You are on page 1of 28

NASA/CR-2002-211662

ICASE

Report

No.

2002-22

Force

Evaluation

Involving
Renwei

Mei, Dazhi

University
Li-Shi
ICASE,

Curved
Yu, and

of Florida,

July 2002

Geometry
Wei Shyy

Gainesville,

Luo
Hampton,

in the Lattice

Virginia

Florida

Boltzmann

Method

The

NASA

STI

Program

Office...

CONFERENCE

Since its founding, NASA has been dedicated


to the advancement
of aeronautics and space
science. The NASA Scientific and Technical

SPECIAL
technical,

The NASA STI Program Office is operated by


Langley Research Center, the lead center for
NASA's scientific and technical information.

largest collection of aeronautical and space


science STI in the world. The Program Office
is also NASA's institutional
mechanism for

to be of continuing reference value. NASA's


counterpart of peer-reviewed
formal
professional
papers, but having less
stringent limitations on manuscript
length and extent of graphic
presentations.
TECHNICAL

MEMORANDUM.

Scientific and technical findings that are


preliminary
or of specialized interest,
e.g., quick release reports, working
papers, and bibliographies
that contain
minimal annotation. Does not contain
extensive

analysis.

CONTRACTOR

or

PUBLICATION.
Scientific,
or historical information
from

NASA programs, projects, and missions,


often concerned with subjects having
substantial public interest.

The NASA STI Program Office provides


access to the NASA STI Database, the

TECHNICAL
PUBLICATION.
Reports of
completed research or a major significant
phase of research that present the results
of NASA programs and include extensive
data or theoretical
analysis. Includes
compilations
of significant scientific and
technical data and information
deemed

PUBLICATIONS.

Collected papers from scientific and


technical conferences,
symposia,
seminars, or other meetings sponsored
cosponsored
by NASA.

Information
(STI) Program Office plays a key
part in helping NASA maintain this
important role.

disseminating
the results of its research and
development
activities. These results are
published by NASA in the NASA STI Report
Series, which includes the following report
types:

in Profile

TECHNICAL
TRANSLATION.
Englishlanguage translations
of foreign scientific
and technical material pertinent to
NASA's mission.

Specialized services that complement the


STI Program Office's diverse offerings include
creating custom thesauri, building customized
data bases, organizing and publishing
research results..,
even providing videos.

For more information


Program Office,

about the NASA

STI

see the following:

Access the NASA STI Program Home


Page at http://www.sti.nasa.gov

Email your question


help@sti.nasa.gov

Fax your question to the NASA


Help Desk at (301) 621-0134

Telephone the NASA


(301) 621-0390

via the Internet

to

STI

STI Help Desk at

Write to:
REPORT.

Scientific

technical findings by NASA-sponsored


contractors and grantees.

and

NASA STI Help Desk


NASA Center for AeroSpace
7121 Standard Drive
Hanover,

MD 21076-1320

Information

NASA/CR-2002-211662
ICASE

Report

No.

2002-22

Force

Evaluation

Involving
Renwei

Li-Shi

Curved

Mei, Dazhi

University

in the Lattice

Method

Geometry

Yu, and

of Florida,

Boltzmann

Wei Shyy

Gainesville,

Florida

Luo

ICASE,

Hampton,

Virginia

ICASE
NASA

Langley

Hampton,
Operated

Research

Center

Virginia
by Universities

Space

Research

Association

Prepared for Langley Research Center


under Contract NAS 1-97046

July 2002

Available

from the following:

NASA Center for AeroSpace


7121 Standard Drive
Hanover,

MD 21076-1320

(301) 621-0390

Information

(CASI)

National

Technical

Information

5285 Port Royal Road


Springfield, VA 22161-2171
(703) 487-4650

Service

(NTIS)

FORCEEVALUATIONIN THE LATTICE BOLTZMANNMETHOD INVOLVING


CURVEDGEOMETRN
RENWEI
MEI*,DAZHIyUI,WEISHYY$,
ANDLI-SHILUG
Abstract, Tilepresent
workinvestigates
twoapproaches
[brforceewfluation
in thelatticeBoltzmann
equation:themomentum-exchange
methodandthe stress-integration
methodon thesurfa.ee
of a body.
Theboundary
condition
fortheparticledistribution
functions
oncurve(]
geometries
ishandled
withsecond
orderaccuracy
base(]
onourrecentworks.Thestress-integration
methodis eornputg_ionally
laborious
for
two-dimensional
flowsandingeneral
difficulttoimplement
forthree-dimensional
flows,whilethemomentumexchange
method
isreliable,
accurate,
andeasytoimplement
tbrbothtwo-dimensional
andthree-dimensional
flows.Several
testcases
areselected
toevaluate
thepresent
methods,
including:(i)two-dimensional
pressuredrivenchannel
flow;(ii) two-dimensional
urfiformflowpasta columnofcylinders;
(iii)two-dimensional
flow
pasta cylinderasymmetrically
placedin a channel
(withvortexshedding);
(iv) three--dimensional
pressure-drivenflowin a circularpipe;and(v) three-dimensional
flowpasta sphere.Thedragevaluated
by using
themomentum--exchange
methodagrees
wellwiththeexactorotherpublished
results.
Key

words,

method,

Boltznmnn

method,

method,

boundary

stress-integration

Subject
1.

classification.

Fhfid

t'erce

evaluasiou

condition

on fluid-solid

tbr

curved

interface,

geometries,

momentum-exchange

accuracy,

3-D

flows

Mechanics

Introduction,

1.1.

Background

equation
is the

(LBE)
particle

density

p and

cedure

the

velocity,

in phase
u)

easy

used

with

lattice

are

a single

kinetic

space

(x,

obtained

to in_plerneni,
method

lattice

Boltzmann

microscopic

velociw

computational

widely

equation

of the

solves

is explicit,

alternative
most

lattice

_)

time

time

t, from

[I,

LBE

dynamics

The

which

the

macroscopic

has

increasingly

problems

due

Because

in various

Gross,

(flow
solution

become

of the
and

Boltzmann
(, t), where

the

systems

version

t;o Bhatnagar,

f(x,
quantities

_, t).

method

of lattice

function

of f(x,

2, 3, 4] is a discretized

approxinmt;ion

method

distribution

integration

the

fluid

equation

method.

for particle

moment

parallelize,

for solving

Boltzmann
relaxation

and

through
and

equation

equation

pro-

an attractive

[1, 2, 3, 4].
model

Krook

mass

The

Boltzmann

(BGK

model)

l
a_.f + _. Vf
where

f(o)

mass

density

is the

*Department

Maxwell-Boltzmann

p and

momenum_

of Aerospace

equilibrium
density

pu

are

= _[f-

f(o)],

distribution
the

first

(1.1)

function

and

_ is the

(D + 1) hydrodynamic

relaxation

moments

of the

time.

The

distribution

Engineering,

Mechanics

& Engineering

Science,

University

of Florida., Gainesville,

PL 32611-6250

(email address:
1Department

rwm_aero.ufl.edu).
of Aerospace Engineering,

Mechanics

& Engineering

Science,

University

of Florida,

Gainesville,

FL 32611-6250

(email address:
*Department

ydz(C..aero.ufl.edu).
of Aerospaee Engineering,

Mechanics

&: Engiaeering

Science,

University

of Florida,

Gainesville,

FL 32611--6250

(email address: ws@aero.ufl.edu).


ICASE, Mail Stop 132C, NASA Langley
(email

a.ddress:

NASA Contract
23681--2199.

luo@icase.edu).
No. NAS1-.97046

This research

Research

Center,

wa.s supported

while the author

3 West Reid Street,


by the National

was in residence

at ICASE,

Building

Aeronautics
NASA Langley

1152, Hampton,
and Space

VA 23681-2199

Administration

l:_esea.reh Center,

ttampt;on,

under
VA

function

f and f(0_ in D dimensions.

and reduced

to a very small

of f and f(0) as well as their


by quadrature

exactly

Eq. (1.1) reduces

It can be shown

set of discrete

velocities

fluxes can be preserved

up to a certain

to a discrete

velocity

the equilibrium
discretized

where

f,_(x, t) :---iI(x,

distribution

in space

_- ..... t/6t

discretization

With

equation

t) and f_o)(x,

t) :----f(o)(x,

of the (tth

discrete

relaxation

time

Boltzmann

by expanding

equation

(1.3) is obtained

series of "a up to second order

and

5._:is the lattice

discrete

se_ {Co,} in D spatial


provides

dimensional

model

nine-velocity

velocity

of three

three-dimensional

offers a better

sets {Co.} that

and

{ properly

discretized,

t) are the distribution

{_,

respectively.

flmction

and

Equat, ion (1.2) is then

solved

function

9 ,

space;

follows,

of computational

the fi[_een-velocity

stability

function

and coefficient

velocity

be pointed

mode]

in Taylor

'w_ depends
units

out that

(D3Q19)

there

[10].

exist other

[6, 7]. A comparative

study

the nineteen-velocity

of accuracy

and

the nineteen-velocity

computational
model

The D2Q9 and D3Q19

and three-dimensional

on the

of 5x = ]

set; {e(_} for the twomodel

(D3Q15),

in terms
that

and accuracy.

in two-dimensional

from

(1.4)

nineteen-velocity

(D3Q27),

vector

2]

for hydrodynamics

aI. [111]. It was found

coherent

in t,he ]attice

t)

we shall use the lattice

It should

symmetry

model

The

fc_eq)(xi,

distribution

- _'tt

lattice

sets of the two models.

including

vector.

as

of coefficient; w_ and the discrete

the twenty-seven-velocity

combination

in general

and the three-dimensional

by Meict

velocity

the Maxwell--Boltzmann

In what

have the sufficient

LBE models

conducted

distribution

of the underlying

the details

used in this study for force e_duation


(1.3) is conveniently

space

can be replaced

(ix = e_dt is always the displa.cement

u) +

dimensions.

(D2Q9)

the discret, e velocity

discrete

3,

constant

and (it ..... 1. The Appendix

has been

integral

moments

(11.2)

e_ is a discrete

[6, 7], and can be expressed

where c _= fixSt;

(D3Q19),

velocity

_,

velocity

The equilibrium

efficiency

_ can be discretized

g/,-_ = lA[f_ ----f(0)]_.


_,

sites.

model

the moment

of the Bol{:zmann

of space and time is done in such a way that

1 shows

because

in { [6, 7, 8, 9].

a lat, tice site to one of its neighboring

Figure

exactly,

space

b}, and the hydrodynamic

x and time t into

is _he dimensionless

velocity

velocity

= 1, 2,...,

order

_,

funct.ion

the particle

model

Otf_ + _"
In the above equation,

that
{g_la

(D3Q]9)

models

flows, respectively.

will be

Equat;ion

in two steps
1

collision:
streaming:
which

is known

is uniform

The simplicity
of constant
number,
1..2.
arbitrary

as t;he LBGK

and requires

spacing

because

f_(xi

+ Coit, t + 5t) = L(xi,

[1, 2]. The collision

computational
nature

effort,

of the LBGK

(5,_. = du) , and

consequently

,_
(1.5b)

step is completely

which makes
scheme,

l),

however,

lead to the

local and the streaming

Eq. (1.5) ideal for parallel


necessitate

unity

of the

the use of the square


local

step

inlplementation.
lattices

Courant-Friedrichs-Lewy

5t = _._:= 1.

Boundary
curved

t) = fo(x_,t) .....:;

scheme

little

and compact

L(_,

condition

wall geometry,

for

a curved

geometry

in the

LBE

method.

Consider

as shown in Fig. 2, where the filled small circles on the boundary,

a part

of an

x_., denote

,,/!

18

, 713

"
"
..' "

9_

7 ....

,
,/

F_G.

].

Discrete

nineteen-vdocity

velocity

(D3Q19)

_'et {e(_}.

0eft)

15

t........

_--

14

Two-dimensional

nine-velocity

--I --

--#

,/12

(D2Q9)

'model.

(right)

Three-dimensional

model

the intersect;ions

of the boundary

the fluid region,

A, is defined

with various

latl:.ice-to-lattice

links.

The fract.ion

of an int:.ersected

link in

by

A .....Ilxf - x"'ll
Ilxs - xbll
Obviously

the horizontal

or" vertical

distance

between

x!, and x_,, is AS:_,on the square

In Eq. (1.5b),

the value of ]_ (xi, t) needs to t)e constructed

the bounda,ry

conditions

b_ck bounda.ry
boundary

condition

well understood
second-order
outside

boundary

at:. the solid boundary


this bourlce-back

accuracy

(for the Couette


node

of straight
halfway

point

where

between

solved

[25]. He and Doolen


except; at the

Eq.

Filippova

and Hgnel

geometry

was developed

[27] and

of the LBE method.


detailed

assessment

in Ref.

tions,

the bounce-back

it is important

that:, it:. is irnt)ossible


condition)

grids
shaped

schemes

analytical

place;

using

boundary

condition

geometry,

number.

a finite

conditions

numerical

scheme

the Knudsen

21]. l_k)r example,


are satisfied

conditions

do not. exist.

grids

method

boundary
in order

to compute

on the accuracy

to impose

work.

a given

are one-h_lf

are imposed

ensional

First

velocity

advantages

]'_(xi, t).

The

Boltzmann

simula-.

of all, one must


(the

and the Couette

[19, 20, 21].

flows.

Dirichlet

realize

boundary

[29, 30, 31] would be mani-

grid spacing

substantial

of

for curved

the

to obtain

works

of the flow field has been given

layer type of phenomena


in the Poiseuille

condition

role in lattice

conditions

f_

tile entire mesh,

to retain

only at the boundaries

of

this difficulty,

throughout

accurate

the

the accuracy

[26]. In the recent

play an important

Nevertheless,

placing

condition

is used

with simple

simply

To circumvent

difIi?rence

it is

with

(A = 1/2)

with interpolation

how the boundary

conditions

solutions

is employed

spacing

on the grid level and degrade


Reynolds

the use of Cartesian

scheme

because

boundary

bounce.-.

condition

and this is only true

takes

flows and in Ref. [11] ['or some three-din,

[32, 19.2(.,

boundary

where the bounce-back


body

with

boundary

collision

coordinates

bounce-back

for us to understand

hydrodynamic

and

the no-.slip

one half lattice

coordinates

boundary

for any kinetic

pa,st, the

to approxirnate

flows) at the location

and higher

of the boundary

type

on a given grid node,

fested in kinet.ie
where

An itlterpolation

and 0 <_:_ < ].

of the boundary

In the

in order

_;he no-slip

Mei e_ al. [28, 11], a second-order

[28] for some two-dimensional

Because

at finite

in conjunction

on the impact

boundary,

grid [19, 20, 21]. For a curved

curvilinear

the

the

satisfies

will alter the geometry

used body--fitted
where

condition

to the lattice

(1.2) in eurvilinear

boundaries

to the location

a solid boundary

and Poiseuille

two nodes

lattice,

[12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 2.-1]. However,

the boimce-back

lines parallel

according

= :*:_ lies t)eyond

boundary

tile flow field and the force on tile body


Mei and Shyy

xi

has been vise to deal with

that

of' a boundary

boundaries

if the grid

condition

(1.6)

flows, the location

away from

For flows around

evidence

shows that

the boundary
an arbitrary
the bounce-

F_(',. 2. Layout
diamo.nds

of the r_gularly spaced lattices

(_) denote

flv, id _odes,

boundary

and curved wall boundary.

locations

(xw),

The circles (o), discs (), shaded discs (,*), and

solid nodes which are also boundary

nodes (xb)

inside

solid, and

solid nodes, respectively.

back

boundary

conditions

at boundaries,
This

poin_

Force
the

several

is crucial

evaluation

schemes,

been

used

to evaluate

He and

Doolen

the

components

body-fitted
grid-cell

pressure
grids

to

instructive
number
Figure

momentmn

the

dynamic

stress
used,

effect

due

accura.te
cylinder

deviat.oric

t,he

solid

to

3 shows

the

tensor

was

to

the

stresses

bounce-back

examine

the

by using
the

pressure

on

the

variation
the

LBE

of the
method

body

by

the
taking

needed

for curved
in

surt)ace
To

grid

spacing

boundaries

gain

integration

total

correction

[26, 23, 24,

respective

on the

for flow

over

the

coefficient

Cp

p - p_'
PU "_

on the

33].

extrapolating

and

[26, 34], have

Even

and

though
the

half-

[27] developed

dynamics
by

the

cylinder

correct

Hfinel

stress

force

method.

to

fluid

fi'om

evaluation

of the

obtained

surface

of a circula.r
of cylinders

stress

gradients.

the

method
accurate

Several

[,BE

in order

was

of

accurate

surface

Filippova

method

surfa.ce

a column

of the

However,

by

into

of surface

velocity

channel

LBE
is the

interaction.

i;he stress

condition.
boundaries.

for the
studied

to say,

context

stress

to obtain

of _he cylinder

pressure

Needless

in the

a two-dimensional

insight

improvement
systematically

in fluid-structure

16] and

boundary

condition

[27, 34].

[14,

by integrating
obtained

been

boundaries.

on a curved

were

boundary

not

especially

exchange
force

placed

one-half

curved

of numerous

has

curved

dynamics,

extrapolation

boundary

tha.t

involving

force

the

of handling

In spite

issue

an

asymmetrically

and

obtained

fluid

[26] evaluated

was

spacing

force

including

capable

works.

of fluid

including

work.

important

dynamic

and

and

present

related

one

to "bhe study

of the

grid

a second-order
a circular

and

interpolations,

accurate

in the

years,

of the fluid

force

with

second-order

demonstrated
evaluation

last

determination
of the

in fact

is also

1.3.
during

are

combined

force

on

integrating

nearby

the

Cartesian

integration,

it

cylinder

at finite

Reynolds

(see Ref.

[28], and

Sec.

is

3.2).

1.5
1.0

0.5

0.0

-0.5

-1.0
-1.5

0o

90

180

270

360 c

FI('.

3. Flow past a colu'mn of 2D circut, ar cylinders.

circular

cylinder

of radius r -- 6.6, and center-to-center

LBE *_sult denoted


multi-block,

on the
those

surface

velocities,

i.e.,

components
the

e2,

the

is compared

grid
using
with

size

with

a Cartesian

unity

of the
and

t;hat

obtained

radial

direction
solver

grid

a nmch
dr

with

of coarser

fluid

In Fig.

..... 0.026

resolution:

(relat;ive

body-fitted

grid

resolution.

force

LBE

Nevertheless,

in the

body-fitted

high
the

e3,

Only

ers, and

links

in the

be

more

clear

how

eT, are
diagonal

noisy.
the

The

noise

(3D)

(indicated

by

coordina_ses,

an(]

pressure-based

the cylinder

solution

still

and

the

is smoother

obtained

the

essentially

cases,
symbol

the smallest

result;

than

of the

force

on
the

st;ress-int;egration

a sphere
body

for a cubic

in which

xb

opposing

directions

LBE

agrees

by
result

with

in suspension

is effectively

lattice.

resides

met;hod,
flow

replaced

The

fbrce

on

in the

solid

of the

neighboring

region]

Lad(]
[14].

used
In

by a series
each

link

results

the

the

flow

simulation

of stairs.

[halfway
from

momentum-exchange

Each

between

the

using

segment,

two

momentum

method

_.he

the
on

lattices

bounce-back

the

surface

at a:f and

exchange

(per

to compute

unit

xb =
time)

e_

an area
(xf

its applicability

Whereas

the

momentum-exchange

and

accuracy

for a curved

boundary

method
have

not

is very
been

easy

to implement

systematically

studied.

of

+ e_jt)

between

lattices

-7 ---e_.

_he

boundary
has

ec_.f a (x f

in which

in

method.

in two--dimensional

Not; surprisingly,

LBE

pressure.

stress-integration

around

resolution

b_j usin9 a 3D

of Cp(0)

result

201 points
to _...... 1).

and

e_,

the

It is not

is manageable
3, the

a 3D multiblock,
finer

by

of C1_ would

pat, tern.

dynamic

integration

cases.

by using

variation

obtained

upstream

i.e.,

intersected

noisy

of a 2D

solution.

Instead

condition,

accuracy

the

far

velocities,

points

a similar

o'n the surface

is located at 0 -- 180 . The

resolution.

Po_ is the

or vertical

considered,

point

The solid line is the result

where

boundary

show

extrapolation

[35, 36, 37] with

Navier-Stokes

Navier-Stokes

fluid

the

also

tensor

coe_[ieient Cp

The stagnation

solver with a much finer

horizomal

If the

in three-dimensional

solver

along

the

the

3.

are

the

in the

laborious

es,

-- 20.

extrapolation,

by

Fi&

stress
aitbct

programming

Navier-Stokes

and

deviatoric
stresses

Navier-Stokes

second-order

by

oj" the pressure

H/r

with r -- O.(i and Re - 40.

intersected

given

e4_ e6,

and

it is rather

using

x_.,

result

of the

pressure

While

by

points,

in the

is obtained

grid, and pressure-based

obtained

b()undary

considered

x)

by s,;rnbols

body-fitted

Distribution
distance

computationally,
To recapi_ulate,

two

there

are two major

of stress

tensor

Cartesian

grids.

the intrinsic

with the method

sirnplicit:y

for example,

stress-integration

method;

can be noisy.

1.4.

resolution
number,

Scope

the resolution

Boltzmann

Whether

method

present

work.

In what

and the momentum-exchange

on the wall in a pressure

momentum-exchange

to compute

method

driven

follows,

methods,

channel

cylinders

using a second--order

reliable

placed

2.1.

for Force

accurate

no-slip
boundary

expansion
dist;ribution

no-slip

for the case

to the cases where


(b) As in the ease of

force,

it is not elea.r what

at a given (moderate)

value

boundary

condition

in detail.

The shear

to assess

['or the

[38 i obtained

by

For flow over a cylinder

is further

flow over a two-dirnensional


medlod

assessed

of Fornberg

method

of the

on the drag three

will be subsequently

grid resolution.

and normal

5he suitability

The results

the result

i.e., the

array

are eompu_;ed
evaluated
of spheres

for force evulua_;ion

and

for threeat finite

gives Nirly

flows.

Method.

boundary

condition

in the lattice

for t;he distribution


function

for t:he force evaluat;ion,

drag anti lift coefficients

momentum-exchange

in LBE

for a curved

for the post-collision


aceura.te

with sufficient

and three-dimensional

Evaluation

the Chapman-Enskog

with

The momentum-exchange

that; _;he simple

for the two-dimensional

Second-order

scheme

pipe flow and for a uniform

We found

conditions

in

The problems

was proposed

the total

incurred.

two methods

a'b Re .... 100, the unst;eady

in _;he literaeure.

fully developed

results

finite difference

in a channel

with the results

2. Methods

boundary

a.eeurate

using these

the errors

at Re = 100 are compared

The drag

number.

for flow field.

flow will be first, examined

when A # 1/2 and analyze

of circular

Reynolds

the

surfa.ee

and the near wall flow variables

will be described

coefficient

dimensional

Second,

are very laborious

can be applied

two methods

consistency.

compared

simulations

fbrce on a bluff body

for flow over a column

asymmetrically

literature.

to the boundary

(a) The scheme

is often lbnited

fluid dynamic

and tile use of

in the

needs to be investigated.

near a solid body

the components

tensor
geometry

this scheme

is not straight,

reliable

of the stress

are as follows.

First,

near the body

of a three-dimensional

exchange
x_,.

resolution

not been addressed

of the lattice

intersection

stress integration.

say, Re _ 0(102).

of the

st;ress-int;egration
stresses

has

eornponents

surface

the boundary

is to obtain

of surface

due to limited

If one uses the momentum--exchange

the adequate
of Reynolds

on the

of n_omentum

with A = 1/2 at every boundary


A 1/2, when,

surface

for Cartesian

of t:he stresses

with

associated

with the method

of such a method

of the extrapolation

t;he integration

comparison

associated

The accuracy

irnplementa.tion
and

problerns

are often noisy on a curved

for curved

Boltzmann

function

geometry.

equation

is accomplished

at the boundary.

on t;he right>hand

side of Eq.

The following

(1.Sb)

The analysis

of

by applying
approximation

can lead t;o a second-order

[11, 27, .8 i
3
t) + :_f2(;ev, t) + 2w_.pTe_.

]; (xb, t) = (1 - X)/_:(xf,

_.,,,

(2.1)

where
f_ I,"b, _) = wop(xf,

t)

1+

c2 ' "

. ,_vf)

zc

2c

(2.'2)

= f2_q)(xs, _)+ ,,_'_,(x_., t)2;_-e,_. (_bs ....._s),


and

UbS .....UH .....US(X s+e_`6t,t),


1

"b_" .... }Z(2A

- 3)W + ,T,A_,,

(2A -1)
X ..... (_---2)

'

(2A - 1)

_ .....(_-+ 1/'2) '

O<A<[,
1

_ < _ < 1.

The

above

treatment

is applicable

for

both

the

two--dimensional

and

three--dimensional

lattice

Boltzmann

models.
By substitution

of Eq.

(2.2),

Eq.

(2.1)

becomes

"

_(eq),.x

t'}]

3
+w_,p(xf.t)_e_
,
,
Thus,

the

above

treatment

viscous

etIi_ct)

near

for the

momentum-exchange

2.2.

Force

integrating

the

the

of curved
wall

total

can

parameter

the

relaxation

effect

due

t.o the

based

stresses

boundaries

(with

evahmtion

on

on the

fi is the

denotes

the

body-fitted
be

near

accuracy

stress

of the

due

to

the

dynamic

force

evaluated

on

using

the

the

a C.aitesian

computed

to further

body

grid

for

was

tbr

(:lose

on

part
the

vectors

lattice

used

(the

accounting

the

force

by

that

the

evaluation

the

number

since

u is not

on

fi_,'s suffers

of the

that

of the

the

body

shear-layer

(with

wall

loss

0iu.i

of
will
the

st.resses

arbitrary

the

can

to obtain

below].

flow

:u

primary

the

deviatorie

in the
near

the

derivative

(2.7)

[26], a

of grids

scheme

[see Eq.

information

rank,
In Ref.

a large

integration

funct.ion

the

of second
operator.

based

a similar

surfaee

d_ when

such

[27] except

upon

tensor

e_f_

dist:.ribution

on the

(2.5)

tranpose

_o:

f_'s

cylinder

} ,

Irl general,

Oi'_l,j,

in

based

size

was

[34] used

pz-;rticle

procedure

a finite

term

[26] evaluated

identity

T is the

of u using

a circular

st.ress

relaxation

to a tbrcing

Doolen

and

gradient.

numbers

of the

I is the

0iu,

I_'ilippova

and

:,*:) + (iV:'_t)r]

stretching

evaluat.ion

force

extrapolation

of accuracy

grid

accuracy.

the

used,

an

the

of the

0g_,

0_,

are

velocity

of two

of the

non-equilibrium

through
loss

and

cancellation

degradation

with

a reliable

simulations

He

cylinder

boundary

components

together

to yield

(2.4)

as a modification

;Xi), in additional

{----pl + p,,[(V

of the

whose

be thought

integration.

boundary

vector

tensor

body

[,BE

in further

since

normal

system

the

in the

result

out
rank

coordinate

placed

variable

be

unit

second

-- uj, -- 2uw). ,

wall.

F = f
d._,_.
J 0f2
where

. (Ubf

7a

were

However,
A)

field.
is not

have

This

to

leads

sufficiently

resolved.
In Eq.
D3Q]9

(2.5),

models,

the

pressure

c_ =

1/3

# can

so that

be easily

p = p/3.

evaluated

The

using

deviatoric

the

stress

equation

of state

for two--dimensional

p = c_o.

For

incompressible

D2Q9

and

flow

=
(:an be evaluated

using

the

non-equilibrium

part

( )

(2.G)
of the

dist.ribution

where
flow

e(_,i and
past

to carry
components
order

e(_,j

are

a circular
out

the

on the

and

numerical

of the

jth

set
given

deviatoric

scheme
boundary

based

on

the

0f_ is eomputed

Jb_f2

the

entire

commenting
code

is taken

here
up

that
by the

for the
above

component
of surmce
by' Eq.

stress

F' = f d..'t
It is worth

O:

Cartesian

a separate
integration

symmetric

extrapolation

exerting

ith

cylinder,

function

tensor

values

(2.51).

on

The

on the

of p and

the
values

surface
Tij at the

velocity
cylinder
of the

points

can

.... [f<_ _ f}eq)]

eo/_ij

of t.he discrete
points

f_,,eq)

(2.7)

e_,
can

be

pressure

respectively.
introduced
and

be obtained

neighboring

fluid

For

t.he

in order

each

of the

using

a second-

lattices.

The

six

forc_e

as

{pl +
two-dimensional
force

evaluation

:u)+
flow

,.
past,

procedure.

a cylinder,

nearly

(2.s)
half

of the

length

of

2.3. Method basedon the momentumexchange.In orderto employthemomentum--exchange


methodefficiently,
twoscalararrays,w(i,j) and w_(i,j) are introduced.
A value of 0 is assigned
to w(i,j)
for the

lattice

inside

the

a value
i.e.,

site

solid

(i,j)

tha.t

t)ody.

The

array

of 1 is assigned.

e_

w(i,j)

(see

-e,_

= 1, the

are occupied

For

Fig.

a value

of I is assigned

'w5 (i, j) is set to zero everywhere

a given

2).

momentum

by fluid;

For

nonzero

a given

exchange

velocity

for those

e,_, e(_ denotes

boundary

with

except

node

all possible

xb

to w(i,j)

boundary

"the velocity

inside

the

neighboring

fluid

for those

lattice

nodes,

Xb, where

in "the opposite

solid

region

nodes

over

with

direction,

wv(i,j)

a time

step

the

total

nodes

..... 1 and
dt = 1 is

c_:/O

Simply

summing

the

by the

solid

on the

body

contribut;ion

all

In

the

the

during
As

the

subsequent

memioned

in the

implicitly

taken

examined

and

the

number

of the
since

accurately

an

arbitrary

the

velocity

circular

with

Refs.

body,

force

(acted

after

been

t-t--St)

the

collision

evaluated.

and

f_(x.f,

The

t-+-5_)

move

of the

v_aria.ble

A is not

of f,,(xb,

t + at).

The

step

is carried

monlentum

out;

exchange

to xf

and

explicitly

xb,

respectively.

included,

applicability

and

occurs

of Eq.

but

(2.9)

it is

will

be

of the

the

the

linearly

with

in tile

(2.7)

gives

accurate

driven

second

above

formula

of F and

in a two--dimensiona.1

in two--dimensions

[38].
the

lift

H/r

reliability

the

exact

solutions

flow.

flow.

[see Eq.

(3.9)],

the

together

with

the

for both

momentum-exchange

the

velocity

method
method
domain.

will

on

The

results

first

case
and

tile

The

details

for t:he flow

in the

field

solutions

for

the

of

centers
are

t;hen

momentum-

two-dimensional
shedding).

literat.ure.

wa.ll shear

the

a column

vortex

pressure

flows

by considering

is the

= 100 (with

f_'s
for

methods
r in the

case

the

exchange

between

radius

is the

for three-dimensional

exact

evaluation

third

that

flow past

distance

the

at Re

with

be evaluated

provided

momentum

4) for which

force

(]rag

profile

(2.5)

force

on the

H is the
two

method.

The

the

Fig.

in a channel

is compared

Eq.

two--dimensional

the

of the

of this
placed

coefficients

tbr

(see

is the

using

that

based

= 20, where

dependence

of three-dirnensional

(2.9)

flow

computed

The

results

of Eq.

considered

I_e = 100 and

cases

in a finite

in the
body

is no doubt

channel

case

two

a sphere

xv

of the

for CD

there

is asymmetrically
and

size

formula

correctness

of the drag

to assess

drag

nodes

the

walls,

fbr r;j
the

pressure

momentum--excharlge

over

of boundary

straight

The

values

that

number

will

driven

flow

in a

stresses

are

known.

be made

first

in this

dra.g on a sphere

computat;ion

can

be

due

to

found

in

[28, 11].
3.1.

on

cylinder

for

flow

the

number

is examined

for which

a uniform

consider

of ][:brnberg

pipe

Firmlly,

Eq.

are known.

result

consider

case.

and

the

dependence

assessment

For

The

also

The

to the

of r.

To demonstrate

at Reynolds

method

circular

has

efl>ct

by pU2r

be independent

cylinders.

a circular

time

the

increase

Discussions.

stresses

flow

We

belonging

is evaluated

f_,(x_,,

to the

nodes

for pressure

exchange

The

(2.1)

determination

is normalized

A, we first

adjacent

over

Eq.

section,
in the

should

computed.

and

compared

force

by
when

is proportional

and

cylinders

of two

step

force

of state

are

streaming

boundary

the

Results

equation

given

account

force

coefficient
3.

xv

v_;lidated.

the

However,

nodes

as

the

boundary

introductory

into

Clearly,

drug

met;hod

of j[_ at the

all boundary

is obtained

mb e_/0

momentum-exchange

value

over

fluid)

the

Two-dimensional
top

wall

(y

pressure-driven
H)

at a given

location

channel
x (i =

flow.
N,_:/2--[--1,

In
for

the

case

example)

of the
can

channel
be

flow,

evaluated

the
using

force
the

t_j

*t

N_j

j:=Ny

.---]

P2
X
P]
j

...............................................................................................
_
/ )

_2:t_
......

AS_

-/
j ===:
]

FIG.

4.

rnornentum--exchange

The

channel

method

x and ;q cornt)onenls

flow

as follows.

total

shear

the

LBE

simutations

The wall is located

with

between

an

arbitrary

A.

j = Ny and Nu - 1 (cf. Fig. 4). The

lhe ith node are

F_ = []6(i,J)

+ f_,(i - l,j

- 1)] e6,.,: + [fs(i,J)

+ L(i

+ l,j

- 1)] es#

/_), = []6(i,J)

+ f._,(i - l,j

- 1)] e6,u + [fs(i,j)

+ A(i

+ l,j

- 1)] cs,:_

+ L(i,j

the jt.h Cartesian

and normal

fluid element

in

of the force oil the fluid at tile top wall near

+[]7(i,J)
where c_,j denotes

co'_figu'rntion

stresses,

(3.1a)

- 1)] or,y,

component

(3.1b)

of velocity

e___.Since 5_. = 1, F,: and F,_ are, effectively,

a_,j and u_y, which include

the pressure

and the deviatoric

stresses,

the

on the

at y = H.

Based on Eq. (2.7), the deviatoric

component

and Nv-

2 (or y .... N_ -4 + A) can be exactly

functions

in the flow field if they

of the fluid shear stresses

evaluated

are eorrectly

given.

at j .... N_ - 1 (or y .....N_ -3 + A)

based oil tile nonequilibriurn


A linear extrapolation

part

of the dis_xibution

of _uhe deviatoric

shear

stresses

to y = H = .N:_ .....3 -+-2A yields


T(neq'
%v(j
_:g,W ......
where

the

superscript

value

at the wall.

"(neq)"

flow, the normal

normal

stress

method

of evaluat.ing

of the surface

stress

using the near

of the deviatoric

to the

negative

stress

of I;he pressure

V_y(y) is expected
C-p).

(neq)

integratk)n

based

on the ext;rapolat;ed

integration

pressure

on the solid surface

profile 'n_.(y) is obtained

wall velocity

.... NY - 2)]

(3.2)

value computed
from f_,eq), and the subscript
w refers to the
(n eq)
stress, v,jy,_ , (!an be similarly computed.
In a fully developed

r,_>.,, given by Eq. (3.2) for I;wo-dimensional

no numerical

After the velocity

the

normal

component

uw('y ) is equal

wall except, that

denotes

The deviatoric

channel

..... ,_7Y _ i) + L_'_'k[%,u(j


.... N v - 1) - %y(j

It needs

channel

to be zero while the


"bo be pointed

flow is equivalent

and the deviatoric

out

that

total
this

to "uhe method

stresses

on I;he solid

is needed.

from f_, the shear stress

%y on the wall can also be calculated

profile as

A
....pu.i-i--+----_i- ['u_(j = I\\,, .....1) .....u;_,(j = N_ .... 2)].
In the above,

a linear extrapolation

is employed

to evaluate

the velocity

derivative

(du_:/dy)lv=u

(3.3)
at the wall.

'-A_JLE3.1
Comparison
the lattice

of fluid stresses

unit._, J\_y

by Eq. (3.13);
Eq. (3.1b);

Column

at y

H in a two-dimensional

35 and T

0.6 as a function

_, -Ty,w
(neq)

given

Cobumn _/, pressure

by b:q. (3.2);

p obtained

p(i

entire

by
Eq.

....F,

1.6333

1.6020

2.5555

0.3333

0.3333

1.603

1.6333

1.6030

2.2309

0.3333

0.3333

0.04

1.604

1.6333

1.6040

2.0685

0.3333

0.3333

0.05

1.605

1.fi333

1.6050

1.9710

0.3333

0.3333

0.1

1.6:10

1.6333

1.6100

1.7760

0.3333

0.3333

0.2

1.620

1.6333

1.6200

1.6781

0.3333

0.3333

0.25

1.625

1.6333

1.6250

1.6583

0.3333

0.3333

0.3

1.630

1.6333

1.6300

1.6451

0.3333

0.3333

0.3333

1.633

1.6333

1,6330

1.6385

0.3333

0.3333

0.35

1.635

1.6333

1.6350

1.6357

0.3333

0.3333

0.4

1.640

1.6333

1.6400

1.6285

0.3333

0.3333

0.5

1.650

1.6333

1.6500

1.6184

0.3333

0.3333

0.6

1.660

1.6333

1.6600

1.6214

0.3333

0.3333

0.7

1.670

1.6333

1.6700

1.6244

0.3333

0.3333

0.8

1.680

1.6333

1.6800

1.6274

0.3333

0.3333

0.9

1.690

1.6333

1.6900

1.6305

0.3333

0.3333

0.95

1.695

1.6333

1.6950

1.6321

0.3333

0.3333

0.99

1.699

1.6333

1.6990

1.6335

0.3333

0.3333

exact:

solution

for the

the

parabolic

accuracy

the

LBE

collision

fluid

shear

step

[thus

the

lattice

and

is noted

and

time

step.

is

are

by

cases

stress

_e_,ct
given
'xv,,,
(3.3).

,_exact for

by

by
Also

with
all

10

the

call

be used

results

Eq.

(3A),

listed

doub]e

solution

to

to different

simulation

I;_ given

y) =

is the

comparison
accuracy.
"\ closer

values
by
the

pipe.
@/dx

(3.13),
,
between

examinat,

view

1 throughout

"/Vy ..... 35,


(_
by Eq.

unique,
In

is normalized

in a circular
case

body

be

converges

precision
of A.

can
constant.

in p(x,

flow

equivalent

arbitrary

flow

for a Wpical

values

result

of an

LBE
an

if p(i,j)

channel

three-dimensional

shear

out

t;he addition

up to

procedure

to tile
of the

Eq.

given
unique

in the

normalization

carried
to

field

exact

evaluation.

for different

density

applied

on:

be

(3.4)
This

through

field

only

is

analysis.

force

velocity
can

stresses

This

given

identical

the

(y ..... I-/)

volume

is enforced

values

computations

T (neq)

given by

h_a .... 3 .+-2A

for the

the

based

wall

H =

control

p(x,y)]

It is also

pt,(du_./dy)l:_:=H

----p. All

the

gradient

the

di_culty,

r .... 0.6)

on

methods

While

field

t.he numerical

and

that.

11].

this

domain.

units,

(3.2),

[26,

to compare

at every

3.1 compares

or simple

pressure

density

To circumven_

= N;,ff2)

profile

the

stress

I eli, H.

aforementioned

it is difficult

case.
2,j

of tile
simuladons_

field

(3.3),

velocity

in

1.602

(3.1b)
It

....p,-_}?__ ,_ lo"

0.03

computational

Eq.

......
,-;'._ x lo"

0.02

Table
in the

6, -b':t,

O.3333

pressure

Column

10 -_

5', ---I_sj
given
'

O.3333

the

in each

Eq. (3.3);

-1.0

Cobumn

3.5294

the

of Eq.

5, -pz/(duj:/dy)iu=_t

flow with dp/dx

1.6010

afl_er the

the

given by Eq. (3._);

1.6333

on

tbrce

channel

7-;ceyxact
_uJ

1.601

._exact

In

Column

driven

2,

0.01

Finally,

assess

pressure

Column

in the simulation.

....._,_!,_ lo _' ....F,; lo _

based

of A.

ion

......

10 -_

r <_,,,,
(neq} given
I,}

given

of t,he

by

shear

stress

profile using Eq.. (2.7) across

profile

_;_,:_ tYJ, which is linear,

extrapolation,
oL z_,_,

despit.e

Eq. (3.2), for _-_linear

in the LBE simulation

(neq]

the channel

the errors

in flle velocit;y

profile therefore

of cha,nnel

(neq) _y)
_ , is also equal to the exact:, shear
that; T_y,,_,

reveals

profile 'a_(?j) for all values of A. A linear

gives t;he em-_et wall shear

flow indicates

st;ress

the reliability

stress.

of the

Thus,

_uhe exact;ness

LBE solution

tbr the

_,_

stress

_.[neq)

field v-,;j "kx,,9/ by using Eq. (2.7. However, as Fig. 3 indicates,


the accuracy of integrating
"7ij
(a,
y) to
obtain the fluid dynamic force in nontrivial geometries needs to be further investigated,
as will be discussed
in the following

sections.

For 0 < A < 1, the


agrees exactly
for the normal

zero.

as A increases

linearly

T (= 0.505,

with

(3.1b)

This is a rather

the method

On the other

based

on the mornenturn-exehange

speeial

quantity

of _uhemomentum

since deviatorie

exchange

method

gives the shear

of z and A.

method

condition

method

component

does give a reliable

exac_ resuh

were carried
0.9,

stress

].0,

of

value

is a natural

out over a range


1.6).

error

depends

The absolute

for F;,_. Although


in conjunction

this method

that

the

is not exaet

error

attains

the frequently

used

with the bounce-back

arid the error in the force

on A and the resolution.

"uhe derivatives

This can be understood

of Ny (= 35, 67, 99, and

The result_s indicate

A = 1/3.

of 4/3H

choice for the three evaluation

in t<,. is due to the fact that

.....3 + 2A),

on the top wall a.s

the relative

method

t_ does not etmnge

_.ex_.c_,_y,_
= ?r(dp/dx)(A_.

1.2, 1.4, and

for A = 1/2, one must be aware that

conditions.

approximation

the

for fixed @/dx,

The error in E_ is zero when

gives

using the momentum-exchange


error

hand,

0.6, 0.7, 0.8,

when A = ], which

momentum-exchange

from Table 3.1 that

computations

0.52,

is, F._ is independent

the maximum

of the

by analyzing

Eq.

velocity

(3.1a).

field are not; considered

At the steady

in the

star:e, and with the

that
f]_ _ f(eq).+

f_l)

ct'

Equation

by Eq.

force, it is observed

/A. Further
0.51,

momentum-exchange

The

on the wall.

if'ore 0.01 to 0.99.

131) and

boundary

given

Nevertheless,

(tangential)

increases

evaluatk)n

stress.

For the shear

boundary

force

with the pressure

the force is identically

That

normal

(3.]a)

......f(eq)

"

9-"

,_,,_,3_(_.

v)(,_,. _'}

(_"

(3.6)

'

at the top wall becomes


3
/%, _ 2u_,:_,.pTe2 - ('a_,s + u f - 2uw)

(3.7)

where the substitution


has A dependence
depends

of Eq. (2.4) for _,_ and J_ has been rnade.

is ubf.

on A because

When

0 <_:A 1/2,

u_, = 0 in this ease

over a set of symmetric

lattice

points,

/_ is independent
[see Eqs.

cancellations

(2.3)].

The only term ira the above equation


of A, and

when

In the case where

in the summation

1/2 _< A < 1, F;_ weakly

F;,_is obtained

may further

which

weaken

by summing

the dependence

of F_: on A.
Table

3.1 also shows

finite-difference,

i:.hat for the shear

die loss of accuracy

other" values

of 2X (> 0.3),

the aeeuracy

of pp(du_:/dy)l

time 7 increases
shows

(from 0.5]

dirnensiordess

the accuracy
_ n

based

to 1.6).

wall velocity,

stress

based on the derivative

is q.ui'_e significant
is comparable

u,,/u_,

obtained

velocity

derivative

of the increasing

by a three-point

]]

of _

wit:h t:hat of i;_.

on "_he near-wall

To see the cause

of the velocity

for small values

obt:ained

by using

(< 0.05) when r .... 0.6.

i-{owever, as shown
deteriorates

For

in Fig. 5(a),

as the

relaxation

error in p,_(du_/d[q)l_,=rr , Fig. 5(b)

second-order

Lagrangian

extrapolation

(b)
h

0.006

0.2

1/a
0.5
0.7

0.004

0.002

FIG. 5. The LBE simulatioms

of the ehannel

flow, with A

0.2_ 1/3_ 0.5_ and 0.Z

The pressure

drop is O:_:p

....1.0 10 --(_

in lattice units. (a) l:_atio between the wall forcG puO_uxl_j=;_, evaluated by using Eq. (3.3), and the exact wdue ....%_y_wexact
-HO:ep/'2,
9iven by t':q. (3-I) as a function
of T. (b) Normalized wall slip velocity u,w/uc as a function of "r.

of the

near

wall

velocity

u_(y)

increasing

relaxatk)n

time

pat;h

tha_

causes

deviation

poor

performance

as

the

Since

for directly

velocity

Ref.

The

are

velocity
must

be placed

in higher

drag.

In this

distance

in the

between

parameters

fixed

is located

about

terminated

study,

increase
30 radii

the following

and
the

drag

behind

case,

c ..... 10 -(_ was

Following

Fornberg

chosen

[38], the

that

difference

7(a)

compares
result

CD

of Fornberg

obtained

the

partMe

with

the

mean

free

It is clear

error

in the near

from

j'_

that

wall velocity

[see Eq.

based

wall

(2.7)i

on the

Re

coefficient

_he

profile

without

evaluation

the
of the

details

20 radii

to the

to

of flow

the

field

simulation

d as Re

1.8%

at, Re -

allow

sufficient

wake

relative

L2-norm

error

100.

rest

The

the

boundary

of the

the

be found

where

force,

and the

center

keeping

of circular
symmetry

can

= Ud/lJ,
of the

cylinder

left of the
while

a column

6 for illustration),

determination

between

on the

of Fig.

cylinder

a consistent,

flow over

Re&_cing

computational

downstream

development.
in t;he fluid

upstream
will result

cylinder.

of the

.... 10 and
over

The

boundary
simulation

region

(3.s)

cylinder

of radius

r is defined

as

i&[
..... pU2r.

a vorticity-stream

(3.9)

method,
function

12

is

ft is satisfied,

100.

a circular

momentum-exchange

in

U is the

_-_-_--_)
..........................................................................
< __.
_: II'-(_,
t
+
1)
..(x_,
t)l?_ ilu(x_,t + :1)11
_
-

from:

[38] using

on

solution.

a uniform

part

of the

by about

CD
Figure

left

to 12.5 radii

cylinder

h)r both
drag

u,_

of increasing

method

For

the

of the

tbr

cylinder

based

directly

distance

coefficient

the

criterion

increasing

diameter

at about
the

E2 =

In this

Most

A shorter

it, is placed

bo_mdary

result;

evaluation

H (see

by the

be noted

force

velocity

hydrodynamk:

of cylinders.

at y ..... H/2.

It must

It is the
the

the

slip

recommended.

distance

is defined

[15].

calculated

the

a eohmm

far upstream.

the

would
25

when

inlet.

be

is not

center-to-center

number

boundary

(2.6)

increasing

from

with

?;.i can

over

imposed

in Ref.

derivatives,

of Eq.

The

solution

is associated
tensor

flow

Reynolds

_miform

the

observed

velocity

r and

for f,_'s

[28].

stress

uniform

of radius

conditions

also

of T.

of _he kinetic

in _.he form

Steady

cylinders

the

completing

gradient

3.2.

v- was

of pL,(d',_/dy)ly::::_

r increases.

need

as a flmction

surface

formulation

stress

integration,

at Re = 100,

H/r

and

finite

= 20, and

()

Uo

iii:: i .................................................................
....................................

FIG. 6. Computational
eo'mputatio'nal

radius

domain,

r ranging

methods
small

give

do'main for the uniform

(left)

from

Unbounded

2.8

to

satisfactory

differences

13.2.

result;s

in CD

could

downstream

direction

cylinder

in Ref.

[38]_ as opposed

the

total

force

variation

CD

can

be

ot)tained,

using

the

diameter

of the

cylinder

by Ladd

in the

results

methods

that

[14].

In

the

the

the

poor

results

>

that

values
integration.

momentum
(due

lhe

adds

error

values

exchange

][:"or sm_,ller

resolution),

the

the

boundarie_s oj"

in

at

300

radius,

stress

i.e.,

radii

validity

than

5%,
lattice

This

coarser

integration

for

a. little

reliable

data

spacings

larger

for

across
with

the

fluctuations

resolutions,

method

the

(2.9)

have

is consistent

produces

The
larger

behind

of Eq.

is, ten

force.

[38].

is much

method

of less

me_hod

integral;ion

Ref.

domain

to the

of the

stress

given

is imposed

an

stress-integration

to insufficient

of 1.248

for r' > 5. That

of 5 < r < 7, the


method.

and

mornentum-exchange

Accepting

reliable

lines indicate

computat;ional

credence

the

method,

to obtain

value

condition

This

Th,e dashed

exchange

[38], the

of CD from

momentum-exchange

are necessary

with
in Ref.

here.

r,

bounded do'main in y_ plane.

momentum

bolmdary

30 radii

stress

the

comparison
fact

downstream

The

and (right)

8, both

in

to 25

t:he range

than

give

the

from

to the

on a body.

less

finding

than

for

flow pa,_t a sphc'r_ o]' radius

i'n xg plane,

t;br

l)e due

in the

evaluating

domain

while

yields

both

much

larger

errors.
Figllre

?(b)

compares

CD obtained

for Re =

:10. The

the

for r > 8, an average

data

a larger
over

of two

resulis

than

7 are

exchange
lattice

l%r r less

larger

1_.
than

as follows:

method

(i) both

gives
across

In other

thal

more
the

consistent

the

from

the

drag;

CD

5, the

result

In contrast,

the

in the

are

Reynolds

give

range

needed
number

accurate

Re*

even

from

(=

U/_)

8).

<

on
has

results

_he stress-integration

100,

the

comparisons

(ii) the

momentm_-

a resolution

consistent

should

Based

method

converged

from

obtain

integration

for r > 8. Averaging

results;

of 10 < Re
to

r (>

between

conclusions

in order

stress

at larger

in CD

The

the

stress--integration

difference

fluctuation

can

and

method

The

met;hod.

(iii)

cylinder

(grid)

a converged

_ 3.319.

evaluation

and

exchange

n_on_ent;um-exchange

or around

for force

of the

lattice

to gives

is obtained.

gives

than

of momentum

momentum-exchange

methods

diameter

words,

seems

integration

in the

methods

_ 3.356

r result

is about

spacings
values.

large

of CD

stress

is much

in Fig.

the

the

method

value

for r > 8, the

methods

method

drag

momentum-exchange

fluctuation

the

from

of ten

and

be less

than

reliable
10 in the

calculations.
In

the

grid,

thus

effect

of the

above

]?resented

t;he comput;ational
mesh

is repeated

with

The

of the

radius

results

symmetry
different
cylinder

in Figs.

mesh
on the
vah_es

?(a)

is symmetric
accuracy

of the

is deliberately

and

?(b),

with

respect;

of t;he force

cylinder
chosen

the

center

13

of the

t;o the

geometry

evaluation,

offset

to be only

center

A_ in the
6.4 lattice

the

cylinder
of the

calculation

z direction,
grids.

In order

is placed

on

cylinder.

To test

of the

flow

or A_j in the
to preserve

a lat;tice
the

at Re ..... 10
;q directioN.
the

mirror

1.8

........

R;=loo

1.7
[]

momentum

[]

stress

3.51 (}))

.....

exchange
inLegraLion

[]

1.6
x

xN

0
1.3

[] []

1.2

momentum

[]

stress

D x

1.4
xD
x

exehange
integration

3.41

1.5

Re=lO

[]

3.31

[]

_-[]

_ --X-

.....

=-_ ....

-_

[]

[]

[]

D-xx

[]

1.1

10

12

[]

3.21

14

10

12

14

F*C,. 7. The drag eoeJJicient


dashed line indieates

for a uniform

the value of CD

flow past a column

1.24: obtained

of cylinders

i.n Ref. [38]. (b) Re

over a range of radius r.


10.

(a) [{e .... 100.

The da,shed lines indicate

the values

The

of

CD

averaged over _ laryest radii.

symmet;ry
(at

of the

y =

H/2).

the

flow

symmetry

the

periodic

when
are
grid

A:_ =

poim

the

For

Ay = 0 while
with

0, but

with
due

than

---_4) to

that

the

walls.

the

The

19;

when

The

coincides

due
This

order

For
to the

Ay
the

the
to

flow

to symmetry

of the

is aJso observed

symmetry

here

of _he variation
is dictated

coefficient
offset
The

outcome
the

and

of the

lift

is

because

when

boundary

variation

Av

coefficient

It is our

respect

CD
from

is precisely

of the

with

studies

conditions

that

This

in CD.

by that
mesh

independent

drag

y direction,

change

maintain
Av, we use

of cylinder
spacings.

to Ay.

in the

which
varying

We notice

in the

computational

in other

13 laltiee

boundaries

boundary

of the

center

perturbation.

A:_ _ 0 results

used

results

due

lower

conditions,

symmetric

of the

that;

and

A_ = 0 while

The

about

than

of magnitude

schemes

0.

change

by mesh

smaller

due

boundary

equivalent

is only

caused

with

same

to

for upper

x direction.

when
due

conditions

symmetric

are

diameter

asymmetry

evaluation

error

of CD

of magnitude

is the

in the

symmetry

cylinder

the

which

order

force

is well bounded.

line

errors

This

which

flow

variation

symmetry

of the

center

the

boundary

A_:, we use

truncation
one

is lost.

O(10--2),

accuracy

solid

an object

mesh

symmet.ry

varying

at y = H/2,

reflect

expected

we use different

to the

3.2.

to A;_, is about

O(10

at the

respect

better

the

Ay = 0 the
mesh

y direction,

conditions

in Table
is less

consistent

when

in the

boundary

presented

in CD

flow

from

observation
conditions

to the

geometry

of

[23, 33].

'raBL>: 3.2
The e_et
cylinders.
H/r

of symmetry

The Reynolds

-- 20. The variation

of the com.putational

number

Re -

10 (:r -

mesh on the force evaluation


0.6),

the _ndius of the cylinder

o]' (7o due to the change of the center

A,

= 0, periodic

of cylinder

boundary

for the stead, y uniform


r - (i.4 (_n the lattice

flow over a column


unit

of &. -

of

1), and

oJ'J_et j'rom a grid point is less than 17

conditions

..................................................................................................................................................................

at y = H/2

A:_

0.2

0.4

I).6

0.8

cr,

3.3661

3.3637

3.3526

3.3526

3.3637

..................................................................................................................................................................

A v = 0, symmetric

boundary

conditions

..................................................................................................................................................................

Aa.
Co,

0
3.3661

0.2

0.4

3.3666

3.3646

in y = H/2

I).6

0.8

3.3667

3.3692

..................................................................................................................................................................

It

is worth

noting

t.hat

the

wall

shea.r

stress

in

1,l

the

channel

flow

obtained

by

using

the

rnethod

of

momentuln
exchange
hasa relativeerrorproportional
totheresolution
across
thechannel.
Foraresolution
ofi10 20latticespacings
across
thediameter
considered
here,therelativeerrorinthedraga,ppears,
however,
smallerthanin thechannel
flowcase.At R.e= 100,withr > 10,theaverage
valueofthedragobtained
by
usingthemethodofmomentum
exchange
hasa 1.7%relativeerrorcomparing
withFornberg's
data;[38].If
theboundary
layerthickness
is estimated
roughlyto be3 x 2r/_ _ 0, thereareonlyaboutsixlattice
spaeings
across
theboundary
layeroverwhichtheveloeidy
profilechanges substantially.
Based on the insight
from

the

channel

that

are

effectively

in the

channel

to the

good

3.3.

flow

over

cylinder
lower

wall

A_

----0.2

for the

located

the

40 radii

requires

lower

of the
tbr

field.

with

wall,

cylinder

inlet,

Re

point.

The

h._ ----4.0r,

center

cylinder

100 based

in Figs.

and

for a laminar
present
fl'om

the

channel
to the

has

center.

on the

Thus

there

average

inlet

flow over

a circular

r =

12.8

of the

in A+

of the
tbr

are

f_

cylinder

of error

have

as

contributed

vortex

is used
cylinder

..... 0.76

shedding.

cylinder

of radius

and

center

the

to the

wall

upper

and

is placed

at four

test

[39].

results

exit

boundary

benchmark
at the

of 564 105 square

U, tile

upper

of

for the

profile

is used

a total

velocity

levels

may

with

a parabolic

specifieation

extrapolation

body

in channel

eenter

results

inlel

of the

comparable

of the

study,

This

surface

7(b).

cylinder

In tile

the

wall

This

that

lattices

use of relaxation

and

time

is

in the
T = 0.55

U _ 0.0fiS1.

At

this

8(a),

Reynolds
8(b),

number,

and

8(c),

the

flow

respectively,

becomes
show

unsteady

the

drag

eoefl3cient

CD

[see Eq.

and

time-dependent
CI:

and

7(a)

on

suf%r

surface

results

distance

The

stresses

exchange
entire

circular

A zeroth-order

of the
=

shear

the

respectively.

aecording

boundary.

= 2rU/v

shown

a channel.

respectively.

downstream

For

drag

over

placed

inside

and

dev]atoric

of momenturn

a set of benchmark

a grid

is h+ .... 4.2r

upstream

in A = 0.2

in the

placed

coin(ides

the

cancella:tion

asymmetrically

[39] reported

and

ures

an

that

method

error

behavior

is asymmetrically

radii

in the

eft%ctive

convergence

tlle

flow

The

and Turek

r that

it is possible

incorporated

flow.

Flow

Sdl_%r

result,

(3.9)],

and

periodic

vortex

behaviors

F_
t)U2r

of the

shedding

is observed.

Fig-

li[% coemcient

-----=,

the

'

pressure

difference

A P = _'--t'---z-Pb
poU _ '
where

p/and

density

Pb are the pressures

imposed

Ref.

[39].

where

We

at the
first

T is the

given

in Ref.

stress-integration

to note

that

Cr)(t)

although

a.nd

the

curves.

and

from

walls.

There

CD

(from

the

range.

note

there

surfi_ce

vortex
the

lift

curve.

that.

the

of Co,

This

0.25%

h+/r
on the

shed

to note

researchers
that,

both

AP

very

between

the

methods

the

range

The

to the

difference

passages

of these

in Ref.

[39].

present

value

obtained

by

peaks

given.

The

of CD(t)

15

(0.2950

between

the

Instead,
of the
the

CD(t),

in

0.3033,
0.3050)
and

the

it is interesting

momentum--exchange
give

existence

in the

results

is about

method

by the

evaluation

h.__./r ..... 40 in the

was

benchmark

of St values

given

of foree

correspond

the

.... 2r/UT

coefficient

results

(t) curve

St

and P0 is the constant

mon_entum-exchange

_.he drag

cylinder.

two peaks

with

number

the

For

respectively,

compared

well with
between

both

the

are

of Strouhal

graphically.

in the

..... 4.2 and

and

in Ct(t)

difft, renee

occurrence

by different

It is interesting

agrees

behind

of the cylinder_

value

method,

two peaks

back

CD,

numerical

diffbrenee

is about

these

and the

is indiscernible

alternately

difference:

to 3.24)

data

presei_t

stress-integration

is no report

3.22

The

front

the

method

Physically,

a stronger

results

that
of the

We

surface

method

entrance

note

period
[39].

at the

two

of a weaker

strength
cylinder
a range
higher

peaks

of the
and

the

in the
vortex
vortices
channel

of the maxinmm
peak

is well within

momentum-exchange

method

3.25
(b)

......
t..... hange
---stress integration

................................

.....................

-1

3.17

\7

,\J

1000

-_

2000

3000

t-t o

2.46

..................................
, , , , , , , , , 1 , , , , , , , , , 1 , , , , , , , , ,

1000

2000

3000

t-t o

F_(:.
drag
the

8.

symbol

methods

are also within


a multiblock

a cylinder
pressure

stress

At
and

value

AP

time

placed

the

the

graph.

+ T/2)

given

lift. eoe]jieient

(a)
(b)

The

_ (a]_er

6'j,(t)

The

The

The

i'n Table

as shown in Fig. S(b).

a channel.
of time

in f_ef. [39].

respectively.

on

in

as functions

to,

bounds

integratian,

of AP(_o

the range,
procedure

the
lower

indistinguishable
the

asymmetrically

difference

/39].

the upper
and

are

>,', indicates

the

in t_ej'.

indicate

exchange

two

past

and

results

lines

momentum

2D flow
Up,

benchmark

horizontal

the

The

eoeJficient

solid

attains
and

3.3,

where

A fl_rther

6:D(t).

T (_

refined

[40] with r = 40 in the fine grid region

its

Note

that

The

computation

lij_ coe]ficient
to)

value

are

the result._'

pressure

C'L,

compared

C)I _.

are the _esults

is the period

nearly

time

maximum

(e)

1296.5)

yield

run

curves

CL (t).

coejficient

of the

initial

dashed

lij_ coej]icient

drag

variations
an

The

the
with

dashed

obtained

by using

obtained

by using

difference

AP(t).

The

of CL(t).

of the present

problem

the same results

using

for Cry(t)

and

cL(t).
We compile
C_ _

and

obtained

in Table

is mea.sured

and T is the periodicity


of 5t ..... 1).
than

clearly

demonstrates

those

schemes

where

the accuracy

model

flow

St, maximum

(-'_'_
_,

in a circular

when

pipe.

over a distance

of one lattice

16

reaches

T is between

minimum

drag coefficient
diffi?rence

AP

given in Ref. [39]. The


its maximum

value C_ 'a_,

] 296 and 1297 (in the lattice

number

St.

With

a resolution

the bounds

given

in Ref.

unit
much

[39]. This

method.
The ste_-_dy-state

with T ----0.52 [11]. Eq. (2.9) is used to evaluate


of the pipe

fluid dynamics
CL(t)

are well within

Boltzm_nn

and

CL
-m-Jin, and the pressure

of the Strouhal

LBE results

of the lattice

and

of computational

to is the moment

of CL (t). For the LBE simulations,

used in Ref. [39], the

Pressure-driven

circumt_rence

and other

at to + T/2,

number

lift coefficient

We use T ----1296.5 in the determination

coarser

D3Q]9

of Strouhal

and minimum

by the LBE methods

value of AP

3.4.

3.3 the values

C_D
n'in , maximum

flow field was obt.ained

the force on the boundary

iu the axial

direction.

The resulting

points

by using
along the

axial ibrce

F_

TAmm 3.3
Vo.lue8

of

"Momentum"

C _max
D

St_

(?mln
"D

and "Stress"

caleulations.

denote,

CL

, amt AP

'-_mln

respectivdy,

]'or the flow over a 2D cylinder

the momentum-exchange

method

asymmetrically

placed in a channel

and the stress-integration

method

in the LBE

The CFD results" are the bounds in [_e[., [39], which does not have data for 6D-_m_and ('L-'H_I_'"

method

S 1;

Mornentum

C}_ a'x

0.3033

Stress

0.3033

CFD

0.2950

is, equivalently,
a fully

C _m_x
L

the

developed

0.3050

force

flow

C}_,'ax

3. ] 7711

1.0045

-1.0347

3.2275

3.1708

1.0040

- 1.0340

3.2200

given

inside

C}_ in

3.2358

by

3.2400

2rrrS_,

a circular

pipe,

0.9900

where
the

t,he normalized

axial

is the

exact

I w

We examine

_,,

fluid

57)

wall
shear

7l-7 .2

diI:

C_ 'i_'

AP
2.49114
2.4914

1.0100

2.4600

shear

stress

and

stress

at, the

r is the

pipe

wall

pipe

2.5000

radius.

is given

For

by

"

force,

_J.....

d_

(13.11)

.w_.2 d_
Figure
close

9 shows
to

as that

1.

the

pipe

there
results
method

drag

flow,

scheme
flow

is no

notieed

in Ref.

error

based

case).

in Fig.

For

5, the

due

from

largest
the

in the

of r:

accuracy

error

of LBE
distribution

of the

flow

field

inaeeuracy
(as seen

of the

that

23.5.

solution

it adds

Except
for the

of values
if the

in the
in the

in F_. is about

corr@exity
sense

3.5

to the

the

exchange

in F;,:. Given

9 a.re satisfa.etory

a range

accuracy

in t;'_, results

>

the

flow

by the

on momentum

error

r_ over

[11] that
channel

is dict;ated

the

systematic

shown

coefficient

two-dimensional

of the

evaluation
channel

It was

fbr the

accuracy
For

t_he normalized

flow

3.5%

fllrther

flow

evaluation

field

and

section

and

boundary

pipe

5, _l is rather
is not

of A around

force

previous

for r

it

in this
credence

the

errors

The

is exaet.
in the

force

two-dimensional

at r =

three-dimensional
to the

pipe.

method

for the

occurs

the

as good

15.5.

Again,
flow,

_.he

momentum-exchange

ff)r ff}rce evaluation.

1.0
[]
g-

0.9

10

15

20

25

FIG. 9. The ratio '7 betwee*_ the tange'ntial

foree [;_: on the pipe and its exact value (;vr '2Vp) over a ra'nge of pipe radius

17

r.

3.5. Steadyuniform flow over a sphere.Tolimit the computational


effort,a finitedomainof
-H/2 < y 51 H/2 and -H/2 _<_"z < H/2, with H/_* = 10 is used to compute the flow past a sphere of
radius

r' (see

Fig.

6).

two-dimensional
In the
the

Two

array

former

following

eases

are

considered:

(a)

the

flow

past

of spheres

(all located

at :r = 0) with

boundary

eondi{;ions

at jv ..... 1 (:{/ ..... ft/2

ease,

the

linear

extrapolation

the

a single

(:enter

sphere,

of the

and

spheres

corresponds

(b) the

forming

to j_

flow

square

.... 2) for f,_'s

over

lattices.

are

given

by

(3._2)
The

veloeit;y

at jv ..... 2 is set

as

u(j_, 2,j_) = u(j_., 3,j_).


Similar
on

treatment

is applied

fee's at Jv =

inlet,

1 by using

a uniform

upstream

velocity

boundary

D)r

flow

at y =
the

values

profile

a sphere,

and

at 7.5 radii
the

z = H/2.

of f_'s

is imposed

is located

over

H/2

drag

In the

at J:v = 3 (see

Ref.

to the

sphere

eoefFmien{;

of the

is often

0 accounts

(_] .... H/2

and

conditions

are

extrapolation

tile

10(a)

5.8,

for

and

r =

l0

wit;h
this

range

factor

(5_ tbr

high

that

that;

the

The

set of data

(3.12)

4.
Boltzmann

LBE

(3.13)

exhibit

Conclusions.
equation

the
The
gives

reliable

do not

guarantee

remarkable
Two

the

self-consistency
methods

for solving

exchange

method

is very

simple

exchange

method

is not

an

fluid

dynamic
consistent

fluid

exact

and

free

Figure

curve(]

error

10(b)

range

value,

on a sphere
that
while

force
geometry

Hence,

in the
wall

5.2,

5.4,

nurnber
the

over

the

the

funct:ion
difference

is averaged

over

solutions

with

LBE

non-Stokesian
H/r"

correction

10, at Re =

of 4.3,5,

there

is 1.1_.

at

_ 1.963,

10.

symmetry

extrapolation
and

It

is little

= S.0.

at r _ 3.5 for a moderate


the

of

actual

an unbounded

largest

drag

5.6,

r = 3.0

by a factor

at; '.y .... H/2

It

value

boundary

conditions

given

z ..... H/2.

Yet,

by
both

resolutions.

in conjmletion
have

flow.

5.1,

2 _ 3.743

The

the

the

of r, the

flow

for _.<_ since


the

where

t;he vorticit:y-stream

for

at

symmetry

is 1.98_ between

..... 10.

shows

resolution

to not-so-coarse

in the

18

Re
data

0.54%.

average

fluid

It. is shown
The

on

at

4.0,

of (5.8/3)

a uniform

LBE

by

condition

the

method.

this

the

t:.he case

unbounded
3.8,

diffbrenee

based

1.7986

z ..... H/2,

for evaluating

t.o implement.

largest,
For

than

flows

involving

With

3.6,

for 3.1) < r <: 5.8 and

coarse

the
The

conditions

where
for

the

3.4,

by a factor

forces

s{;ream

t?'om

3.2,

0.7.

1.7986

the

more

at ._1..... 2t-f/2

to simulate

If the

surface

from

t;he boundary

results

met;hod

force.

of

t;he case

_5 _

of the

deviation

"hypes

investigated.

of spheres

improvement
largest

specified

drag

array

At

(3.14)

the

roughly

f'rorn

second

posed

in all simulations.

for

at r ..... 3.2.

differs

for the

a planar

for cO, is inherently

be exactly

and

with

solution

which

lattices).

and

correction

is r =

coefficient

ease).

first

.oo_ denotes
in order

difference

is 1.36%

two

= 4.3..5. The

eases

a finite

a drag

values

over

2370/546

are

as

for r = 3.0,

increases

in the

gives

consistent
flow

time

using

r/5 _-. 1.8086,

z = H/2)

relaxation

eonditions

two--dimensional

the

center

t,br

and

0_

a ratio

result;s

drag.

coefficient

sphere

resohJtion

LBE

in c),(r).

of lie.

and

the

very

is (:lear

can

m!, gives

least

obtains

variation

condition

The
of the

resolution

and

to note

systematic

the

a uniform

is important

10.

eomput;at;ion

result

of r, one

the

symmetry

F,
_b ............................
'
6"JrrUpu '

non-Stokesian

and

between

expressed

z .... H/2)

:i:H/2

surface

nodes

have

3.0 < r < 5.8 yield

Eqs.

a.t Re

of

the

and

non-Stokesian

on the

an independent

between

denot;es

at (y =

the

boundary

formulation

the

3.2 that

sphere,

0_

effect

at (?j ..... H/2


is used

nodes

of the

non-Stokesian

..... H/2),

shows

H/r

boundary

counts

the

imposed
for f_

Figure
and

for

ease,

[28] for the

way

Fr
24
C'D ........ f--77.2-------_ = w--- _
_pt,-rc_ _-'
t_e
'
where

latter

at j_, ----1.5 (half


left

(3.13)

with

{;he method

of la{;tiee

been

examined.

The

momentun>

channel

flow

simulation

that

momentum-

shear

stress

is inversely

proportional

to

1.8g

(a)

Re=lO

HJr=m

(b)n

2.00

'

[]
1.82

[]
[]

R LIo H/ =IO

[]
[]

[]

[]

[]

[]
1.81
[]

-o-

[]

[] []
.....
0---[]

[]
"_

[]

1.95
[]

1.80
[]
1.79

[]
1.90

1.78

[]

I,,,,I,,,,I,,,,

I,,,,I,,,,I,,

F_C.
radius

10.

Flow

r at Re

field

(If/r

.--- oc).

the

resolution.

when there
stresses

past

10.

addition,

dashed

The

]tow

(b)

are at least

because

from f_rther

the former

unstable

than

two force evahmt;ion


curved

boundaries

methods,

Acknowledgments.
work is partially
for Aerospace
Engineering
National

supported

Langley

Research

Science

Appendix

Center

Foundation
A. LBE

The nine-velocity

(ERC)

(or 9-bit)

velocity
(with

lattice

models,

(D3Q27)
zero

equilibrium

such

models,

velocity)

distributions

The met;hod

in implementing

is limit, ed. In

the extrapolation

exchange.

superior

does

exchange

lhe

but. it exhibits

t;o the stress-integration


while the latter

not yield

Boltzrnann

is recommended

is derived

method

Often extrapolations

that

values

of integrating

resolution

of momentum

with the lattiee

to Prof.

A.J.C.

Science

under

_lses

a,re more noisy

very aiccurate
method.

local

Among

the

for force evaluation

on

Three

as _he fifteen-velocity

partners

D3Q15,

of Innovative

square

19

This

Algorit;hms

support

of lhe

of Florida,

the

of the ERC.

lattice,

ninet;een-veloeity
[10]. All these

{e_l-(_ = 0, 1, ...,

D3Q19,

partial

at the University

denoted

flows. For three-dimensional

(D3Q15),

set

comments.

Dimensions.

which have been used in the lit;erature


velocity

for his helpfid

the program

and Techno]ogy

and Industrial
and

Ladd

R. Mei also acknowledges

of two-dimensional

for l.he D2Q9,

sphere

unbounded

drag

the momentmn-exchange

resolution

l:{esearch Center

for Particle

in the discretized

of

in an

when the resolution

is perhaps

LBE model on a _wo-dimensional

has been widely used for simulations


cubic

exchange

In addition,

Optimization.

in Two

sphere

and robustness.

(EEC-9402989),

Models

100.

is sufficient

uses exr,rapolations.

are grat;eful

and

there

method

a coarse

as a function

a sin.gle

it; (:an give accurate

more efforts

of moment;urn

accuracy,

Analysis

past

based on the distribut, ion f_nctions

with

by NASA

body,

with the method

can be accomplished

The authors

Engineering

considerably

functions.

the method

for it;s simplicity,

-b_,/OTrrUpt,

flow

at Re _

when

of momentum

method

Even

force evaluation

The

---. 10).

;he body
results

requires

is directly

interpolations.

accurate

across

of t,he distribution

(a)

flows over a bluff

t:he momen_tum-exehange

method

]_ctor

over" r.

(lt/r"

in comparison

while the stress-integration

information,

of spheres

in the method

surface

correction

averaged

gives similar

ion method

to note that

processiug

interpolations

arr'ay

spacings

body

{.ban that

on the body

(
o] O,r)

three-dimensional

ten lattice

the stress-integrat,

non-Stokesian

rabies

'past a planar

of the

fluctuations

It is interest;ing

of the

li'_.es are

In two- and

and int.egration

method

Variation

The

on the surface

much larger

and

sphere.

and

D3Q27

as t;he D2Q9 model,

flows, there

(D3Q19),
models

and

twent;y-seven-

have a rest

(b ....._)}. For athermal


models

are several

particle

fluids,

are all of the following

the
form

[6, 7]

f(fq)=vGp
where
the

w_

is a weighting

factor

constant

the

lat;tice

and

I +

and

time

eo

step,

e<_=
{

and

the

va,lues

of the

weighting

_(e_'u)+_

=c"

is a discrete

velocity,

respectively.

The

c -

5_/5,

discrete

is the

velocities

unit

for

(l, 0)c, (0, =:1)c,

a .....1, a, 5, 7,

(0, 0),
(1,
1)

,_ ..... 0,
(_ = 2, 4, 6, 8,

factor

c,

speed,

t;he D2Q9

a,nd 5:_, and


models

at are

are

(A.2)

wo are

,,_ =

g_

(t = O;

},

_ ....1, a, a, 7,

(f.a)

_,
For

the

D3QI9

model,

the

_o, =

and

the

weighting

factor

w_

discrete

_ .... 2, 4_ 6, 8.

velocities

are

(i,o, o)< (o, 1, o)(:, (0, 0, _)_,

_ = :1 <

(0, 0),
(1,
1,

O) c

c_ = 0,
c_=7
18,

is given

by [7]

wo

(0

1,

l)c,

(1,

discrete
The

velocity

density

set;s {G_}

and

velocity

for the

D2Q9

and

('.an be comput:ed

(A.4)

(_ ..... 1 0,

a_
The

O, ])G

(A.5)

(_..... 7 18.
D3Q19

from

models

are

shown

in Fig,

1.

f,-,,

, .....Zs,, .....EI,7 '),


c_,

(_

= }2 <,s2

= }2
The

speed

of sound

of _he above

LBE

models

<A.sb)

is
1
Cs

and

the

equation

of sl;ate

is that

of an

ideal

gas

such

_C_

that

p .... c7 p.
The

viscosity

of the

fluid

is

1/

for

l;he discrete

f:_eq) is in fact
form

makes

velocity
a Taylor

the

LBE

(A.7)

model
series

method

of Eq.

expansion
valid

only

.....

Cs

(1.2).

It should

of the

Maxwellian

in the

be

incompressible

2o

noted
f(0)

thai;

the

[6, 7']. This


flow

limit

equilibrium
approximation
t_/c-----> 0.

distribut;ion

flmction

of .f(_q) in algebraic

Equation(1.2)is oftendiscretized
in space
x andtimet
fa:(xi
where

r .... ),/at.

above

lattice

Fbr

this

+ ea5_,

LBGK

Boltzmann

t + at) - f,_(xi,

model

equation

-1/2

correction

in the

e_ at, t + at) in Eq.


method

formally

computational

(A.8)

above

formula

a second-order

CttEN,
gas

[2] Y.H.

S. CHEN,

Q_AN,

D.

that;

for

from

series

Europhys.

Lett.,

Phys.

Rep.,

[4] S.

CHEN

AND G.D.

(1998),
[5] P.L.

17 (1992),

S. SUCCI,
222

pp.

solving

the

second--order

equation

derivatives

correction

incompressible

derived

from

the

of f_, when

in l/makes

flows

[7].

the

lattice

Obviously,

the

f_(xi

Boltzm

+
an n

physical

and

Rev.

CES

Recovery

A, 4,5 (1992),
LALLEMAND,

pp.

484.

479

VERGASSOLA,

pp.

145

The

of the

pp.

AND P.

DOOLEN,

Lattice

lattice

Navier-Stohes

R5339

equations

using

a lattice-

R5342.
BGK

models

t3ottzmann

for

eq'_atior_,:

Navier-Stokes

Theory

a_d

W'_tation,

applicatiorzs,

197.

Lattice

Boltz.ma_r_

method

.[or.fl'_,id

flows,

Annu.

t_ev.

Fhfid

Mech.,

30

364.

Bttz\TNA(-_AR,
arwolit_Me

AN[) M.
(1992),

329

(A.8)

(i.9)

in _,. This

MATTttAEUS,

Phys.

D'HUMI_;a>;S,

BFNZI,

t)]

r > 1/2.

AND W.H.
method,

[3] R.

t) - f},eq)(xi,

equation

c_O_.

for p comes

in a Taylor
method

require

Boltzmar_,n

Boltzrnann

in "the Navier-Sl;okes

_......

REFEREN

[1] H.

lattice

is

is expanded

stabilit;ies

tile

1
(
-_[f,:,a;i,

t) =

[1, 2], t:he viscosity

u=
The

into

E.P.

GROSS,

processes

AN13 M.

i_ char:qed

KROOK,

ar_,d ,_eutral

A model

.for" collision

o'ac-compor_ent

system,

processes

Phys.

iv, gases,

Rev.,

l. Small

94 (1954),

pp.

511

525.
[(3] X.

[7] X.

HE AND L.--S.
pp.

R(3333

HE

AND L.--S.

lattice
[8] T.

ABE,

[10]

AND X.
Rev.

l:l.

pp.

MEL

W.

bou_dar:q,
[12]

D.P.

[13]

1171

I. GINZBOURO
Boltzrnanr_,

lattice

of the lattice

lattice
Rev.

Boltzrn, a'nr_ eq,tation,

AND S.A.

method

Phys.,

131

of the

pp.

65

BoItzman_t

E, 5(3 (1997),

Boltzmann

Discretizatiov.

80 (11998),

Corrtputatiortal
195

of the
Phys.

J. Comput.

HE,

S. SuCCI,

Phys.

R_ev. E, 55 (1997),

From,
6817.

6811

b,v meaus

(1997),

velocity

method:
pp.

pp.
space

the

of the

241
i_

[3oltz'ma_rt

equation

ordinate

m.ethod

discrete

to the

for

the

24(3.
the soh, tion

@ the

Boltzro.ann

cquatior_,

(38.

ORSZAC4,

Physics,

Recer_t

Vol.

advar_ces

Ill,

edited

by

in lattice
D.

Bottz__,a_r_

Staufli?r,

(World

co'm,p'_,tir_.9, in A_,rtv.al
Scientific,

Singapore,

242.

StlYY,

D.

J. Comput;.

Zm(_LER.,
pp.

of the

Lett.,

of

199,5),

Theory

equatior_.,

Phys.

QIAN,

derivation

eq_,atior_.,

Derivatior_.

Reviews

[11]

LUO,

Boltz'man,_

SIIAN

Y.H.

./][pr'iori

R633(3.

Boltzmanv,
[9] X.

LUO,

Bo_u'udar"q

YU,
Phys.,

AND
161

co_Mitious

L.-S.
(2000),
for

LUO,
pp.

Lattice
680

lattice

Boltzruamt

method

for

3-d

flows

with

curved

71

(1993),

699.

Bolt._rnanv.

simulations,

J.

S_;at.

Phys.,

1177.
AND P.M.
model,

ALDErt,

J. Phys.

Bov.,_dary flo_v

II, 4 (1994),

pp.

191

21

conditior_,
214.

a_.al'qsis

for

the

thrce-diruensiorzal

lattice

[141A.J.C. LAJOD,
Nu'merical
Part
[:151 D.R.

2, N_merical

NOBLE,

S. CHEN,

condition
[161 O.

for

the

Solid

Rev.

(1995),

E, 52

[171 T.

INAMUf{O,

[181 S.

C,HF, N, D.

simulations,

8 (1996),

Phys.,

HE,

Q.

Zou,

condition

for

Luo,

Q.

zou

[23]

M.

[24]

A.J.C.

Phys.

Fluids,

104

AND W.

G.

ci'_rular

cylinder,

F_LWPOVA
219

MEI,

L..-.S.

143

209.

Bottzmann

condition

in lattice

methods

1591

'method,

for

lattice

Boltzmann

for lattice

and

Phys.,

87

equation

boundar'y

Phys.

Boltzmann

method,

Phys.

[)oltzmann

m, odels,

conditions

analysis
(1997),

on
pp.

for

simple

for

the

non-slip

115

boundary

136.

flows,

lattice

J. St;at.

Boltzm,

Phys.,

ann

BCK

1598.

LALLF, MAND,

Lattice

solutions

J. Star.
t3oltzmann

velocity

pp.

Momentum

transfer

of a Boltzmann-lattice

fluid

4,52 3459.

Boltzmann

simulations

d_fference-based

(1998),

pp.

426

of particle-fluid

suspensions,

J. Stat.

H;*iNEL,

Grid

lattice

Boltzmann

method

in curvilinear

coordi-

448.

method

Phys.,

LUO,

AND W.

on curviIinear

134 (1997),

pp.

refincro.ent.for

An

York,

Phys.,

Rarefied

Gas

Introduction

An

SHYY,

155

306

coordinates

syste'm,s:

Flow

around

315.

lattice-BCg

accurate

(1999),

Dynamics

'models,

J. Comput.

Phys.,

147 (1998),

to the

pp.

curved

307

(Plenum,
_lzeor'y

boundary

treatment

lattice

Boltzmann

in the

330.

New

5%rk,

1969).

of the Boltzrnann

Equation

(Holt,

Rinehart

and

Winston,

1971).
The

Boltzrnann

CORNUBERT,

D.

D'HuMIgaES,

Physica.

(1991),

D, 47

D'HuM_gRES,

M.

Rev.

E, 63

[341 O.

FI[LIPPOVA

(private

[351 W.

SHYY,

pp.

1994).

SHYY,

THAKUR.,

S.S.

and

AND D.

241

its

Applications

LF_VERMOaE,

(Springer-Verlag,
A

Knudsen

New

layer

theory

York,
for

1988).

lattice

gases,

259.
AND P.

LALL>;MANI),

Thirteen--velocity

3D

lattice

Boltzrnann

revised

printing

model,

066702.

communication).

Computational

Transport

Equation

BOUZIDI,

(2001),

Amsterdam,

plex

conditions

Analytic

lattice

Boltzmann

CEeC_CNA_q,

[361 W.

boundary

boundary

model,

13 (2001),

Lattice

J. Comput.

HAaRIS,

Phys.

DEMBO,

the finite

[32]

[331 D.

203

via lattice

boundary

1251.

[311 C.
R.

pp.

hydrodynamic

3000.

boundary

[dGK

pp.

J. Comput.

AND D.

KOGa_'4,

New

1191
On

DOOLEN,

2928

second-order

AND P.

Fluids,

Phys.,

equation,

228.

method,
[291 M.N.
[301 S.

SHYY,

BoItz'mann

971.

and

VEaB,:aC,
pp.

J. Coil:lpuC.

HE AN[)

pp.
[281 R.

(2001),

7 (1995),

A non-slip

Local

pressure

Phys.

Fluid,

simulations

On

927

9 (1997),

AND R.

pp.

of linearized

FIaDAOUSS,

boundaries,

nates,

[271 O.

M.

Phys.

MEI,

AND M.

solutions

On

a discrctized

3111 339.

}:_UCKI[US, A consistent

suspension

OG[NO,

Boltzmann

HE,

LADD

[251 1:I. ME[

Luo,

AND X.

via

pp.

2536.

pp.

lattice

926.

Phys.,

[261 X.

L.-S.

913

BOUZTDI,
with

2527

pp.

model,

F.

D'HuMI_;RI-:S,

the

(1994),

1175.

7 (1995),

84 (1996),

Analytic

88 (1997),
[22]

method,

AND

Fluids,

pp.

suspensions

271

AN[) R.().

in particle

t)l;). 1164

AND D.

J. Slat.

[211 L.-S.

lloltzmann

YOSH[NO,

Phys.

Mech.,

GEO-H(_[ADIS,

MAf{TfNF, Z, AND R.

[191 I. GINZBURG

[201 X.

J.G.

boundaries

M.

of particular

J. Fluid

lattice

BEHREND,

Fluids,

simulation

results,

Modeling

H.

Phenomena

for Fluid

OUYANG,
(Cambridge

J.

Flow

LIU,

and

htte@tcial

AND E.

University

22

BLOSCU,

Press,

Transport,

Computational

Cambridge,

1997).

_I_chniqv.es

(Elsevier,

for

Com-

[37]S.THAKUH,
W. SHYY_
H.S.UDAYKUMAR_
ANDL. HILL,Mu[tibtock
based flow
[38]

B.

[39]

M.

FORNBERG,
pp.

655

Heat

incompressible

AND S. TUREK,
with

Meehar_ies
D.

Steady

Numer.

Tr_msfer,
flow

Pa.rt
past

B, 33 (]998),

a row

pp.

of cir(mlar

intc_faec
367

cytir_der,

t'reatmer_ts

in a pressure-

396.
J. Fluid

Mech.,

225

(1991),

671.

SCH;{FEa
ulation

[40]

solver,

Be.vchmark

High-t-'erforrnanee
(Vieweg,

Yu,

R.

Int.

J. Numer.

MF. L aND

Braunschweig,
W.

SHYY,

Methods

Fhfids

cornp,_tations

Cornpvters
1996),
A multi-block

H,

of laminar.flow

edited

\%1. 52, pp.


lattice

(2002).

23

by

E.H.

547

Bottzrna.nn

tfirschd,

over

a c:qlinder,
Notes

in Ho'w

in Numerical

SimFluid

566.
.method

fur fluid

]tows,

to appear

in

REPORT

DOCUMENTATION

PAGE

Form

Approved

OMB No. 0704-0188

Publicreporting
burdenforthiscollectionof information
isestimated
to average
] hourperresponse,
includhTg
thetimeforreviewing
instructions,
searching
existing
datasources,
gatheringandmaintainlng
thedataneeded,
andcompleting
andreviewing
thecollection
of informatlon.
Sendcomments
regarding
thisburdenestimateoranyotheraspect
of this
collectionofinformation,
including
suggestions
for reducing
thisburden,to Washington
Headquarters
Services,
Directorate
for Information
Operations
andReports,
1215Jefferson
DavisHighway,
Suite1204,Arlington,VA222024302,andto theOfficeof Management
andBudget,Paperwork
Reduction
Project(07040188),Washington,
DC20503.
1. AGENCY USE ONLY(Leaveblank) 2. REPORT DATE
July 2002

3. REPORT TYPE AND DATES COVERED


Contractor
Report

4. TITLE AND SUBTITLE

5. FUNDING NUMBERS

FORCE EVALUATION
INVOLVING
CURVED

IN THE LATTICE
GEOMETRY

BOLTZMANN

METHOD
C NAS1-97046
WU 505-90-52-01

6. AUTHOR(S)
Renwei

Mei, Dazhi

Yu, Wei Shyy, and Li-Shi

Luo

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES)


ICASE
Mail Stop 132C
NASA Langley Research
Hampton,

ICASE

Center

Report

No. 2002-22

VA 23681-2199

9. SPONSORING/MONITORING
National

8. PERFORMING ORGANIZATION
REPORT NUMBER

Aeronautics

AGENCY NAME(S) AND ADDRESS(ES)

and Space

Administration

10. SPONSORING/MONITORING
AGENCY REPORT NUMBER
NASA/CR-2002-211662
ICASE Report No. 2002-22

Langley Research Center


Hampton,
VA 23681-2199

11. SUPPLEMENTARY NOTES


Langley Technical Monitor:
Dennis
Final Report
To appear in Physical Review E.
12a. DISTRIBUTION/AVAILABILITY
Unclassified

M. Bushnell

STATEMENT

12b. DISTRIBUTION CODE

Unlimited

Subject Category
34
Distribution:
Nonstandard
Availability:

NASA-CASI

(301) 621-0390

13. ABSTRACT (Maximum 200 words)


The present work investigates
two approaches
for force evaluation
in the lattice Boltzmann
equation:
the momentumexchange method and the stress-integration
method on the surface of a body. The boundary
condition
for the particle
distribution
functions
on curved geometries
is handled
with second order accuracy based on our recent works. The
stress-integration
method is computationally
laborious for two-dimensional
flows and in general difficult to implement
for three-dimensional
flows, while the momentum-exchange
method is reliable, accurate,
and easy to implement
for
both two-dimensional
and three-dimensional
flows. Several test cases are selected to evaluate
the present methods,
including:
(i) two-dimensional
pressure-driven
channel flow; (ii) two-dimensional
uniform flow past a column of
cylinders;
(iii) two-dimensional
flow past a cylinder
asymmetrically
placed in a channel
(with vortex shedding);
(iv) three-dimensional
pressure-driven
flow in a circular pipe; and (v) three-dimensional
flow past a sphere.
The
drag evaluated
by using the momentum-exchange
method
agrees well with the exact or other published
results.

14. SUBJECT TERMS


lattice Boltzmann
method,
method, stress-integration
3-D flows
17. SECURITY CLASSIFICATION
OF REPORT
Unclassified
NSN 7540-01-280-5500

force evaluation
on fluid-solid interface, momentum-exchange
method, boundary
condition
for curved geometries,
accuracy,

15. NUMBER OF PAGES


28

16. PRICE CODE


A03
18. SECURITY CLASSIFICATION 19. SECURITY CLASSIFICATION 20. LIMITATION
OF THIS PAGE
OF ABSTRACT
OF ABSTRACT
Unclassified
StandardForm 298(Rev. 2-89)
Prescribedby ANSIStd. Z39 18
298102

You might also like