1590 AIAA JOURNAL VOL.

10,NO.2
Analysis of Flow of Viscous Fluids by the Finite-Element Method
TINSLEY ODEN* CARTER WELLFORD JR.-J

J.

ANDL.
University of Alabama, Huntsville, Ala.

General finite-element models compressible incompressible fluid flow derived. These involve local

of

and

are
approximations of the velocity field, the density, and the temperature for compressible fluids and the velocity,
temperature, and pressure for incompressible fluids. Theories of local solenoidal approximations and mixed finite-
element models for compressible flow are derived. A number of computational schemes are developed for the
numerical solution of both transient and steady nonuniform flow problems involving incompressible fluids. Numerical
results obtained from several test problems given. shown that finite element method great potential

are

Itis

the

has
for use in flow problems, and represents a powerful new tool for the analysis of viscous flows.

Introduction of fluid under consideration. Effectively, develop finite

the

we
element analogues of the equations of continuity, linear momen-
T present paper concerned with application tum, energy arbitrary fluids. models obtained from
HE

is

the

ofthe

and

of

The

are
concept finite elements formulation solution local approximations of the density, velocity, and temperature
of

tohe

and

of a wide range of problems in fluid dynamics. The method is fields in each element and represent generalizations of those
sufficiently general to treat a variety of unsteady and nonlinear proposed earlier.7'9
flow phenomena in irregular domains. An intrinsic feature of In addition, we treat the problem of fluids characterized by
finite-element approximations that mathematical model equations state which thermodynamic pressure
is
a

is

of

in

the

isnot
generated patching together number purely "local" given explicitly function density, temperature,
by

of

approximations of the phenomena under consideration. This

as

ofthe

and
velocity gradients. There we develop mixed finite-element models
aspect of the method effectively frees the analyst from traditional by approximating locally mean stress thermodynamic

the

(or
difficulties associated with irregular geometries, multi-connected pressure) each element obtain general model which
domains, and mixed boundary conditions. Moreover, applications
in

We

in
equation state satisfied average sense over each
firmly rooted physics problem hand
the

of

is

ina
element. then consider special important case
are

inthe

ofthe

at

and

preliminary studies indicate that, for a given order of accuracy,

We

the

but

of
viscous incompressible fluids, with emphasis on isotropic New-
the resulting equations better conditioned than those obtained tonian fluids with constant viscosities. There we address ourselves
are

by, say, finite difference approximations of the governing dif- certain problems connected with imposing continuity
ferential equations.1
to

the
equation (incompressibility condition) in the discrete model and
Certain underlying ideas finite element method to special boundary conditions. A notion of solenoidal finite
ofthe

ofthe

were discussed 1943 Courant.2 However, formal element fields is introduced. We then describe computational
in

by

the

presentation method generally attributed 1956 schemes for the solutions of the equations governing the model
ofthe

is

tohe

paper of Turner et al.3 While the method has found wide applica- for uniform steady flow, nonuniform steady flow, and unsteady
tion in solid and structural mechanics,4 its application to flow flow of viscous fluids. Numerical results obtained from applica-
problems come only rather recent times. Early uses tions number representative example problems
has

in

ofthe

toa

of

are
method were always associated with variational statements presented.
ofthe

problem under consideration, that natural that steady,

so
its

potential flow problems were first solved using finite

the
tobe

elements. We mention, in this regard, the works of Zienkiewicz, Finite Element Models of Fluid Flow
Mayer, and Cheung5 on seepage through porous media and
Martin6 potential flow problems. Finite element models To fix ideas, consider the motion of a continuous medium
on

of

unsteady compressible and incompressible flow problems were through some closed region R of three-dimensional euclidean
obtained Oden.7"10 Applications finite element methods space. We establish in -R a fixed inertial frame of reference defined
by

of

to a number of important problems in fluid mechanics have been by orthonormal basis vectors it = 1,2,3). spatial coordi-
(i

The

reported recent years; among them, mention work nates place denoted xt components
ofa

PinRare

andthe

of
in

we

the

of

Thompson, Mack, and Lin11 on steady incompressible flow and velocity of the medium at P at time t are denoted v^x^, x 2 , x3, t) =
Tong,12 Fujino,13 Argyris al.,14~16 Reddi,17 Baker,18 vt(x, t). The density and the absolute temperature at place P at
et

and

Herting, Joseph, Kuusinen, MacNeal19 various special time t are denoted p(x, t) and 6(x, t\ respectively. If T0 denotes a
uniform temperature at some reference time t0, we may use as an
and

on

incompressible flow problems. recent book Zienkiewicz4

alternate temperature measure the temperature change T(x, t) =
The

of

can be consulted for additional references.

present investigation, extend finite element 0(x,t)-TQ.
To construct a finite element model of the fluid, we replace R
Inthe

we

the

method to general three-dimensional problems of heat conduc-

tion flow compressible incompressible fluids, wherein by domain ^consisting finite number subdomains re,
a

ofa

Eof

so that R R = uf = t re. The subregions re are the finite

and
of

and

no restriction is placed on the constitution or equation of state

elements; they are generally of simple geometric shapes and are
designed so as to represent a good approximation of R when
Received December 1,1971; revision received May 24,1972. Support appropriately connected together. geometry each finite
of this work by the U.S. Air Force Office of Scientific Research under
The

of

element re is characterized by a finite number Ne of nodal places

Contract F44620-69-C-0124 is gratefully acknowledged.
Index categories: Viscous Nonboundary-Layer Flows; Boundary (the number may vary from element to element) and the nodes of
Layers and Convective Heat Transfer-Laminar. typical element defined spatial coordinates xfj e) ;
a

ear

bythe

* Professor and Chairman, Department of Engineering Mechanics. N = 1,2,..., JV_e; i = 1,2, 3; e = 1,2, ...,. The global finite
Member AIAA. element model R is obtained by connecting the E discrete elements
t Graduate Student, and Engineer, Teledyne Brown Engineering. at appropriate nodal points by means of simple incidence
 DECEMBER 1972 FLOW OF THE VISCOUS FLUIDS BY THE FINITE ELEMENT METHOD 1591

mappings which merely identify the desired correspondence Momentum Equations for a Finite Element
between local global nodal labels. These mappings
equations governing motion typical finite element
and

are
described elsewhere20 and need not be discussed here. The

The

the

ofa
important feature of the model is its local character; that is, the obtained constructing Galerkin integral Cauchy's

canbe

by

of
behavior of the medium can be idealized locally in a typical first law of motion over the element and by using the velocity
element independent of the behavior in other elements in the interpolation functions \l/N(x) weight functions this integral.

as

in
If this approach taken, linear momentum balanced
model independent ultimate location model.

is

is

ina
average sense over element. arbitrariness choice
and

ofits

inthe
final global model then obtained routinely connecting

the

The

ofthe
of \l/(x) as weight functions, however, is removed if an alternate
The

is

by
elements together through mappings which depend only upon
but equivalent approach based energy balances employed.9
the topology of the model.

on

is
The behavior of the medium must be consistent with the
Following guidelines provided by the notion of determinism,22 principle conservation energy:
we shall take fundamental dependent variables, velo-

of

of
asour

the
city, mass density, temperature change obvious measures D/Dt(K + U) = Q + Q (7)
and

as
of these primitive characteristics. Therefore, consider a typical where K is the kinetic energy, U is the internal energy, Q is the
finite element re with Ne nodes isolated from the global mechanical power, and Q is the heat:
model R. v(x, p(x, T(x, denote global approxima-
= ^1 f pVjVjdv f
Let
t),
t),and

t)

tions of the velocity, density, and temperature change to be K U psdv

U=

=
determined time r, v(e), p(e), T(e) denote their Jv Jv

Q = f pFjVjdv + f SjVjdA r r q/ijdA

atxsR

andlet

and

restrictions to element re. Then finite-element approximations of (8)

these restrictions constructed which form Q= \ Phdv +
Jv JA Jv JA
are

areofth

ejW (la) e internal energy density, Fj components

InEq.(8)
isthe

areth
P(e) e)t of the body force vector per unit mass, Sj are components of the
=

T(e) = $\x)T}(t) (Ic) surface tractions, h is the heat per unit mass supplied from
Here henceforth repeated nodal indices summed from internal sources, q} are the components of heat flux, and rc, is the
normal boundary surface. Noting that material deriva-
and

the

are

1 to Ne; v(^(t), p"e}(t\ and T"e)(t) are the values of the velocity,
tohe

the
density, relative temperature node element re tive of the kinetic energy for element e is
and

at

Nof

at

time r; i.e.
DK/Dt =

etc. functions ^(x), <p(ff(\\ ffi(x) are local interpolation and introducing Eq. (5) and (la), we obtain
The

and

functions defined so as to have the properties (D/Dt)K = LcMNLpMvLk + (9)

(3) where

Notice that Eq. (1) implies that different forms of the interpolation C
MNL Jv (10)
functions may be used to approximate different local fields over
the same finite element. In certain cases (some of which are to be and M, N, L, F = 1,..., Ne. The first quantity in parenthesis
discussed later), this may require that certain of the functions represents local inertial force node /cth direction,
the

at

Ninthe

vanish certain nodal points that indices each member and the second term in the parenthesis represents the convective
at

or

Nin

of Eq. (1) may have different ranges. Note also that "higher order" inertial force at node N in the /cth direction.
local representations obtained also specifying values Since the local form of the energy balance is
canbe

by

of derivatives of v, p, and T at the nodes. p(Ds/Dt) = ps + pvfj = Tikvki + qkk + ph (11)

We must also remark that, in the case of incompressible fluids,
our formulation requires that, instead of the density p(e)(x, t) we wherein Tik is the Cauchy stress tensor, the material derivative
approximate the pressure field p(x, t) over R. Thus, if p(e)(x, t) is of internal energy element written
the

forthe

canbe

as

the restriction of p(x, t) to re, we assume

DUJDt (12)
=

P(e) = (4)

interpolation functions ju also obey Eq. (3). where Qe = Ve(qkk + ph)dv is the heat of the element. For an
The

element of fluid of volume ve and surface area Ae, the power of

Mechanics of a Finite Element external forces
is

Kinematics

With local velocity field given (la), relevant

where
[L CRNkpR vNk (13)
the

byEq.

al

kinematical quantities associated with motion element

the

ofthe

are determined by the nodal velocities \N(t). For example, the

components of local acceleration are
a, = DvJDt = dvjdt + vimvm = ^ 4- $ *,&*$*% (5) first second quantities parenthesis represent
The
and

inthe

the

where v? dv*(t)/dt, commas denote partial differentiation force at node N in the /cth direction due to the surface stress
with respect to the spatial coordinates (i.e. \I/N m = d\l/N(\)/dxm\ distribution force node /cth direction
andthe

at

Ninthe

dueto

the repeated indices are, again, summed over their admissible body force, respectively. define generalized force
the

Ifwe

the

ranges (i,m = 1,2,3; M, N = 1,2, . . . ,JVJ, and the element pNk according to
identification label (e) has been dropped for simplicity. Likewise,
the models of the rate-of-deformation tensor dtj and the spin PNk = + CRNkpR (14)
Ae
tensor w/7 are given by
then generalized forces develop same amount mechan-
the

the

of

ical power external forces continuum element,

asthe

inthe

and

local vorticity field co, co, = ^Nk^ where s &e PNkVk

andthe

is

isthe

permutation symbol. Local approximations various other Finally, introducing Eqs. (9), (12), and Qe into (7), we obtain an
of

kinematical quantities can be calculated in a similar manner. energy equation typical finite element Then, making
fora

e.

the
 1592 ODEN WELLFORD AIAA JOURNAL

J.T

ANDL.C

JR.
argument that this result must hold for arbitrary values of the Constitutive Equations, Equations of State,
nodal velocity vk, we obtain as the general equations of motion and Mixed Models
(momentum) for a compressible fluid element:
Equations (15), (18), (22) (23)] describe general

and
[orEq.

the
rikilsNidv - pNk = Q (15) equations motion, continuity, energy heat conduction)

of

and

(or
v ye of typical element finite-element model arbitrary

ina

ofan
The constitutive equation for stress in terms of the approximate fluid. To apply these equations to a specific fluid, it is necessary to
velocity pressure expressions must introduced complete eliminate a, Tip.qk, and possibly Y\ by introducing appropriate
and

be

to
this equation. constitutive equations which uniquely define these functions in
terms v (e) , p(e) T(e). This customary procedure

of

and

isthe

in
The Continuity Equation for an Element finite-element formulations.
The local form of the continuity equation is In the case of compressible fluids, however, the mean stress or
thermodynamic pressure often appears implicitly

the

inthe
equation state fluid, impossible

of

ofthe

anditmybe

orim-
where = dp/dt. Introducing Eqs. (la) (Ib) into (16) yields practical to obtain Ttj explicitly as a function of p, r, p and T.
p

and

at point x the residual
re(x) = (pN(x)pN + ((pN(x)i//R (x))kpNv%
We guarantee that residual vanishes average sense
can
In such cases, we propose that a "mixed" finite-element formula-
(17) tion be used, the basis of which is now to be described.
Consider a class of fluids described by constitutive equations
for stress tensor form

the

ina

the

ofthe
over the element by requiring that it be orthogonal (with respect
inner product #> Vefgdv) subspace spanned Tu = dun(p, drs, T} + 7^(p, drs. T) (25)
tohe

</,
=

tohe

by functions <pM(x). Then <r e , (pMy = Vere(pMdv Here n is the so called thermodynamic pressure and ffj is the
the

=0,and

we obtain the finite element model of the continuity equation: dissipative stress. Generally T^ given explicitly function

is

as
aMNpN + bkMNRpv* = 0 (18) of drs, (e.g., class Stokesian fluids, fu 2^dip
p,
andT

fora

of

=
H being viscosity); however, defined implicitly
Here M,N,R 1, 2,..., Ne aMN y^NR denote local

the

nis

byan
equation state :
=

and
and

the

arrays
of
F(TC, dip T) (26)

p,
=0
<PM<PNdv, (19) Assuming that cannot choose eliminate

we

(ordnt

to)

n
from (25) (26) propose mixed finite-element
Eq.
byuseofEq.
we

a
model which restriction ;re(x, f) element re assumed
in

the

ofnt

is
Energy Equation for an Element to be of the form
By introducing (la) into local form first law, ne = ftv(x)7rA'(r) (27)
Eq.

the

ofthe

(11), obtain residual The interpolation functions /?N(x) have properties similar to
Eq.

we

the

re = <PRPR + (PRpR^N^Sti - Tik\ljNivNk - qkk - q>RpRh (20) (3). Introducing (27) into (26), obtain residual
Eq.

Eq.(1)and

we

a
Here it is understood that e, Tik, and qk are functions of the local rc*. Then, the condition <r*,/? N > = 0 leads to the Ne local
fields defined Eqs. (1). before, require that residual equations
in

As

we

the

orthogonal subspace spanned #M(x) functions: N

be

tohe

bythe

) dv = 0 (28)

-i dv = Q (21)
Equation (28) represents finite-element analogue
the

ofthe
Introduction of Eq. (20) into Eq. (21) gives the general finite equation state, (26), insure that satisfied
of

Eq.

and

its

ina

element analogue energy equation: weighted average sense over each finite element. Introduction
ofthe

of (27) into (25) incorporation result into Eqs. (15),

Eq.

and

ofthe

f f (18), (22) (23)] (along with constitutive equations

<PRWM<IVPR +
and
[orEq.

for

Jve Jv Jve qk, e, and /? as functions of p, T, and vt j) yields a complete system

of equations nodal values velocity, density, temperature,
inthe

of

+ (22) and thermodynamic pressure.

Ae

In order to obtain a useable expression for the energy equation

constitutive equations must be introduced for stress Tik, internal
Incompressible Fluids
energy density e, heat flux qk, and heat produced by internal We shall now consider purely mechanical behavior of incom-
sources pressible fluids. Here two principal considerations are involved:
h.

We remark that an alternate equation of heat conduction can 1) motions volume-preserving stress tensor
obtained element rewriting (11) terms
al

are

and2)the

is

not completely determined by the motion. The first condition

be

forthe

by

Eq.
in

ofthe

entropy density rj(x, internal dissipation cr(x, reveals that known constant continuity equa-
t)andhe

t)=

pOq qkk these choices variables, procedure

pisnowa

andthe

tion reduces incompressibility condition

ph.For

of

similar to that used to obtain Eq. (15) yields the general equation
tohe

of heat conduction for an element, div dn = t'i i (29)

v=

=0

The second consideration suggests that n = p, p being the

M
< V+ hydrodynamic pressure, and that the local pressure of Eq. (4)
Jve
Jv e
should selected unknown place
be

asn

in

ofp.

= qN + aN (23) Considering now p to be known and following essentially the

Here qN and GN are the generalized normal heat flux and the same procedure used obtain (15), obtain general
to

Eq.
we

forthe

generalized internal dissipation node equation of motion of incompressible elements

at

h%Ndv q^^dA aN = (24) ^k + + hMNkpN + = PMk

=

Jve jAe Jve

Again, procedure used derive (23) equivalent where m and n^ are the mass and convected mass "matrices"
the

to

Eq.
is

to

Galerkin's method specific forms obtained when respectively, hMm array pressure coefficients, pm
and

canbe

isan

of

and

constitutive equations Y\ qt furnished. are the components of the generalized force defined by Eq. (14)
for
and
are
 DECEMBER 1972 FLOW OF VISCOUS FLUIDS BY THE FINITE ELEMENT METHOD 1593

-i
(31)

derivation of. the incompressibility constraint best

The

can
be
cast in physical terms: Observe that the work done by the pressure
p due to a change in volume dv is clearly peV \dv = pev\ej dv.
pe perform work since incompressible flow
But
can

no

for

Vvis
zero. It follows that for a finite element of an incompressible fluid,

= 0

This relation must hold arbitrary pM-; thus finite element

for

the
model for the incompressibility condition is,
Fig. 1 Finite-element model for the calculation of Couette flow.
^" = r, MN HM\I/N t (32)
0;

J ve
quantity pNk generalized nodal force prescribed

The

isthe

dueto
We note that the energy equation, Eq. (22), is also modified pressures boundary.

ponthe
slightly for incompressible elements due to the fact that (pNpN is Similar procedures used various other special

canbe

for
replaced presumedly known value fik depends boundary conditions. example, uncommon
now

byits

and

For

itsno

tousea
pN. These alterations straightforward need a condition at the impermeable boundary with zero velocity the
on

are

and

notbe

displayed here. specification of the pressure gradient as a function of the body

force acceleration (i.e., n = pFn pan where sub-
andthe

the
script refers normal direction wall). This boundary
Incompressible Newtonian Fluids
n

tohe

tohe
condition can either be applied by retaining the momentum
The special case of an incompressible Newtonian flow with equation at the nodes in contact with the fixed wall or by applying
constant viscosity special interest because number discrete version nodes contact with wall. first
a

athe

in

the

The
isof

of

classical solutions to various flow problems are available for method, which followed herein, statisfies boundary con-
is

the
comparisons. Of course, since the viscosity is constant, the energy dition in an average sense over the boundary elements. The
equation is uncoupled from the momentum and continuity second method satisfies the boundary condition exactly at the
equations and they may be solved separately. For the Newtonian wall nodes.
fluid the dissipative stress tensor is simply fik = 2^dik, \JL being the first method obtained directly from discrete momen-
The

is

the
viscosity. Thus, finite element, fik fj[\l/N kViN ijsN p^ tum equation, Eq. (33), by setting vk = 0 (for boundary nodes N)
forthe

and Eq. (30) reduces to and fik = 0, since fik depends only on vk there. Then
v hMNkpN ZMNV? w^ pMk
-I pnk\l/MdA
+

v?=

(33) + hMNkPN =

where where fMk = |Ve pFk\l/Mdv is the generalized force at node M due
and wlMN, _f (34)
body forces Fk. Transforming surface integral
tohe

the

viathe
MN Green-Gauss theorem collecting terms, arrive
Jv e
and

we

discrete analogue pressure-gradient boundary condition, athe

To these equations we add the incompressibility conditions,
ofthe

Eqs. (32), form determinate set. 9MNkPN fMk ~ "iMNVk (36)

=
to

wherein g J \l/^k
=

dv.

In the second method, a discrete version of this type of boun-

Boundary Conditions
dary condition obtained element merely intro-
canbe

for

eby

If either the total stress or the velocity is specified at the boun- ducing Eq. (4) into the local statement of the boundary condition
dary finite element model, particular difficulties and evaluating the result at the coordinates of each of the wall
ofa

no

are

encountered; the prescribed stress is introduced directly into nodes. While this leads cruder approximation, neverthe-
toa

its

Eq. (14) to obtain consistent generalized nodal forces, and the less much easier to apply in actual calculations.
nodal velocities prescribed satisfy "nonslip" boundary
are

to

the

condition at a fixed or moving wall or the specified velocity Solenoidal Approximations

distribution on other surfaces. In this respect, the procedure
differs very little from that employed in the finite element analysis The pressure term h^ eliminated from discrete
canbe

the

of solids. However, in fluids, specification of the stress at a momentum equation Eq. (33) by constructing solenoidal finite-
boundary may not uniquely determine the pressure; moreover, element approximations of the local velocity. We accomplish this
the boundary conditions may represent constraints on the by introducing called "bubble functions" af(x) which vanish
so

pressure or its gradient rather than the total stress. In these boundaries each element which satisfy least
onthe

of

and

(at

situations, it may be necessary to develop special analogues of approximately) condition

the

the boundary conditions. a M = -*NtM (37)

Consider, example, generalized force (14)
Any particular integral (37) vanishing dRe form
for

the

ofEq.
inthe

case incompressible fluid. Ignoring body force term

ofEq.

on
isofthe
ofan

the

temporarily, observe that a;(x) = a/N(jc)^ (38)

we

f f incompressibility condition, (29) satisfied locally

The

Eq.
isnow

PNk = Tikn^NdA = (-pdik by local velocity approximations form

ofthe

JAe JAe

or
vfa t) = ^Njt(x)v^, $NJt = cVWx) + a/N(x) (39)
The remaining terms in Eq. (33) are altered accordingly.
= PNk + I Tikniil/NdAi pNk = - pnk\l/NdA (35) local solenoidal velocity fields makes possible
Theusof

it

to

JAe jAe eliminate hydrodynamic pressure term from discrete

the

the
1594 J. T. ODEN AND L. C. WELLFORD JR. AIAA JOURNAL

EXACT SOLUTION
VELOCITY (in/sec.)
0.5 .10 FINITE ELEMENT
SOLUTION

Fig. 2 Velocity profile at various

time points for transient Couette
flow.

r
T T T T
t^.OOOIOsec. t=.OO042sec. t=OOI67sec. t*OO376sec t=OO669sec. t=oosec.

momentum equations, Eqs. (33). Thus momentum equations of this type. With minor changes, these techniques could be
the

can be solved directly for the local velocity fields. Then the applied to three-dimensional incompressible flow problems or,
pressure calculation performed independently based in general, isothermal incompressible viscous fluid.
toany
canbe

on

the computed velocity field. Consider the solution of the system of equations consisting of
The pressure must satisfy Poisson's equation. momentum equation, (33), continuity equation,
the

Eq.
andthe
Eq. (32). Three cases can be analyzed based on the properties of
Pii = T^x, itf) pFit (40)
the flow: steady uniform flow; steady nonuniform flow; and
+

-pan

where 7}/x, v*) is the dissipative part of the stress tensor. This transient unsteady flow. models developed earlier lead
can be verified by taking the divergence of the local momentum
or

The
to systems of linear algebraic equations in the first case, non-
equation.22 solution this equation means finite linear algebraic equations in the second, and nonlinear dif-
The

of

by

of

elements very well documented.4 ferential equations third.

is

inthe

Numerical Techniques Steady, Uniform Linear Flow

In order obtain preliminary estimate applicability steady flow problem, term mMNv* con-
Forthe

the

andthe

vective term n^NPv^ in Eq. (33) vanish and the linearized

to

ofthe

of the foregoing theory, we consider a series of classical problems

in two-dimensional incompressible Newtonian flow. Several momentum equation takes the form
numerical techniques developed herein handle problems hMNkPN ZMNI% vO? pMk (41)
are

to

.08T

.07-

.06-

.04-
Fig. 3 Time history of the ;c velocity
components at nodes 7, 8, and 9 for
transient Couette flow.
03-

.02-
NODE 7'

.01-

C/ .obi .002 .003 .004 .005 .006 .007 .008 .009 .010 .012
TIME (SEC.)
.01-
 DECEMBER 1972 FLOW OF THE VISCOUS FLUIDS BY THE FINITE ELEMENT METHOD 1595

.001 .002 .003 .004 .005 .006 .007 .008 .009 .010 .Oil .012

Fig. 4 Time history of y velocity component at nodes 7 and 9 for transient Couette flow.

Equations (41) and (32) represent, for the two-dimensional case, r

MNF = jvc dv. Thus, we obtain the system of non-
a set of 3Ne algebraic equations in 3Ne unknowns. Taking note linear equations
of fact that r^ h express these equations
frk + hMNkPN + + " = pMk (43)
the

=
wecan

in matrix form as
A natural choice of a technique for solving Eqs. (43) is the
incremental loading method used nonlinear structural

in
Z
MN
2
tN n
L
MN2 [fc_ 2 ] (42) mechanics.20 This conclusion is based on the observation that if
1
r'NM 'NM
r 2
u O] density p is assumed to be the loading parameter, when the
0

density equals zero the set of equations reduces to the stationary

_

coefficient matrix square symmetric. addition, linear system, Eq. (42), which can be solved using methods
The

is

and

In

upon assembling the elements, the global form is sparse and described above.
banded. Boundary conditions must be applied in accordance with
This solution process can be described concisely in vector
previous discussion resulting equations notation. Suppose that the collection of momentum equations
the

andthe

setof

canbe

solved for the velocity and pressure variables using standard

and continuity equations are expressed in vector form as
Gauss-elimination codes.
f(v, p) = 0 (44)
Steady, Nonuniform FlowThe Method Incremental Densities
where v is the vector of unknown velocity components and
of

The momentum equation for steady nonlinear flow can be pressures density parameter. solution
andpisthe

Letvba

obtained from (33) setting local inertial term mMNvkN vector corresponding particular value density
toa

of

p,andlet
Eq.
by

the

equal to zero and using the fact that n^NP = pr^NP, where v + d\ be a solution vector corresponding to the value of density

Slider Bearing
_.3000
c

o .2333
Fig. Finite element model
5

of

steady flow of a lubricant through

plane slider bearing.
a

T7////////7////////// X
06667 1.333 2000 (inches)
Moving Surface
1596 J. T. ODEN AND L. C. WELLFORD JR. AIAA JOURNAL

FINITE ELEMENT
SOLUTION

Fig. velocity component lubrication

6x

forthe
problem.

0 -05 ., 3333 >

VELOCITY(in./sec.xlO')
DISTANCE FROM LEADING EDGE OF THE BEARING (ia)

p 4- Then d\, also vanishes. Taylofs series and mLM * is the inverse of the mass matrix mLM. The first
dp.

f(v+
p+d)

expansion can be introduced in the neighborhood of v if f(v, p) derivative pressure variable formulated similar

ofthe

canbe

in
is continuously differentiate at v. Thus fashion. We differentiate Eq. (46) with respect to time to obtain
f(v + v, p + 6p) = f(v, p) + B(5v + Cdp + higher order terms
where matrix vector: B,-/v, where
BisanNx

andCisNx1

p)=

dffa p)/dvy9 C(v, p) = dftv, p)/dp ij = 1, ... , N. Thus, within

terms first order, have linear system B<5v Cdp.
By differentiating (32) twice observe that r^Lv^
of

we

the

Given the initial value of the solution vector v0 and the initial
Eq.

we

=0;
value density p0 series expansion forms basis consequently
ofthe

=0,the

the

of iterative scheme which 1 TH iterate given yv> X

UN rvLmLMlhMNk> yu ruLmLlHMk (48)
an

for

then+

is

by

=
Vn = v. - B- Hv,, pn)C(vn, pn)dpn+ ! ; p n+ ! = dpt (45) If hw inverse of XUN, obtain explicit expression
isthe

we

the

forthe
1 derivative pressure variable
=1

ofthe

This procedure amounts to a piecewise linearization and the Pv hwyu (49)

number iterations should determined accuracy require-
=

Equations (46) (49) form F(x, which

of

be

by

ments.
and
arenowfth

x=

r),

can be directly integrated by standard Runge-Kutta schemes.

Unsteady Flow

We shall present results in the next section in which the solution Some Numerical Results
of the system of nonlinear differential equations, Eqs. (32) and
(33), were integrated numerically using self-correcting order We shall now cite representative numerical results obtained
a

4th

Runge-Kutta technique. To outline the essential features, con- by applying theory methods presented earlier specific
the

and

to

sider first Eq. (33) rewritten in the form problems in two-dimensional flow of incompressible Newtonian
FMk (46) fluids. demonstration purposes, shall employ six-node
For

we

triangular elements of the type shown in Fig. 1, for which the local
where velocity pressure fields given quadratic polyno-
and

are

bythe

pMk (47) mials

FINITE ELEMENT SOLUTION

Fig. 7 y velocity component for

lubrication problem.
the

I 2 3 T I 2 33 I I 3 ~ | 3 |
3I2

I23?

.3333 .6666 I.OOOO 1.3333 1.6666 2.0

VELOCITY (in./sec. x IO4 )
DISTANCE FROM LEADING EDGE OF BEARING (in.)
 DECEMBER 1972 FLOW VISCOUS FLUIDS FINITE ELEMENT METHOD 1597

OF

BYTHE
vk = (aN + bmxl .004-1
> FINITE ELEMENT
p (aN bNixl cNijxlxj)pN (5) SOLUTION

=
+

+
Here independent constants %, bNh cNij (cNij cNji;
thesix

=
ij, 1,2,3; 1,2,..., depend only local coor- .003-
k=

N=

6)

onthe
dinates of the six nodes of the element. These local approximations "I
determine relevant arrays matrices local \
u:
alofthe

and

inthe
momentum, energy, and continuity equations described earlier m
~ .002-
for each element.
Concerning convergence accuracy this particular
and

of
approximation, we mention that Zlamal24 has obtained the error
estimate .001-
v-v\^ (K/sin 9)h2 (51)
where given continuously different?able field, finite-
visa

element interpolant (i.e., coincides with nodal points), visa

v

vathe

h is the maximum diameter of all finite elements in a given mesh, 0 1.0 2.0

6 smallest angle between sides triangle, DISTANCE FROM LEADING EDGE OF BEARING (IN.)
isthe

anytwo

ofa

and

K is a constant independent of 9 and h. Clearly, long flat elements

lead to poorer conditioned systems than networks of isosceles Fig. Variation pressure fixed slider bearing wall.

of

onthe
triangles. case elliptic parabolic problems, estimates
Inthe
of

and

of type (51) lead directly convergence error component nodes included. seen,
the

inEq.

to

and

estimates also involving I/sin and, energy convergence, K/z4.

at

7,8and9re

Ascanbe

in
the initial starting period negative tangential velocity com-
9
for

While study stability convergence finite-element ponents occur several nodes. This apparently either
the

ofthe

and

of

at

was

dueto
approximations hyperbolic problems scarcely beginning, the applied stress boundary condition coarseness
of

is

orthe

ofthe
preliminary results seem indicate that local character model used. Because of this constraint small transverse velocity
to

the

of

approximation lead inherently better conditioned systems components were computed. They were symmetric with respect
the

to

than conventional difference schemes equal accuracy. to the lines connecting nodes 11 and 15 and 3 and 23. The time
of

histories transverse velocity component nodes

ofthe

at

7and9
Couette Flow are given in Fig. 4.
The problem of unsteady Couette flow through the domain
An Incompressible Lubrication Problem
indicated in Fig. 1 is considered. The following boundary
conditions were applied: 1) The x velocity component was The two-dimensional flow of lubricant between a slide block
assumed equal 0.1 in./sec. velocity component slider bearing moving surface determined. fifteen
tobe

to

andthey

ora

and

was

A
was assumed to be equal to zero at y = 0.2 in. 2) The x and y element model, as illustrated in Fig. 5 was constructed. The
velocity components were prescribed as zero at y = 0. 3) The following boundary conditions were applied: x-velocity

1)The
stress on the boundaries x = 0 and x = 2.0 in. was set equal to component was 0.01 in./sec and the y velocity component was
zero. 4) The gradient of the pressure in the direction perpendicular zero along velocity components were
y=0.2)Thexand

to the wall was zero at y = 0. As initial conditions, we set the equated to zero along the slider bearing wall. 3) The stress at all
pressure and the velocity vector equal to zero at all interior nodes unconstrained boundary nodes was prescribed as a hydrostatic
at t = 0. The value of the mass density used was 0.00242 Ibf.-sec.2/ pressure 0.001 lbf./in.2 gradient pressure perpen-
of

4)The

ofthe

in.4 viscosity here subsequent results dicular to the bearing wall was set equal to zero.
andthe

andil

was

assumed 0.00362 lbf.-sec./in.2 In initial calculations, convective inertial terms were

the

the
tobe

In Fig. 2 the tangential velocity profile at x = 0.5 in. is presented ignored. The finite element solution for the velocity profile in the
at various times and shows good agreement with the exact x direction is presented in Fig. 6. The transverse velocity profile
solution.21 In Fig. 3, time histories of the tangential velocity is presented in Fig. 7, and the pressure along the inclined bearing

DISTANCE FROM LEADING EDGE (in.)

1.0 2.0
I I
LU

1
Fig. 9 x velocity component at
section A as computed by the in-
cremental densities technique. fe

LU
o

I
Q

-7002 0 .002 .004 .006 .008 .01

X VELOCITY COMPONENT AT SECTION A (in./sec.)

1598 J. ODEN WELLFORD AIAA JOURNAL

T.

ANDL.C

JR.
DISTANCES FROM LEADING EDGE (in.) NOT SCALE
"cvi

IN
888 jj o
Sod d !

8
9-
w/////////// /?// ///////////777/7 /////7/7T*

Fig. Finite-element model flow over

10

of

a
NODE A flat plate time histories velocity
'L-

and

ofthex
components nodes

at

AandB.
NODE B

z
.3 .4 .5
TIME (SEC.)

is indicated Fig. transverse velocity profile changes were 10. in./sec. and the y velocity components were zero at all
in

8.The

direction between x = 0.3333 and x = 0.6666 in. resulting in a unconstrained nodes at t = 0. The pressure was also zero at all
back flow in the bearing. The nonlinear convective terms were then nodes time zero. mass density lbf.-sec.2/in.4 assumed.
at

of1.

was
included, and the bearing problem was solved by the incremental Computed time histories velocity components

ofthex

at
densities method described earlier. resulting velocity profile nodes A and B are presented as representative variations in
The

in direction Reynolds number included Fig. 10. In Fig. 11, the steady state finite-element solution is
thex

at

of240is

in

Fig. Reynolds numbers larger than more significant compared Blasius boundary-layer solution specific
9.For

240

tohe

at
changes were observed in the velocity profile; these results are points on the plate. Again, excellent agreement is obtained with
now being evaluated with a more detailed model. rather coarse mesh.
a

Boundary-Layer Flow
The problem of transient boundary-layer formation over a References
flat plate analyzed using finite element grid Fig. 1
Fix, Larsen, Convergence Iterations
was

the

of

10.

This model primarily useful study velocity profile

G.and

K.,"Onthe

ofSOR

for Finite Element Approximations to Elliptic Boundary Value

is

for

ofthe

away from the leading edge of the plate. Thus results are presented Problems," SI AM Journal of Numerical Analysis, Vol. 8, No. 3, 1970,
for spatial points at least 1 in. from the leading edge where the pp.2 536-547.
model converges to the exact solution. In order to predict flow Courant, "Variational Methods Solution Problems
R.,

forthe

of

patterns at the leading edge, where high velocity gradients are of Equilibrium and Vibrations," Bulletin, American Mathematical
encountered, a more detailed model of the flow would be required. Society, Vol. 49, Jan. 1943, pp. 1-23.
3
following boundary conditions were specified: Turner, M. J., Clough, R. W., Martin, H. C., and Topp, L. P.,
The

1)Thex

velocity component equal in./sec. velocity "Stiffness Deflection Analysis Complex Structures," Journal
and

of

of

the Aeronautical Sciences, Vol. 1956, 805-823.

waset

to10.

andthey

component was equated to zero at x = 0.05 in. 2) The x and y

23,No.9

p.

4
Zienkiewicz, O. C., The Finite Element Method in Engineering
velocity components were zero at y = 0. 3) The gradient of the Science, McGraw-Hill, London, 1972.
pressure in the direction normal to the wall was zero at y = 0. 5
Zienkiewicz, O. C., Mayer, P., and Cheung, Y. K., "Solution of
4) The pressure was zero at all boundary nodes at which the Anisotropic Seepage Problems by Finite Elements," Proceedings of the
velocity specified. American Society of Civil Engineers, EM1, Vol. 92, 1966, pp. 111-120.
wasnot

6
initial conditions specified that velocity components Martin, H. C., "Finite Element Analysis of Fluid Flows," AFFDL-
The

thex

I.O-,
BLASIUS SOLUTION

3 FINITE ELEMENT
SOLUTION

a .7-
UJ

I6" Fig. Comparison finite-element

1

of

and

Blasius solutions for the steady-state bound-

ary layer on the front portion of the plate.
i*
Q

.2-
.1-

3 4 5 6 7 8 9 10

DISTANCE FROM THE LEADING EDGE OF THE PLATE (ia)

 DECEMBER 1972 FLOW OF VISCOUS FLUIDS BY THE FINITE ELEMENT METHOD 1599

TR-68-150, Wright-Patterson Air Force Base, Ohio, 1969, pp. 517-535.
7
Oden, J. T., "A General Theory of Finite Elements; II Applica-
tions," International Journal for Numerical Methods in Engineering,
the Method of Singularities," The Aeronautical Journal of the Royal
Aeronautical Society, Vol. 73, Nov. 1969, pp. 959-961.
15
Argyris, H., Scharpf, "The Incompressible Lubrication

J.

D.W,
Vol. 1969, 247-259. Problem," Aeronautical Journal of Royal Aeronautical Society,

The

the
1,No.3

p.

8
Oden, Somogyi, "Finite-Element Applications Vol. 73, Dec. 1969, pp. 1044-1046.
J.Tand

D.,

in
16
Fluid Dynamics," Journal Engineering Mechanics Division, Argyris, J. H., Mareczek, G., and Scharpf, D. W., "Two and

ofthe
ASCE, Vol. 95, No. EM4, 1969, pp. 821-826. Three-Dimensional Flow Using Finite Elements," Aeronautical

The
9
Oden, J. T., "Finite Element Formulation of Problems of Finite Journal of the Royal Aeronautical Society, Vol. 73, Nov. 1969,* pp.
Deformation and Irreversible Thermodynamics of Nonlinear Con- 961-964.
17
tinua," Recent Advances in Matrix Methods of Structural Analysis and Reddi, "Finite-Element Solution Incompressible

M.,

ofthe
Design, University Alabama Press, Tuscaloosa, Ala., 1970. Lubrication Problem," Transactions of the ASME, Journal of Lubrica-
of

10
Oden, T., Finite-Element Analogue Navier Stokes tion, Vol. 91, July 1969, pp. 524-533.
J.
"A

ofthe
18
Equations," Journal of the Engineering Mechanics Division, ASCE, Baker, J., Numerical Solution Technique Two-Dimen-

A.
"A

for
Vol. 96, No. EM4, 1970, pp. 529-534. sional Problems Fluid Dynamics Formulated with

in

theUsof
11
Thompson, E. Mack, T., Lin, "Finite-Element Discrete Elements," TN-TCTN1005, Bell Aerosystems Co., Niagra
G.,

L.
and

F.S,

Method for Incompressible Slow Viscous Flow with a Free Surface," Falls, N.Y., 1970.
19
Developments in Mechanics, Proceedings of the llth Midwestern Herting, Joseph, A., Kuusinen, R., MacNeal,

D.N,

J.

L.
and
Mechanics Conference, Iowa State University Press, Ames Iowa, R. H., "Acoustic Analysis of Solid Rocket Motor Cavities by a Finite
Vol. 5, 1969, pp. 93-111. Element Method," NASTRAN: User's Experience, Vol. NASA

I,
12 Tong, "The Finite Element Method Fluid Flow Analysis," TMX-2378, 1971.
P.,

in

Recent Advances in Matrix Methods of Structural Analysis and Design, 20 Oden, J. T., Finite Elements of Nonlinear Continua, McGraw-Hill,
edited by R. H. Gallagher, Y. Yamada, and J. T. Oden, University of New York, 1971.
Alabama Press, Tuscaloosa, Ala., 1971, pp. 787-808. Schlichting, H., Boundary-Layer Theory, McGraw-Hill,

New
13
Fujino, T., "Analysis Hydrodynamic Plate Structure York, 1968.
of

and

22
Problems Finite Element Method," Recent Advances Matrix Eringen, A. C., Mechanics of Continua, Wiley, New York, 1967.
bythe

in

23
Methods of Structural Analysis Design, Edited Gallagher, Ladyzhenskaya, O. A., The Mathematical Theory of Viscous
and

byR.H

Yamada, Oden, University Alabama Press, Tuscaloosa, Incompressible Flow, Gordon and Breach, New York, 1969.
Y.

andJ.T

of

24
Ala., 1971, pp. 725-786. Zlamal, M., Finite Element Method," Numerische

"Onthe
14
Argyris, J. H., "Two and Three-Dimensional Potential Flow by Mathematik, Vol. 1968, 394-409.

12,

p.