incremental approach of analysis is closely related to the theory of a body under initial stress.
Variational principles for initial stress problems and also for large displacement analysis have been
treated by Washizu [ 13,14]. Incremental formulations of hybrid displacement and mixed models
were developed by Pian and Tong [ 151, who derived the associated incremental variational prin-
ciples from the second variations of the appropriate functionals. This work was later extended by
Pirotin [ 161 who also developed a modification of the hybrid stress model for nonlinear problems.
A unified approach to linearized buckling analysis based on mixed and hybrid formulations was
described in [ 33 and [ 17 f .
In the past few years, several applications of hybrid and mixed finite element models to non-
linear problems have been reported in the literature. These include linear buckling problems
[ l&--20], large displacement analyses [ 2 l-281, and problems involving both geometric and
material nonlinearities [23,25,26].
The present paper aims at presenting a unified and consistent development of incremental varia-
tional principles and associated finite element models for large deformation problems. However,
the formulations are also valid for material nonlinearity when appropriate incremental strain
energy potentials can be defined [ 291: The incremental variational principles are derived directly
from an incremental formulation of the principle of virtual work. The derivations of the associated
finite element models are given in detail, adopting the terminology suggested by Pian for linear
finite element models. A more comprehensive treatment of the subject will be published later
1301.

2. Description of motion

Consider the motion of a material body in a three-dimensional Euclidean space (see fig. 1). The
motion of the body is a continuous one-parameter family of configurations Ft, where the parame-
ter t denotes time. In order to describe the kinematics of the continuum, reference is made to
three configurations, the initial configuration F,-, corresponding to t = t,, the updated configura-
tion Fl at t = t,, and the current configuration Fz (t = tz). It is assumed that F2 is close to configu-
ration F1, i.e. t, = t, + At where At is a small quantity,
Several choices of describing the motion are possible due to the freedom in selecting a reference
configuration. Two methods commonly employed in continuum mechanics are the total Lagrangian
formulation where the initial state Fe is taken as the reference configuration, and the updated
Lagrangian formulation in which all statical and kinematical variables are referred to the updated
configuration l? 1. A comprehensive treatment of the desc~ption of motion of a material body is
available in [ 3 11, fn the present paper the updated Lagrangian approach is chosen. However, it is
worth noting that the following derivations could also have been formulated in a total Lagrangian
reference frame in essentially the same manner.
In fig. 1 the coordinates describing the initial stress-free configuration F0 are denoted by Xi
(i = 1,2, 3). For convenience, rectangular Cartesian coordinates are used throughout. The coordi-
nates xi (i = 1,2, 3) are associated with the reference configuration I’i . The quantities necessary
for defining the state of stress and deformation of the body in configurations Fi and FZ are listed
in table 1. It is emphasized that all statical and kinematical variables are measured with respect to
the reference configuration l?, .
 G. Horrigmoe and P.G. Bergan, Incremental variationalprinciples 203

Fig. 1. Coordinate systems.

Table 1
Statical and kinematical variables associated with configurations rl and rz

Updated configuration r 1 Current

(reference configuration) configuration l3

Time t1 t2 = tl + Af
Volume V’
Surface s1=s;+si
Prescribed volume forces Fi Pi + M-i
Prescribed boundary tractions Fi T; + ATi
Displacements ui Uif AUt

Prescribed displacements iii Ui + Aiii

Stresses Oij 0~ + A&j
Tractions Ti Tt + ATt
Strains q EV+ AE..Y

3. Incremental formulation of the principles of virtual work and complementary virtual work

3. I. Incremental virtual work

Using the principle of virtual work, the equilibrium of the body in configuration rz require
that
204 G. Horrigmoe and P.G. Bergan, Incremental variational principles

j(a,+ASii)GAEijdV- j(F,+A1;;,)6Au,dV- j(~i+AT,)6A~idS=0, (3.1)

V’ V’ J

where aii are Cauchy stresses, ASij are the incremental components of the second Piola-Kirchhoff
stress tensor, and SA is the portion of the total surface S’ where tractions are prescribed. The
stresses ASij are referred to coordinate axes xi (i = 1, 2, 3) and volume V’ in the reference state
I’r . For a closer description of the second Piola-Kirchhoff stress tensor the book by Fung 1321
is recommended. .Moreover, it should be noted that both the incremental prescribed volume
forces AFi and the incremental prescribed surface tractions A Ti are measured with respect to I’, .
The components of Green’s strain tensor due to the increments in displacements are given by

AEii = ;(Au,~ + Au,~ + A~,,iAuk,j) = AC, + Anti 3 (3.2)

where
A~, = ~ (A~ij + AUj,i) and Aqij = i A.U,,iAU,,i . (3.3)

Eq. (3.2) is now substituted into the virtual work equation (3.1). If the reference state Fr is in
equilibrium, the incremental form of the principle of virtual work arises when the virtual work
equation for state I’, is subtracted from eq. (3.1):

j AS, 6Ae-, dV + j a,jsAqii dV- j AF,SAUi dV- j ATiSA~i dS= 0 ) (3.4)

V’ V’ V’ SC+

where higher order terms in the incremental strain expression have been neglected.

3.2. Incremental complementary virtual work

The equilibrium of the body in configuration Fz may also be expressed by the principle of
complementary virtual work,

j (fij + AEii)sAS, dV - j(iii + Ac\ui)GATidS= 0 , (3.5)

Vl si:

where eij are the components of the strain tensor in configuration I’, , and Sh is the portion of S’
where displacements are prescribed. The virtual stresses must be statically admissible, i.e. they
must satisfy the equilibrium equations in Vv’ and the mechanical boundary conditions on S$
Assuming that the reference state I’r is an equilibrium configuration and neglecting higher
order terms in Green’s strain tensor AE,, eq. (3.5) is reduced to

jAe,&As,dV- jAu,SAT,dS=O, (3.6)

V’ s:

where the complementary virtual work equation for state Fr has been subtracted from eq. (3.5).
The above equation is a mathematical statement of the incremental form of the principle of com-
plementary virtual work.
 G. Horrigmoe and P.G. Bergan, Incremental variational principles 205

4. Incremental formulation of the classical variational principles and related finite element models

4.1. Stationary incremental potential energy and the compatible displacement model

In order to transform the principle of virtual work given by eq. (3.4) into an incremental form
of the principle of stationary potential energy, the following assumptions are made
1) There exists an incremental strain energy function U,, defined by

z= AS, or 6U, = ASii6Aeii. (4.1)

2) The increments in volume forces AFi and boundary tractions ATi afe locally conservative
- this means that there exist incremental potential functions for AFi and AT,. Thus, non-
conservative loadings may usually be treated as piecewise conservative loads.
With these assumptions eq. (3.4) may be written as

6 $ ‘AdV+ $ ciil ‘Au,,~AQ dV- $ A&Aui dV- s ATiAui dS =0 (4.2)

( V’ Vl Sk
I
or
6AII, = 0, (4.3)

where the incremental potential energy functional AII, of the body is given by

AI-I, = j- U, dV + s aii ;Au,,~ Au,,~~V- s AFiAuidV- JATiAuidS. (4.4)

Vl Vl V’ SA

Eq. (4.2) represents the incremental form of the principle of stationary potential energy.
It can easily be verified by calculus of variations that eq. (4.2) leads to the equilibrium equations

ASii,i + (oki Au,~), k + A& = 0 in I” (4.5)

and the mechanical boundary conditions

ATi = AS, vi + aki vk Au,.~ on SJ , (4.6)

where vi are the direction cosines of the surface normal.

Compatible displacement model

In the following, piecewise linear material behaviour is assumed for each increment, i.e.

AS, = Ciikr AeR, . (4.7)

Eq. (4.4) is now recast into finite element form:

206 G. Horrigmoe and P. G. Bergan, Incremental variational principles

AI& = C s ;Cii,,AeiiAekl dV + J CJ,~~AU~,~AU~,~

dV- J AFiAUi dV- J ATiAUi dS ,
n
Vn vn Vn s% I
(4.8)
where the summation extends over a total of N discrete elements of the system. For brievity, the
superscript “1” for the volume V, and the surface S, of the nth element is dropped. However, it
should be noted that all integrals are taken over the current volume and surface of the element in
configuration r ,.
In this finite element model, approximate displacement functions Aui are interpolated in terms
of generalized displacements at a finite number of nodal points of the element. In order that the
functional AII, be defined as the sum of the contributions from the individual elements, the
interpolation functions must satisfy certain compatibility requirements along the interelement
boundaries.
In matrix notation the assumed displacement field within each element may be expressed as

Au=NAv, (4.9)

where Av is the vector of incremental nodal point displacements. The corresponding strain distri-
bution is

AC =BAv . (4.10)

The incremental displacement gradients Au’ entering the second term of eq. (4.8) are obtained by
differentiating eq. (4.9):

Au’ = N’Av . (4.11)

Accordingly, the incremental potential energy of the finite element assemblage may be stated as

AI-I, = c (;Av’(K, +k,)Av - Av’ A01 , (4.12)

n
where

k,= j-B’CBdV (4.13)

Vn

k, = JN%N’~V (4.14)
Vn

Aa = A& + A&- = s N’APdV+ j- N'ATdS . (4.15)

vn s%

In the above equations G is a matrix containing the initial stresses in configuration rl, and AF
and AT are the increments in body forces and boundary tractions, respectively. Moreover, it is
observed that k, is the usual linear element stiffness matrix, k, is the initial stress stiffness matrix
 G. fforrigmoe und P. G. Bergan, Incremental variationd principles 207

(by many authors called the geometric stiffness matrix), and AQ is the incremental load vector.
The stationary conditions are obtained by taking the variations of eq. (4.12) with respect to
AU, i.e.

C{kAv-A&=0, (4.16)
n

where the instantaneous or incremental stiffness matrix k is given by

k = k, + k, . (4.17)

Although a majority of authors using the compatible displacement model favour the total
Lagrangian approach [33], the updated Lagrangian formulation has been employed frequently,
especially in a simplified form for thin structures such as beams, plates and shells. The survey
articles [ 11,12] give extensive references to nonlinear applications of incremental models of this
type.

4.2. Stationary incremental complementary energy and equilibrium models

Assuming the existence of an incremental complementary strain energy function U,, defined by

or 6U,, = Aeii 6AS, , (4.18)

the principle of complemental virtual work, eq. (3.6), may be stated as

6AII, = 0, (4.19)

where the incremental com~~emen#a~ energy functional AII, of the body is defined by

A&= s U,,dV- JAg$AridS. (4.20)

V’ st

Eq. (4.19) represents the incremental form of the principle of stationary complementary energy.
This principle is also subject to subsidiary conditions in the form of fulfilment of the equilibrium
equations (4.5) and the mechanical boundary conditions (4.6).

Equilibrium models
In a finite element application of the conventional principle of stationary complementary energy
the assumed stresses AS, must satisfy the equilibrium conditions throughout the whole region. In
small displacement analysis this is easily achieved by use of either appropriate stress functions
(equilibrium model 1) or polynomial expansions (equilibrium model 2) [ 61. For geometrically
nonlinear problems, however, the assumed stress approach becomes far more complicated. This is
due to the direct coupling between incremental stresses ASjj, initial stresses ajj, and incremental
displacements Aui in eqs. (4.5) and (4.6). To the authors’ knowledge, no applications of equilibrium
208 G. Horrigmoe and P. G. Bergan. Incremental variational principles

models to geometrically nonlinear problems have yet been published.

The authors feel that it may be worthwhile to consider the formulation of the complementary
energy principle in terms of the unsymmetric first Piola-Kirchhoff stress tensor instead of AS,
to avoid the complications described above. For further information the works by Fraeijs de
Veubeke [ 341 and Koiter [35] are recommended.

4.3. Incremental Hu- Washizu principle

The principle of stationary incremental potential energy derived in sect. 4.1 was based on the
assumption that the incremental displacement field Aui satisfies the equations

Ae, = ~(Au,.~ + AUi,i) in V’ > (4.21)

Aui = AU,. onS,’ . (4.22)

These equations may now be treated as conditions of constraint and incorporated in the variatio-
nal expression (4.4) by means of Lagrangian multipliers. By introducing nine Lagrangian multi-
pliers AS, and Ap,., defined in V1 and on Si, the generalized incremental functional may be
written as

AIL,,, = J UA dV + $ Oiif Au,,i Au,,i dV- J AF,AUi dV- JAT,Aui dS

V’ Vl V’ SJ (4.23)

- JAS,[AE~~ - 4(AUi,i + AUi,i)] dV- J Api(Aui - Aii)dS >

V 1
&A

where the independent variables subject to variation are eighteen in number, namely, Aui, Ae,,
AS, and Api with no subsidiary conditions.
The incremental form of the Hu- Washizu variational principle states that the actual solution
is given by the stationary condition of the functional AIL,, i.e.

&AII,=O. (4.24)

It can easily be verified that this leads to the equations

-=
au, AS, in V’ , (4.25)
aar,

Aeii = ~ (AU,i + AUi,i) in V’ , (4.26)

ASii,i + (U,iAUii),k + AFi = 0 in V’ , (4.27)

ATi= AS,vi + o,ivkAUii= ATi onS,‘, (4.28)

Aui = AZ, onS,' , (4.29)

 G. Horrigmoe and P.G. Bergan, Incremental variational principles 209

Api = ASiivi + aMVk AUii = ATi ons,l . (4.30)

4.4. Incremental Hellinger-Reissner principle and mixed models

The generalized incremental functional (4.23) may be reduced to the incremental Hellinger-
Reissner functional AII, if Api are eliminated by use of eq. (4.30), and U,, is introduced through
the equation

U,, = A$ Ae, - U, . (4.31)

Thereby follows

AI-I, = - $ U,, dV+ j- AS&(AU~,~ + Au,,~) dV

Vl Vl
(4.32)
+ $aii;A~k,iA~k,idV- j-AF,Au,dV-JAT,Au,dS- $ATi(Aui-Aii,.)dS.
Vl V’ sli s:

The incremental form of Hellinger-Reissner’s variational principle states that the actual solu-
tion is obtained by requiring

6Al-I, = 0. (4.33)

Carrying out the variations of eq. (4.32) with respect to the field variables AS, and Aui, the
following equations are obtained:

-au*,=;(A++
8AS,
AQ) in V1 , (4.34)

in addition to eqs. (4.27)-(4.29).

Another form of the functional can be deduced by integrating eq. (4.32) by parts:

Any = $U~,dV- $ [AS,, + (~,iA~ij),k + AFil AUi dV

Vl Vl

+ j- (ATi - ATi)Aui dS+ s ATiAiii dS. (4.35)

s: 5:
In the case when the stress equilibrium equations (4.27) and (4.28) are satisfied, the incremental
Hellinger-Reissner functional AII: reduces to the incremental complementary energy AII,.

Mixed models
The incremental strain-stress relationship may be expressed as

Aeii = Dijkr AS,, , (4.36)

210 G. Horrigmoe and P.G. Bergan, Incremental variational principles

where Dijkl are the components of the incremental compliance tensor. Substitution of eq. (4.36)
into eq. (4.32) and rewriting the result in finite element form yields

An, = c - s ;Diik,ASiiASR, dV + s As, ;(AUii + Au,,~) dV

n
Vtl Vn

+ JOijiA~k,i~~k,i dV- j-A&AuidV-j- ATiAuidS , (4.37)

V, V?l s an I

where it has been assumed that

Aui = Aiii on S,+ . (4.38)

In the finite element formulation the stresses ASii and the displacements Au, are assumed to be
independent of each other within each element. These quantities must obey certain continuity
requirements across the interelement boundaries in order to maintain the validity of the functional
AII, of eq. (4.37). Various mixed formulations arise from the use of different continuity require-
ments for combinations of AS,, and Aui. Here, it will now be assumed that the displacements Aui
are continuous. Other possible mixed formulations have been discussed by Pian and Tong [ 6,7].
Mixed finite elements have the extra advantage compared to equilibrium models that the assumed
stresses may violate the equilibrium conditions (4.5) and (4.6). Hence, the restrictions on the as-
sumed stress field as discussed in sect. 4.2 are removed, and the stress interpolation functions are
easily obtained.
As mentioned above, displacements AU and stresses AS are assumed independent within each
element, i.e.

Au = NAv (4.39)

and
AS=AAp, (4.40)

where Ap is a vector of generalized nodal point stresses.

Following the notation of sect. 4.1, the incremental Hellinger-Reissner functional of a finite
element assemblage may be expressed as

AII, = c {-;Ap’fAp + Ap’gAv + ;Av’k,Av - AdA@ , (4.41)

n
where

f= s A’DA dV, (4.42)

Vtl

g= j-A’BdV. (4.43)
V,

The other symbols in the above equations have been previously defined in sect. 4.1.
 G. Howigmoe and P. G. Bergan, Incremental variational principles 211

Setting the variation of All, equal to zero yields

(4.44)
?[[E :A +b :J[:J-!:!~=[:I
or

c {(k; + K;) Av* - Aa* } = o , (4.45)

n

where the asterix is used to distinguish the above mixed method from the standard displacement
method of sect. 4.1. Although the diagonal submatrix of zeros in Ki in eq. (4.44) may necessitate
some modifications of existing solution algorithms, this should not create any serious problems
or substantially reduce the computational efficiency.
The use of the mixed formulations for geometrically nonlinear problems is still at an early stage
of development and only a few applications have been published [ 2 l-231.

5. Incremental variational principles with relaxed continuity requirements - hybrid models

5.1. Modified incremental potential energy and hybrid displacement model I

The difficulty in achieving conforming shape functions for plate and shell elements is one of
the major shortcomings of the conventional compatible displacement model. However, it is per-
missible to modify the incremental potential energy functional by relaxing the interelement con-
tinuity requirements for AZ+. This can be done by adding a jump term to the functional AII, of
eq. (4.8). The modified functional then becomes

“mm = c SSCjk,AEiiAEkl dV+ J OiiiAUk,iAUk,i d’

V?l V,

- $ AF,Aui dV - s ATiAui dS- s ATiAui dS , (5.1)

Vn s sfl
%

where S,, is the interelement boundary of the nth element. The jump term in the above equation
may be interpreted as a condition of constraint for the discontinuous variable Aui. The corre-
sponding Lagrangian multipliers are recognized as the boundary tractions A Ti which now appear
as additional variables in the functional AIIMP1. In general, the resulting finite element equations
contain both displacement and stress parameters as unknowns [ 241.

5.2. Modified incremental potential energy and hybrid displacement model 2

Another modification of the incremental potential energy functional consists of using displace-
ment expansions for the interior of the element that are different from displacements along the
element boundary. Compatibility between these two displacements fields is introduced in a
standard manner by use of Lagrangian multipliers representing the boundary tractions A Ti. Thus,
212 G. Horrtgnwe and P.G. BeEan, Incremental variational principles

the modified functional may be expressed as

AII mPZ= C s ~CijklA~ijAEkldV+ $ aijaAUk,jAuk,j dV

n
VP2 vn

- $ AFiAUj dV- S ATiAU, dS- S ATj(AUj - A~i) dS ) (5.2)

s Cr, I
vn av?Z

where AGi are the boundary displacements and aV, is the total boundary of the nth element, i.e.

av, =s, +sf;n+sn.

In eq. (5.2) Ai& is subject to the condition that

Aclu”,
= Ai+ on S,,, . (5.3)

The stationary conditions of the above functional are obtained by taking the variations with
respect to Au,, AGi, and ATi.
In applying this finite element model, three completely independent fields can in general be
used. By eliminating unwanted variables, AlImp2 may be expressed in terms of nodal displacements
only. However, this operation is rather cumbersome. A substantial simplification is achieved if the
two displacement fields are expressed in terms of the same set of generalized nodal displacements
and the boundary tractions are computed from the displacements within the element.
As a first step, ATj are calculated from the interior displacement field, i.e.

AT, = ASijvj + OkjVkAUj,j on aV,, . (5.4)

Substitution of eq. (5.4) into the modified functional (5.2) yields

- S AF~Au~ dV_ S A~jAui dS- $ (ASijVj + UkjUkAUi,j)(AUj -A~i) dS . (5.5)

V?l son av, I

The interior displacement field AL\U

may now be assumed as

Au=NAv, (5.6)

where the interpolation functions N need not satisfy the interelement compatibility conditions;
they may, in fact, be chosen with a considerable amount of freedom. Further,

Atl=LAv, (5.7)

where the interpolation functions L are applied only to individual segments of the boundary W,.
 G. Horrigmoe and P. G. Bergan, Incremental variational principles 213

Hence, it becomes an easy task to achieve compatible displacements for adjoining elements. The
incremental stresses are computed from eqs. (4.7) and (4. lo), i.e.

AS = CAE = CBAv. (5.8)

Substitution of eqs. (5.6)-(5.8) into eq. (5.5) gives

AI-ImPz = ~{;Av’(k,+k,)Av - A&@], (5.9)

n
where

R, = $ B’CB dV - f (M’(iV-L) + (IV-L)‘M) dS, (5.10)

V, avn

k,= $N”aN’dV- S(N”aB(N-L)+(N-L)‘a,N’)dS, (5.11)

Vn ah

A@=A&+Aij,= JNtAFdVt $ L’ArdS. (5.12)

avn s
an

In the above equations M is a matrix which relates the incremental boundary stresses to the nodal
displacements Av, whereas the matrix ug contains initial boundary stress terms.
From eqs. (5.10) and (5.11) it is observed that this finite element model leads to expressions
for the incremental element stiffness matrix which are closely related to those derived in sect. 4.1
for the conventional compatible displacement model. The main difference is the corrective
boundary terms due to possible gaps between Aui and A&. However, the integrals over V, are
considerably simplified due to the new freedom in selecting interior displacement functions.
Perhaps even more important, eqs. (5.10) and (5.11) may be considered as a basis for improving
already existing nonconforming elements. Kikuchi and Ando [ 25,261 have employed this model
in the nonlinear analysis of plates and shells considering both geometric and material effects.

5.3. Modified incremental complementary energy and the hybrid stress model

The extension of the hybrid stress model to geometrically nonlinear problems is due to Atluri
[ 271, who gives a detailed derivation of the hybrid stress model; therefore, only a brief outline of
this model is given below.
As stated in sect. 4.2, the assumed stress approach has certain inherited drawbacks in large dis-
placement analysis due to the complicated nature of the equilibrium equations. However, the
interelement equilibrium and the mechanical boundary conditions may be treated as conditions
of constraint and introduced in the incremental complementary energy functional by means of
Lagrangian multipliers. Thus,

AS,AS,,dV- s ATiAiiidS- JAT,Au,dS- $ (AT,-ATi)AuidS

sun h %I I
214 G. Horrigmoe and P.G. Bergan, Incremental variational principles

or

j ;Dii~rASiiASkl dV- S ATiAu”i dS+ j ATiAiii dS , (5.13)

Vn ah s
%

where AC, is written for the displacements now appearing over the entire boundary aV,. The quan-
tities subject to variation are the interior stress field AS, and the boundary displacements A&..
As shown in ref. 127 I, the equilibrium equations (4.5) can be satisfied by introducing the dis-
placements Aui within each element as additional variables. The assumed stress field AS is separated
into three parts:

AS=PB+‘P&+EAv, (5.14)

where P p satisfies the homogeneous equations

ASii,i = 0 , (5.15)

and PF#lF is any particular solution corresponding to the prescribed incremental volume forces
AFi. The matrix E results from the initial stress term in the equilibrium equations (4.5). Interior
displacements Au and boundary displacements Aii are interpolated in the same manner as in the
simplified hybrid displacement model discussed in sect. 5.2. The final matrix equations are ex-
pressed in terms of Au only. Formally, they coincide with the standard form of eq. (4.12). See
[27] and [ 281 for examples of nonlinear applications of the hybrid stress model.

5.4. Modified incremental Hellinger-Reissner principle and the modified hybrid stress model

Several possible schemes for relaxing the interelement continuity requirements of the Hellinger-
Reissner variational principle have been suggested by Pian [ 31. Here, such schemes will be dis-
cussed in connection with the incremental formulation of the Hellinger-Reissner principle derived
in sect. 4.4. A complete derivation of a modified hybrid stress model is also given.
In sect. 4.4 it was assumed that displacements were continuous across the interelement bounda-
ries. When this condition is removed, additional surface integrals must be added to the functional
of eq. (4.37) to account for the discontinuities in the derivatives of Aui across the boundaries.
This approach is analogous to the modification of the incremental potential energy functional dis-
cussed in sect. 5.1.
Another possibility is to relax the interelement compatibility conditions by use of separate
variables for displacements A ui within the element and displacements A Gi along the element
boundary. This approach has already been discussed in sect. 5.2 in connection with hybrid dis-
placement model 2. The modified Hellinger-Reissner functional becomes

A%R = F - J ;Diik,~SiiAsk, dv+ j ~s~$fh+ + AU,~) dv

V, V,
(5.16)
+ s o$Au, i
Auk i dV- jAFiAuidV-
Vn
j A&A&dS- jATi(Aui-A&)dS
ah
vn ‘%
 G. Horrigmoe and P. G. Bergan, Incremental variational principles 215

where the variations are to be taken with respect to Aui, AS,, AU,., and ATi. However, the large
number of variables involved makes this form impractible. An obvious simplification is to elimi-
nate ATi by means of eq. (5.4). Moreover, integration of the second term of the above equation
by parts leads to

AI-L2= Cn - J fDijk,ASijASkl dV - J AS,jAUi dV + J Oij~AUk,iAUk,i dV

V, VP3 V, (5.17)
- s aFiAui dV - J AT;AUi dS+ J ASiiViAii; dS- J OkjVkAUi,j(AUi- Ani) dS .
vn s avn ah
on

In applying this model, the incremental stresses are expressed in terms of a set of unknown
parameters fI,

AS=Pfl. (5.18)

The interior displacement field Au and the displacements Ati along the element boundary are as-
sumed in the same manner as in sect. 5.2; that is, they are given by eqs. (5.6) and (5.7), respec-
tively. With these assumptions, the functional AII’,, of eq. (5.17) becomes

WilR = q {-; flfHfl + fl’(G-S)Av + ;Avfk,Av - Av’A@ , (5.19)

where

H= JP’DPdV, (5.2Oj
V,

G= s R’LdS, (5.2 1)
ah

S = JP”NdV, (5.22)
vn

and k, and A0 have been previously defined by eqs. (5.11) and (5.12). In the above equations
the matrix P’ contains derivatives of the assumed stresses, and R is a matrix relating incremental
boundary stresses to the parameters fl.
The incremental stress field is assumed independent for different elements; hence, the stationary
condition with respect to fl yields

fl= H-‘(G-S)Av . (5.23)

With the help of the above equation p is eliminated from eq. (5.19). Thus,

ATnR= ~(;Avt(k,+k,)Av - AvfAijl , (5.24)

216 G. Horrigmoe and P.G. Bergan, Incremental variational principles

where the element stiffness matrix k, is given by

k, = (G-S)'H-'(G-S) . (5.25)

It is worth noting that a substantial simplification may be achieved by making the incremental
stresses of eq. (5.18) satisfy the homogeneous equilibrium equations (5.15). In this case S becomes
identical to zero, and eq. (5.25) is reduced to

k, = G*H-'G , (5.26)

which coincides with the linear stiffness matrix of the hybrid stress model (see [ 71). The reason
for calling this approach a modified hybrid stress model thereby becomes apparent. Pirotin [ 161
has applied this model to the large deflection analysis of arches and cylindrical shells.

6. Conclusion

In the present paper the major variational principles of solid mechanics have been formulated
for nonlinear problems by considering incremental deformations of a continuum. The principles
treated are the incremental forms of the principle of stationary potential energy, the principle of
stationary complementary energy, the Hellinger-Reissner principle and the general Hu-Washizu
principle. It has been demonstrated how these principles can be adopted to the finite element
method and how associated finite element models arise. The principles mentioned can also be
modified by relaxation of compatibility requirements between adjoining elements. Hybrid ele-
ments resulting from such modifications are also discussed. A consistent notation has been used
throughout the mathematical treatment of the various nonlinear finite element models. Similar-
ities and differences between the various models are pointed out.

References

[ 11 J.H. Argyris and S. Kelsey, Energy theorems and structural analysis (Butterworth, London, 1960). (Originally published in
a series of articles in Aircraft Engineering during 1954-1955.)
[2] B. Fraeijs de Veubeke, Displacement and equilibrium models in the finite element method, in: O.C. Zienkiewicz and G.S.
Holister (eds.), Stress analysis (Wiley, London, 1965) 145-197.
[3] T.H.H. Pian, Finite element methods by variational principles with relaxed continuity requirements, in: C.A. Brebbia and
H. Tottenham (eds.), Variational methods in engineering (Univ. Southampton, 1972).
[4] T.H.H. Pian, Formulations of finite element methods for_solid continua, in: R.H. Gallagher, Y. Yamada and J.T. Oden (eds.),
Recent advances in matrix methods of structural analysis and design (Univ. Alabama Press, 1971) 49-83.
[5] T.H.H. Pian,Variational formulations of numerical methods in solid continua, in: G.M.L. Cladwell (ed.), Computer aided
engineering (Univ. Waterloo, Waterloo, Canada, 1971) 421-448.
[6] T.H.H. Pian and P. Tong, Finite element methods in continuum mechanics, in: C.S. Yii (ed.), Advances in applied mechanics
12 (Academic Press, 1972) l-58.
[7] T.H.H. Pian and P. Tong, Basis of finite element methods for solid continua, Int. J. Numer. Meth. Eng. 1 (1969) 3-28.
[8] S. Atluri and T.H.H. Pian, Theoretical formulation of finite-element methods in linear-elastic analysis of general shells,
J. Struct. Mech. 1 (1972) 1-41.
[9] R.H. Gallagher, General potential energy and complementary energy models based on stress parameters. Mixed variational
principles and hybrid formulations. Lectures given at ICCAD Course on Advanced Topics in Finite Element Analysis,
Santa Margherita, Italy, June 3-7,1974.
 G. Horrigmoe and P.G. Bergan, Incremental variational principles 211

[lo] J.P. Wolf, Generalized stress models for finite-element analysis. Doctoral Thesis, Institut fur Baustatik, ETH, Zurich, 1974.
[ll] J.T. Oden, Finite element applications in nonlinear structural analysis, in: W.H. Rowan, Jr. and R.M. Hackett (eds.), Appli-
cation of finite element methods in civil engineering (School of Engineering, Vanderbilt Univ., 1969) 419-456.
[ 121 H.C. Martin, Finite elements and the analysis of geometrical nonlinear problems, in: R.H. Gallagher, Y. Yamada and J.T.
Oden (eds.), Recent advances in matrix methods of structural analysis and design (Univ. Alabama Press, 197 1) 343-381.
(131 K. Washizu, Variational methods in elasticity and plasticity (Pergamon Press, London, 1968).
[ 141 K. Washizu, Some remarks on basic theory for finite element methods, in: R.H. Gallagher, Y. Yamada and J.T. Oden (eds.),
Recent advances in matrix methods of structural analysis and design (Univ. Alabama Press, 197 1) 25 -47.
[ 151 T.H.H. Pian and P. Tong, Variational formulation of finite-displacement analysis, in: B. Fraeijs de Veubeke (ed.), High speed
computing of elastic structures (Umv. Liege, Belgium, 1971) 43-63.
[ 161 SD. Pirotin, Incremental large deflection analyses of elastic structures, Ph.D. Thesis, Dept. of Aeronautics and Astronautics,
Massachusetts Institute of Technology, ‘971.
[ 171 P. Tong, ST. Mau and T.H.H. Pian, Derivation of geometric stiffness and mass matrices for finite element hybrid models,
Int. J. Sol. Struct. 10 (1974) 919-932.
[18] D.J. Allman, Finite element analysis of plate buckling using a mixed variational principle, Proc. third Conf. on Matrix Methods
of Structual Mechanics (Wright-Patterson Air Force Base, AFFDL-TR-71-160, 197 1) 683-705.
[ 191 F. Kikuchi and Y. Ando, Simplified hybrid displacement nethod applied to plate buckling problems, J. Nucl. Sci. Technol.
9 (1972) 497-499.
[20] W. Altman and F. Venancio-Filho, Stability of plates using a mixed finite element formulation, Computers & Structures
4 (1974) 437-443.
[21] J.J. Connor and G.T. Will, A mixed finite element shallow shell formulation, in: R.H. Gallagher, Y. Yamada and J.T. Oden
(eds.), Recent advances in matrix methods of structural analysis and design (Univ. Alabama Press, 1971) 105-137.
[22] R. Harbord, Berechnung von Schalen mit endlichen Verschiebungen, Gemischte finite Elemente. Doctoral Thesis, Univ.
Braunschweig, Germany, 1972.
[23] J. Backlund, Finite element analysis of nonlinear structures. Doctoral Thesis, Department of Structural Mechanics, Chalmers
Institute of Technology, Gothenburg, Sweden, 1973.
[24] M. Link, Zur Berechnung von Platten nach der Theorie II. Ordnung mit Hilfe eines hybriden Deformationsmodells,
Ing.-Archiv. 42 (1973) 381-394.
[25] F. Kikuchi and Y. Ando, Application of simplified hybrid displacement method to plate and shell problems. Preprints of
the Second Intern. Conf. on Structural Mechanics in Reactor Technology, Berlin, lo-14 September, 1973.
[ 261 F. Kikuchi and Y. Ando, Application of simplified hybrid displacement method to large deflection analysis of elastic-plastic
plates and shells, J. Fat. Eng., Univ. Tokyo (B) 32 (1973) 117-135.
[27] S. Atluri, On the hybrid stress finite element model for incremental analysis of large deflection problems, Int. J. Sol. Struct.
9 (1973) 1177-1191.
[28] A.L. Deak and S. Atluri, Nonlinear hybrid stress finite element analysis of laminated shells, Proc. Intern. Conf. on Computa-
tional Methods in Nonlinear Mechanics (Univ. Texas, Austin, Sept. 23-25, 1974).
[29] R. Hill and J.R. Rice, Elastic potentials and the structure of inelastic constitutive laws, SIAM J. Appl. Math. 25 (1973)
448-461.
(301 G. Horrigmoe, Thesis, Division of Structural Mechanics, The Norwegian Institute of Technology, Trondheim, to be
published.
[31] J.T. Oden, Finite elements of nonlinear continua (McGraw-Hill, 1972).
[ 321 Y .C. Fung, Foundations of solid mechanics (Prentice-Hall, 1965).
[33] J.H. Argyris, P.C. Dunne, T. Angelopoulos and B. Bichat, Large natural strains and some special difficulties due to non-
linearity and incompressibility in finite elements, Comp. Meth. Appl. Mech. Eng. 4 (1974) 219-278.
[34] B. Fraeijs de Veubeke, A new variational principle for finite elastic displacements, Int. J. Eng. Sci. 10 (1972) 745-763.
[35] W.T. Koiter, On the principle of stationary complementary energy in the nonlinear theory of elasticity, SIAM J. Appl. Math.
25 (1973) 424-434.