You are on page 1of 26

NASA Technical

Memorandum

102847

A One-EquationTurbulenceTransport
Model for High Reynolds Number
Wall-Bounded Flows
Barrett S. Baldwin and Timothy J. Barth

August

1990

NQI-I0252
" (NA_A-TM-IOZG4/)
A ONE-EgUAT_C]N
:_oR_ULFNCE
TRANSPORT
MOu_L
FnR t_IGH REYNOLDS
N!JM_ER
WALL-_OUNOED
FLOWS
(NASA)
23 P
CSCL
200
_3/54

National Aeronautics and


Space Administratbn

uncl,_s
O310m31

_L

K_

NASA Technical

Memorandum

102847

A One-EquationTurbulenceTransport
Model for High ReynoldsNumber
Wall-Bounded Flows
Barrett S. Baldwin

August

and Timothy

1990

National Aeronautics and


Space AcIministratbn
Ames Research Center
Moffett Field, California 94035-1000

J. Barth, Ames Research

Center,

Moffett

Field, California

SUMMARY
A one-equation
is derived
based

from

on well

other

flows

the

turbulence

a simplified

form

established

has

that

of the

properties

are compared

model

model

with

predictive

avoids

standard

of the

experiment.

and

investigation
and refinement.
robust
solution
methods
are

need

for

an algebraic

k - c model

equations.

over

plate,

flow
The

numerical

the

a flat

preliminary

properties

model

After

is also

of several

presented
interest

indicate
to merit

analyzed

scale

calibration

predictions

results

of sufficient

The one-equation
presented.

length

that
further

numerically

and

INTRODUCTION

One

motivation

for

of well-established
predict

several

shopt.

These

reported

flows

turbulence
verified

and

model,

such

From

our

Sugavanum

_ equations.

was

found

limited

that

also

of difficult

The

model.
a degree

phasis

was placed

methods
ical

avoids

solvers.
t Held

flows.

analysis

one-equation

in conjunction

The

from

model.

Numerical

model
the

has

show

layers
and

Jespersen,

to avoid

un-

algebraic

models.
reports

by

others

(e.g.,

to investigate

the

overall

incorporate
implemented
Aerospace

one-equation

The

main

applicability
of the
and

behind

the

design
extremely
section
the

to a

development

of the

one-

to demon-

equation,

numerical
some

is

of the

wake

model

robust
gives

cases

improvement

one-equation

Sciences

of this

model

two troublesome

turbulent

in a number

model

purpose

of the

significant

for a self-similar
so that

25th

shear

by Pulliam,
need

and

scales.
the

reported

Implementation

AIAA

one

5) and

is cumbersome

a self-consistent

rationale

In the

been

The

computations

are

to properly

reported

(ref.

multiple

be worthwhile

length

and

considerations

needed

with

model

the solution

of the

on numerical

it would

for algebraic

the

(ref.

by only

King

two-equation

models

investigation,

explain

contains

were

than the basic physical


variables
in an effort
that
occur in the solution
of the standard

Workshop

sections

by

scales

blowing.

_ or related

that

Results

Airfoil

next

be used.

and

The

two

the need

were

predicted

computed

two-equation

of that

Work-

results

code

in which

length

surface

apparent

one-equation

of generality

could

theory

the

The

strate

with

Airfoil

poor

ARC2D

adequately

configuration

of k -

to adequately

code.

of algebraic
airfoil

inability

of Baldwin-Lomax

4) as reported

flow problems

to variables
other
numerical
difficulties
course

were

(ref.

uniformly

the

the

Transonic

models

by

cases

King

tangential

it became

Transonic

following

equation

Two

to treat

Coanda

exploits

turbulent

in the Viscous
achieved.

predicted

and

determination

In the

flow and

was

models

Viscous

turbulence

an independent

experience

8))

is to present

range

the
is the

of transforming
the well-known

k -

3.)

to a consideration

(ref.

possibility
to avoid

in ref.

6) using

7) which

leads

results

report

turbulence

in the

separated
algebraic

is the need

that

(ref.

scales

report

(ref.

An example

Barth

(The

of Johnson

motivation

present

reliable.
length

PuUiam

by Coakley

Another

2).

the

in this

algebraic

contained

significant
using

that

using

flow fields

contained
(ref.

documented

solvers

turbulent

by all participants

by Maksymiuk

and

developments

Navier-Stokes
of the

1) or Cebeci-Smith

are

the

of the
model

em-

solution
numerinto

flow

of central-difference

Meeting

(January

1987).

and upwind finite-volume Navier-Stokessolversin both two and three spacedimensions


and generalized coordinates. The computer code for these implementations can be obtained by contacting
the second
author
(barth@prandtl.nas.nasa.gov).
Finally
in the
appendix,

we completely

The
and

authors

summarize

are grateful

for reviewing

the

the

to Drs.

one-equation

P.R.

with

Spalart

and

T.J.

k-

RT

for compressible
Coakley

flow.

for useful

discussions

report.
DERIVATION

We begin

model

a standard

OF

form

Dk

THE

of the k - equations
vt

- V.(v+--)Vk

Dt

MODEL
(see

Patel

et al. (ref.

9)):

+P-

6r k

(1)

D_

_2

D---/=V.(_ + V*)W + c,,_P-c,,--


where

denotes

form

a third

form

equivalent

discussion

the

by considering
systems

in a later

Reynolds

substantive

number,"

"_vt

linear

and

which

section.

derivative,

have

Dt
D

--

From
nonlinear

improved

In particular,

at
8

4-V"

these

two

combinations.
numerical

we consider

and

the

equations
In our

properties.

production

we are

case,

free

to

we do this

to

We will return

a field equation

for the

to this

"turbulence

RT
k 2

RT

= --

(turbulence

Reynolds

number)

(2)

V_

The

field

RT

of RT,
of k and

RT

the

certain

is obtained

= 2 d k/k

e as well

transforming
omit

equation

dRT/RT

as their
diffusion

terms

from

k -

source

(which
the

the

It should

respective
terms

arising

from

d e/e.

e equations

be clear

terms

are

transform

modeled

transformation

that

in both
to obtain

by

considering

differentials

the

substantive

derivatives

without
the
a new

approximation.

k and

In

e equations),

diffusion

model

we
for the

equation.
vRT
D(vRT)
Dt

,_,

--(2-c*}Tr

+ (c' -

vt
2)k + (v +--)V2(vRT)_r_

1__
_r,(Vvt)"

V(VRT)

(3)

where

v_ = c.(vRr)
Note

that

VRT

= k2/e

since

Equation

vt should

rather

than

not

depend

on v at large

RT,

the

appropriate

field

variable

is

RT.

(2) can be rearranged

in the

k2

= -vRT

form

(k! + k2) 2

vRT
2

(4)

Without loss of generality, we can assign


k2a = vRTP
In regions

where

k2 <<

(at

kl this will result

k s =(kl+ks)

large

(5)

RT)

in P .._ e. Note

that

the

k_-s =vRTP(I+

s =k_(l+_)

relation

ks)s
kl

or

ks

(6)

k=_(l+K)
is still completely
leads to

general.

Substitution

of equation

(6) into equation

(3) and

rearrangement

D(vRT)
Dt

(7)
-(2-

The syste_a

ks
_+ ks

c,,)kl

can be closed

by substituting

Dk
Dt
ks = k -

but in a way
available.
THE

that

the

is of interest

one-equation

one-equation

purpose,

RT

to note

of shear

portion

further

k equation

to obtain

V_)vk
o'1

that
model

layers.
model

all of the

The
that

previous

comment.

resulting

procedures

ONE-EQUATION

major
that

(5) requires

approximation

rational

self-consistent
to this

(6) in the

24/
P___p__k2_ k________2
+V.(v+
VvRT
vRT

(4) and

for exploiting

It

equation

(S)

V"-_-gTP.

Use of equations
means

(2- c.)ks

for

from

departures

MODEL
by

is obtained
main
relations

RT. To arrive at a model that is applicable


number
RT is split into two factors

last

that

should

two

can

be viewed

taken

in near-wall

in equation
approximation

is to develop

in all parts
regions,

the

functions

f3 is a damping
commonly

function
used

such

that

in k - e models

RT

(ref.

a = _c./.Rr

,'_ RT

over

of a shear

layer.

at sufficiently
turbulence

For
large

Reynolds

(o)
at

large

9) are introduced

= -c././3Rr
3

(7),

modifications

nr = -_rf3(RT)
where

remain

FLOWS

terms

to be applicable

as a

to dissipation,

approximation

be a valid

of this report

it to be used
are

steps
production

WALL-BOUNDED

the

that

objective

will allow

from

FOR

neglecting

These

equating

RT.

In addition,

damping

so that

(lO)

and
k2

d=e-D-

(kl + kz) 2

YRT
The

definition

of kl applicable

at small

(and

RT

(11)

vf3RT
all RT)

at

is taken

to be

k_ = V.RTP
This

will

allow

function

kl

to be the

f3 is designed

dominant

part

to accomplish

that

O(vRT>
Dt

suitable

approximation

over

a flat

functions

is used

purpose.

near-wall

The

so that

plate

and
the

with

to help

field equation

gradient
model.

P is approximated

(thin

shear

dam_ing
for RT

is used

The

thin

is

(13)

(Vv,).V(vRT)

pressure
the

if the

calibrate

production

P = vt(u_)

zero

region

resulting

+ (v + --)V2(v._T)

flow

damping

of k in the

v,

-(c,,fz-c,,)

Incompressible
mine

(12)

to

deter-

shear

layer

by

layer assumption)

(14)

Then

vkrP = c.(.kru_)2/.y3
and

the

model

equation

reduces

to

_,, cX/_-_]J"f3RTu_
+(v+ _)(RT)_

DRTDt
--(c"f2--c")
At sufficiently
layer
to the

occurs

high

momentum

in a region

shear

negligible,

stress
the

where

at the

the

wall.

z-momentum

thickness
total

shear

In the

equation

v <<

u_vt,

is the friction
and

velocity

vt = vc_,RT

stress

log region
becomes

2
(_,+ _',)_y = _,_
where

Reynolds

1 ,_',V_
_ ,_-_ _
_,

number

the

beginning

is approximately
and

below,

where

constant
advective

(15)

of the
and

log

equal

terms

are

simply

(log

x/rw,,u/pw,,u.

region

and

In the

log

(16)

below)

region where % = u_/(,cy),

we have
2

vt -

lcu,.y

(log

region)

Uy

where

t is the

Karman

constant
RT

and
=

vt

__ _

VCg

In this
unity

case
and

(RT)v_
advection

= 0 and
terms

substitution

zero

produces

u,-y

C$_

(log

in equation
the

(c,,

region)

following

(15)

with

damping

functions

set

to

formula:

_,,)v,_;/,_

(17)

In the re,on
definitions
of RT

below the
and ]3(RT).

log layer an additional


After a study of the

.RT-

t u,-y
C/_

We can ensure
requiring
procedure

that

this

relation

(log

relation
is needed
to determine
consequences,
we have imposed

region

and

the

below)

(18)

is consistent

with

the

field equation

that the damping


function
f2 (Y+) be adjusted
that will be described,
the following
damping

for RT

(eqn.

(15))

by

to accomplish
that purpose.
By a
functions
have been determined:

where
DI=I-ezp(-y+/A+),
D2=
03
where

y+

that

does

functions
it occurs.

. From
neither

implementation
and

these

+/A +),

A + =10

require

than

f_, nor

Substitution

model

fa, it is convenient

(c,,fa

(17)

and

(16)

to replace
(18)

the

D1D2.

It

However,

it will

the evaluation

of the

Since
product

in equation

= (c',

to remove

c 1

only

f_,fa

f3 individually.

- c,,)v/%DIDayu)j

of equation

that

to D1 or D2.
requires

of f_, and

of equations

we have

f3 is equal

knowledge

f_, and

B3 = 5.2, A + = 15

expressions

of the one-equation

not

Substitution
produces

l-ezp(-y

= 1 + B3ezp(-y+/A+)[1-exp(-y+/A+)],

= u,-y/v

emphasized

A + =26

(15)

- c,,,

uv

and

equations

D1D2

)(

Dv/-D-_I
D_

D1

f_,f3
with

_2

and

D2

with

should

be

be seen

that

product

fg fa

are

simpler

D1 D2 wherever

DRT/Dt

= 0 yields

and

to remove

(10)

(18)

vt

(19)

+ D,/N-N D
For small

y+ we have

the

following

lira

limit

f2(y +) = c---z_
+(1-

y+ --=*0

After

assignment

damping
y+.

function

For our

choice

of the

C(e2

parameters

f2(y +)

when

of constants

value:

c____
)
C(_

c1, c 2, to, equation


the

combination

we have

a limit
5

(19)

can

ft, f3 = D_D2
value

of f2(0)

be used
is known

_ 0.781

to determine
as a function

the
of

From equation (18) it is seenthat

y+ in the

y+ = c.
so that

the

damping

functions

the turbulence
Reynolds
pressure
gradients,
the
determine

whether

In this
recommend
results

this

from

prediction

the

be expressed

functional

dependence

for simplicity,
as a final
(inner)

of vt from

(when

functions

can be replaced

in terms

choice.

adhered

We have

calibrated

Cebeci-Smith

that

model

on RT

we have
model

(20)

of RT

or (implicitly)

in the

log region

in terms

of

involve nonzero
Experience
will

or y+ is preferable.
to dependence

(ref.

by

= 0)

number
RT = RTf3(RT).
For general
flows that
dependence
on RT or y+ is no__ttinterchangeable.

the

report,

can

damping

on y+,

these

functions

2) using

equation

and

below

is

D1) 2 -

1]

by

but

we do not

comparison

(16).

The

with

resulting

(vt)cs

For the

present

model,

= _ [V/1 + (41y+

substitution

of equation

(20)

in equation

vt = v ft, f3 tc y+ = v DiD2

For either

model

u + is obtained

by the

integration

u+ =

Figure
ment

1 shows
of the

designed

for that

From
inner

a comparison

two models

of equation

variations

because

(10),

(12),

(14),

(16),

_-D_ID2

the

and

(18)

explicit

(log

p+

was

realistic
paid

of vt/v
damping

and

u +. The

functions

close

D1 and

agree-

D2 were

formulas

for k + and

P+ in the

i y+

I+DID_

A plot of the

(16):

are obtained

more

toy+

dY +
_o _+ 1 + vt/v

predicted

surprising

yields

purpose.

equations

region

of the

is not

(10)

k + variation
turbulent

to that

purpose

ny+D1 D2
(l+ny+
D1 D2) 2
is shown
energy
in the

region

and

below)

(log region

and

below)

toy+

in Figure

distribution
design

2. It is somewhat
than

of D1 and
6

several
D_.

k-e

surprising
models,

that

although

kl provides
no attention

4O

4 ..........
-;".z::."
..........
_...............
i...............
_...............
........ [

30-

Or re-equation modal

[ .................

I o u"
I
[_
t: COrgi-Smith)
I
I

20 _ ........

vt

I ....

..a._'._................

__>r
09,_i

"_'_.2

_ [.....
/........
i...............
i--:-:_::-:-:_-:-:-::
"
i
_
.
"

Vt (Cebeci" Smith ) I 0,0'"_""-........ _ ....................

:
_

: [Dne-equaUon
il_._ k+

mode_
l

+_-2
...._..........
! i...............
q-- _:

:
!

J_..............

I0

i
0

25

50

75

0;

100

20

40

y+
Figure
1.
Comparison
vt between
Cebeci-Smith
equation

Since

been

introduced,

function

was

although
complete

neither
e nor
the definition

calibrated

Mansour,
For

the

although,

sake

e distribution
The

found

Kim,

from

realistic

f.

Moin

that

a reasonable

(ref.

of completeness,

was

necessary

evaluated

they

direct

of e in the

based on our
D in equation

Energy

(f2)

distribu-

The

above

near-wall

one-equation
(11) is taken

of RT,RT,

f3,

for implementation

that

reported

is a byproduct

defined.

not needed
k _ kl,

D3 was designed

simulations

distribution

are

assuming

function

I00

D3

region,
model.
To
to be uk_

10)).

plots

earlier,

function

it be fully

variation

80

y+
Turbulent Kinetic

2.

and damping
tions.

D3 is needed
in calculations
of RT and RT, the quantity
and

damping
the

it seems

to produce

as mentioned

below.

of U + and
and the one-

model.

f3 has

(from

Figure

6O

f_, and

(i.e.,

to make

in reference

that

e are

k2 <<
the

kl)

shown

in figures

of our
in the

e distribution

10.

However

the

3-5,

model.

The

log region

and

resemble

that

resulting

rather

was not anticipated.

800'
...................

1.75

.......... t..........
.......
I.... g; _
I

..........

1.50]"
1.25"

I y....

2-

i .........

'"

1.000.75 t e

0.50-

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

.rf

.......

_ ........................................................

s r

20001 :_A'"
0

...............i!............................
50

100

150

0.25-

..... y

t ........................................................................

0.0_
0

50
y+

y+
Figure

3.

RT,

RT

distributions.

100

Figure

4.

Damping

functions.

150

0"25I
.............................

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

0.20 l ............................
"_
._,

0.15-

_-,

0.10-

+
W

0.05 -

0.0t3
0

50

1O0

150

y+

Figure

To

complete

i,c_,,c,l,c,

2 are

0.41, c_ = 0.09
zero.

calibration

needed.
and

We have

ible flow over


the

the
set

The

value

is not the

level

_ in the
of RT

an airfoil)

used

on record,

to impose
a relatively
that a constant
value
k and

will not

the

free

the

last

in Hopkins

Reo

is 1.2, which

be large

compared

adopted

that

the

and

+ 6.012

that

Inouye

(ref.

(Karman

report,

r't in the

free

molecular
8

low value
equation
production

we adhere
stream
viscosity.

t =

for incompress-

(and

by

11):

"standard"

to our

values

is well represented

- Schoenherr

outer flow. From


in regions
where

in this

used

parameters

(7) will be identically

coefficient

data

the

of the

widely

in equation

is well below

to the

values

skin friction

9. An alternative

However,
such

model,

of experimental

larg value of RT in the


of RT is self-preserving
stream

distribution.

term

calculated

see reference

k - _ equations).

in the

we have

Reo) _ + 25.111og10

of c,1 currently
lowest

report
so that

given

epsilon

one-equation

to a compilation
formula

= 17.08(log10

this

c, 1 to match

plate

Karman-Schoenherr

1
C!

In

of the

c,2 = 2.0,

adjusted
a flat

Near-wall

5.

formula)

value

of 1.44 but

of c,,

would

13 it can be seen
is zero (unlike

to c,_ = 1.2 and


at the

be

leading

a low
edge

of

5'

4O

t+
.............

35- "'I
One,..equat[on
model
+
I .... u..i,_,,,.,b,_.
I_
"IN--4 -- Log l_er
$_..........
_
_,v I o _,_.=32._._o(M=o.t) li
._
25

I.!_

Re0=29'700_ = 2.0)

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

,{
"it

....

_............. i............. _..............

5t.............
_

10 i

10 2

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

10 4

_ .... Theory (M=O.l)


i I_
Calculation
(M=2.O)

6.

In this
has

been

10000

20000
Re 0

Figure

we extend

for algebraic

to compressible

flow with

the

model

models

6 graphs

2).

(ref.

Mach

law-of-the-wall

numbers

where

by Hopkins

and

an empirically

Inouye

(ref.

formula

11)

plate.

a practice

which

relations

apply

above

with

(u + = u*/u,(see,

versus

for example,

y+)

Rubesin

for

and

subsonic
Horstman

momentum

Finally,

temperature

thickness
value

the

pressible
friction

v/F/pwo, d u
to the conventional

several

to compressible

determined

compressible

contains

Too = Tedge(1

alternative

flow.

In the

law of the wall.


procedures

procedure

is good.
was
Figure
several

for applying

by Sommer

and

paper
the
Short

+ 0.035M_dge)

Reynolds

number

+ 0.45(T,_,tt-

Reo

is then

T_dg_)

adjusted

to a corresponding

in-

Reo:

resulting

Mach

The

Ts_ is defined:

C/

from

Reo

= Reo#(T,

dg,)/l_(T,,)

the

Karman-Schoenherr

formula

is adjusted

value by multiplying
by the temperature
ratio Tedg_/T,o.
Figure
for compressible
flow over a fiat plate with the foregoing
"theory"

supersonic

using

solutions

flow this plot corresponds

Karman-Schoenherr

model

all the

on fiat

12)):

For incompressible

bers

friction

pl; t

u* is defined

u* =

The

that

v and

Skin

flow following

We assume

a temperature-dependent

computed

and supersonic

7.

to compressible

(ref.

[.l t

Figure

30000

10 5

Law-of-the-wall.

report,

used

I
/

<>,,=_.0)t....

y+
Figure

[
[ ....

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

"i...........

10 3

One-equation model
Calcul-tion (M--O.I)

x.,iel""_

/
O,l -:--..-.:.,'". .......[ ........[ ................ :, ........
10

3 _'_i--_e_

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

I
?"1 --v--

""*( ...............

............
+.............
i..............

i............. ,L"......

'+1.............
+__.
10 -i

numbers.

The

excellent

calibrated

in the

8 contains
values

The

agreement

agreement

at low

incompressible

law-of-the-wall
of

for both

free-stream

Mach

and

numbers

was

comskin
and

supersonic
expected

Mach
because

numthe

limit.
plots

RT"

subsonic

to the
7 compares
at subsonic

Note

for
that

flat

plate

vt/v

boundary-layer

= c, RT

= 0.09RT

flow
when

computed
damping

functions
are at unity.
free stream
RT tested,
unity

in the

free

The plotsindicate
except
(RT)

insensitivity
to free-stream
= 100 which is the only value

values for all values of


at which vt/v exceeds

stream.
30

....

:,

/ I _-_,,,_=_,
251 "I ....
,_,,,
_b'.m
/
_
I

204'""'1

/
I *

li
_ ._/
|I .................
_'"_ll_
.....

Log t._
I"
Et
P_= o.I (ftee-tu_un)
I :"
.dt"
_ffi 1.o (tttt-_)
[.l.....'""_'"i
................
I%ffi lO.O (ftet-s_tun)]
i ._"
!

151................
_................
_"/""-I"i
.................
_................
10l................
::
ii................
_i................
i.............
i_.
/

OJe--_-:_r,,,,;........
; ........
l ........
', ........
I0 -I

100

101

102

10 3

I0"

y+
Figure

To further
formed

assess

show

M_

= 0.75,

the

coefficients

mesh

solution

distribution

to values

one-equation

in the

Viscous

for viscous
model

The

layer

of the

a Reynolds

solution

CD = 0.0279.

coefficient

cases

one-equation

Baldwin-Lomax

of CL = 0.895,
pressure

and

a = 2.72 , and
for the

the

of boundary

the performance

for two of the troublesome

9-11

with

Sensitivity

8.

flow

number
were

9.

RAE

2822

model,

Transonic

over

the

RAE

CL = 0.771,

plotted)

which

shock

position

is substantially

is in much

better

0.75).

mesh.
10

produced

agreement

2822

airfoil

Computed
This

lift and

10.

per-

Figures

geometry
lift

drag

improved
with

were

Workshop.

CD = 0.0352.

(also

RT.

computations
Airfoil

of 6.2 million.

Figure
Figure

of free-stream

and

and

at
drag

is compared
coefficients
the

overall

experiment.

Mach

contours

(Moo

-1.5

-1.0

...........

-0.5 ........... |
0.0........... d
I

0.5 ..........
|

1.0 ........... |
1.5
-0.2

0.6

o.i

0.i ol

oi

16

x/e
Figure

The
turbulence

,_ext

geometry

modeling.

a = 2.26 , and
at
(CL

these

flow

= 0.340,

one-equation
flow only

Pressure

flow

conditions

The

geometry

a Reynolds

number

conditions.

Figure

CD = 0.035),
model

(CL

and

11.

the

with

= 0.589,

provide

standard

of 9 million.
14 plots

comparison.

a much

NACA
Figures

12-14

12.

NACA

show

model

= 0.531,

removed

from

(CL

the

model

severe

computed

for

the

the

test

grid

and

one-equation
CD = 0.048),

equation

case

for

at Moo = 0.799,

on

solution
model
and

the

upper-surface

= 0.048).

Figure
Figure

more

0012

Gp distributions

Baldwin-Lomax

advection

(70

is the

coefficient

0012

0.8).

mesh.
11

13.

Mach

contours

(Moo

-1.51

-1.o]

: '_

.....7-_

::

io,
.............................
i

[ --

ORe-equation

Model

1.0...........
r "l ....B,t_in-t.om,x
1.5

.2 0.0 0.2 0._ 0.i

I........

0.i

1.6 1.2

x/c
Figure
The

improvement

wave

is moved

in the
forward

full

that

(ae,:p

we

have

2.86).

used

The

Pressure

model

20 percent

with and without


influences
obtained

the

corrected

discrepancy

wake
the

In this

section

flow.

Following

z momentum

simplified

aT

we examine

pressure

and

viscosity

by

the

FREE
model

arguments

the

effect
does

SHEAR
for the

of Tennekes

model

shock

with

experi-

advection
clearly
indicates
terms in this separated
flow.

correction

FOR
RT

upper-surface

agreement

suggested

angle-of-attack

MODEL

calculation

experimenters
of upper-surface
not

adequately

LAYERS

self-similar

turbulent

and

(ref.

Lumley

reduces

to the

13),

following

form:
UU_

Using

of attack

the one-equation

for an eddy

The

is in good

upper-surface
via advective

angle

the order-of-magnitude

equation

and

in lower-surface

ONE-EQUATION

comparison.

is dramatic.

chord

shock location
seems to indicate
that this
account
for wind tunnel wall interference.
TH_

coefficient

one-equation

almost

ment.
The solution
obtained
the importance
of upstream
Note

14.

similar

arguments,

equation

= (vtUv)

(7) reduces

--

--

(21)

to

--[v't(RT)vv

--

(vt)v(RT)v]

(22)

-(2 - c,1)kl k2
+ k2 _P
where

two
that

vt = r,c,RT,

P = vt(U v)2,

We are interested
terms
of equation
approximation

dissipation.

Wake

and

kl :

(2 -

V/g_.

in the one-equation
model
that
(22).
However,
it is worthwhile
can

flows

apply
provide

more

c,2) k2

generally

an interesting

than

in regions

example.
12

results
from neglecting
to gain an appreciation
where

In this flow,

the last
of how

production

equals

production

is zero

at the centerline,
that

but

production

dissipation

is equal

is not.

to dissipation

of a wake flow we can rationalize


the centerline
of a wake the first
vanishes

there.

the last

Also

term

wake,
and

analyze

the

for wake

becomes

that

many

The

primary

k -

from

valid.

modelers

question

in a simple

situation

flows,

away

the

centerline,

Fortunately,

the assumption

even

at the

centerline

neglecting
the last two terms in equation
(22). Clearly at
of these two terms is identically
zero because
production

approximation

we can
Lumley

note

identically.

the one-equation

Moving

c,2 = 2, which

still remains

wake.

in detail.

choose

as to the

Assuming

Following

the

would

overall

self similarity
scaling

remove

validity

of

of a turbulent

procedure

of Tennekes

we set

U: Vo + V.f(_)

(23)

RT = U'lh(_)

(24)

(25)

v, = vc,,RT = U.lc,,h(_)
where

_ = y/l,

equations

(21)

Uo = Ax -1/2,
and

(22)

I = Bx 1/2,

(with

the

last

fl(f

two

+ _f,)

and

U,

terms

<<

of eqn.

omitted)

in

we obtain

- (h_) 2] +

--If_lh

_;2

+ --

[hh_

= 0

(27)

0"_

where

UoB
flnumerical

on the

value

definitions

centerline

such

If h(_)

of/9

was

Us = [U that

]V -

is a known

evaluated

2A
from

U0 ],,_az, and

Uo]/]U

function,

=.08

U0]maz

equation

expression

be expressed

is then
in the

differentiated

following

experiment

is linear

in h_ and

and

I as the

Lumley,

distance

based

from

the

= exp(-1/2).
(26)

can be evaluated

in the

form

_ a_

(29)

Ire Ih with/9_f/ct,.

to replace

) h_ = P,(O

by Tennekes
a length

Equation

(27)

can then

form:

h_

where/91

(28)

by choosing

/ = exp _
This

Substituting

assumed.

(26)

= 0

O"c

The

(22)

+ c_,(hf_)_

Cp

fl_h_

[To are

/92 = fl

With

the solution

h_ =

P1 and

(30)

-Q,(O

Q, considered

known,

is

Ql(_)ezp[PlI(_)

13

-- Pl/(_)ld_"

the

resulting

equation

where
can

Plx

= f:

P1 d_.

be obtained

compare

By iterating

the

10 steps.

In figure

in about

it with

results

from

the

quadratures

constant

for

15 we plot
eddy

and

the

viscosity

h, a numerical

solution

of this

solution

as well

solution

equation

and

as results

from

experiment.
1.0

0.8-

.................X'""? ......I-_
\

_
i

az_.c_=la

| .... _2=i._1=1.,_
t ....... c,_,_t nu_t

I
I

0.6-

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

0.4-

iiii
iiiiiiiiiii
iiiii
iiiiiiii ....?
...............

0.2-

;"-..

0
1

Figure
The
our

15.

data

Comparison

points

one-equation

model.
for

The

the

curve
used

are

model

solid

wall-bounded
from

the

(in wakes)

Townsend
the

is based
flows

data

of experiment,

from
with

curve

values

eddy

viscosity

In this

section

Current

we consider

implementations

numerical

which

might

where

Rij

As

we will

produces

the

vRT(zi,j,

c,2 from

[_1,1,

see,

a system

_1,2,

our

...,

the

original

that

sections.
dotted

was

computed

using

"standard"

provide

The

curve,

flow.

the

departure

which

k -

best

results

of the

is from

the

solid
widely

IMPLEMENTATION
of the

one-equation

model

in two-

and

solver for the scalar equation


analysis,
implicit
unfactored
system

model.

operators
for the

is a prerequisite

We begin

for discrete
one-equation

three-space

in a

systems.
model.
dimensions

(decoupled
from the mean
schemes
will be considered.

for the

more

general

situation

factorization.

solution
Yi,j)

curve

wake

approximation.

one-equation

of this

matrix

define

of the

discretization

of the

behavior

include

We first

dashed

setting
by introducing
the notion of positive
us to construct
extremely
robust
algorithms

employ
an implicit
factored
ADI
flow equations).
In the following
Good

for self-similar

of parameters

in previous

NUMERICAL

rather
abstract
This will allow

The

of c,1 and

that

theory

14).

values

considered
than

and

(ref.

on the

is no worse

constant

vector

77. on a two-dimensional,

logically

rectangular

mesh

with
7_1 ,M,

one-equation

of ordinary

_2,1,

_2,2,

model
differential
-_t+

...,

with

_2,M,-..,

discretized

equations
M(_)_=
14

AN,l,

"_N,2,

advection

-..,

and

"_N,M]

T.

diffusion

terms

of the form
D_

(31)

where M(_)
advection
source

is a matrix
and

term.

diffusion

operator
and

We can

(possibly

construct

implicit
...... _

[I + AtOm(
or after

diffusion

1. M(_)

2.

with

explicit

schemes

the

positive

discretization

entries

of the

of

representing

the

form

)1( _

4_7
_) = At(-m(x')-"--_- " + D)'/_
'_'---*

(32)

rearrangement

for all 0 E [0,1].

onally

and

representing

matrix

__.+n+l

7_

[I + AtOM(R'_)]_

and

nonlinear)

D is a diagonal

As we will show,

which

guarantee

is a diagonally
dominant

M(-_)

"+'

has

zero

we can

the

with

row

This

will

be due

'_

(33)

approximations

for advection

of M(R'_):--

(M-type)

diagonal

'_) + AtD]_

numerical

properties

monotone

positive

sum.

design

following

dominant

matrices

= [I - (1 - O)AtM(-R

matrix.

entries

and

to the

use

M-type

matrices

negative
of the

are diag-

off-diagonal
chain

rule

entries.

form

of the

equation_
A well-known
elements
matrix

property

of the

of M-type

inverse

are

matrices

non-negative.

is that

is an M-type

matrix

and

non-negative
1 that

inverses,
the

i.e.,

left-hand-side

right-hand-side

R'_)]

consequently

[I+

matrix

AtOM(_n)]

-1 > O,

is also unconditionally

non-negative

for 1 > 0 > 0. Assuming

__,_+1

positivity
We

of _,_+1
now

2, it is a simple

_0

> 0, from

is guaranteed

the

whenever

stability

matter

properties

(see,

'_) +AtD]

for 0 = 1 and

= [I + AtOM(-_'_)]-'[I

investigate

0 > 0

of (33)

[I - (1 - O)AtM(_

and

have
property

of (33)
[I + AtOM(

The

they

We see from

(34)

non-negative

solution

update

under

for example,

numerical

Barth

II[/+ AtOM(-_=)]-'[Io_
15

'_) + AtD]-_

is a nonnegative

of the

and

condition

equation

- (1 - O)AtM(_

(34)

a CFL-like

operator.
scheme.

Lomax

< 1

'_

(ref.

Given
15))

properties
to show

that

We alsohave that
[[[I under

the

source

term

CFL-like

condition

(D = 0) and

like condition

(1 - O)AtM(_)][[oo

for

positivity.

Dirichlet

< 1

If we first

boundary

consider

conditions,

the

we have

system

stability

without

under

the

the
CFL-

for non-negativity

ll_"+Xl]=
=II[1
+ AtOM(_")]-_[I
- (1- O)AtM(_)]_"II=
.._+_

_<11[I
+ AtOM(_")]-_[I= I1[I- (1 -O)AtM(_)]I[=]IT_

11=

_<lf_"ll=
In the presence
estimate:

of the

linear

source

term

with

D > 0, we obtain

11_"111=< (1 + AtllOll=)lt_"ll_

the

following

stability

_ (1 + At m_(Oj))]l_'_ll_
3

This

result

is expected

because

the

differential

equation

admits

presence
of the source
term (with positive
coefficient).
We now turn to the actual
discretization
of the individual
rewrite

the

diffusion

terms

(v + -)v_(._T)
vt
Using

this

0(_Rr)

ot

We first

identity,

(assuming

equation

= 2(_+ _)v_(._r)

(13)

approximate

the

advective

(u + = (u + N[)/2,v

v v(_r)~

sort

in the

For convenience

terms

v.

O"c

by using

a standard,

first-order

accurate

upwind

+ = (v + Ivl)/2):

_
.._ct_,_j+l,k

_
+ fl._j,k
Y

+a_7_i,k+l

+ fl._j,k

+ %7_j-_,k
.

+ "/_'R.j,k-1

where

---

a_--AzUj,
_1

Note

that

this

discretization

- k,

we

(S 2 = \ 0_j + 0_ ] 0_, J

+V.V(._)=2(.+_)V_(.Rr)-lv

approximation

terms.

of this

v is constant)

(vv,). v(_r)

we rewrite

growth

7_ = ---u+
Az

automatically

J'_'

lv+

satisfies

previously.
16

both

properties

1 and

2 mentioned

The diffusive/antidiffusive terms are approximated by central differencing:

[
+

1 r(vt)j+l/2,k
O"

(7_j+1,_

- 7_j,k)

O" L

these

two

(r't)j-ll2,k

(7_j,k

- _j-l,k)

h;_2

By combining

Ay 2

Ay 2
expressions

we obtain

where

_ -A_1 2

vt
_-)j,-

[2 (v+

a_ - Ay1 _ [2(t,+-_)

(-_)

_+1/5,4]

J,k -(-_)j,k+_/2]

_
7d-

1
A-z2

[2 (t,+-_)

7:-

Ay1 _ [2(v+-_)

j,k -(-_)j-1/2,k]

_,_ -(-_)j,___/_.]

_=-(a_+_)

_: = _ (_ + _)
.Note
require

that

in this

that

both

It is important
result

a and

to realize

of poor

in a Taylor

form

we do not

7 be positive
that

grid resolution.

series.

have

if these

automatic
coe_cients
coe_cients

To see this,

For example

we have

assume

satisfaction
(property
become

of property

1 which

2 is automatically
negative

a smooth

this

variation

effect
of t't and

would

satisfied).
is entirely
expand

a
t't

that

(_,)_+1/:,_ = (_,)_,_+ O(A_)


and
1
Ay_

a_From
Az.

this

equation

If we assume

it is clear
that

vt >>

that

this

2v + (_)
coefficient

v, we obtain

the

J,_ + O(Ax)
is guaranteed
following

positive

restriction

a and 7:

(v_)_+_,_< 3(_)s,_, (_)___,_ <__


3(_)_,_
(_,)s,_+_< 3(v,)s,_, (_,)_,_-_ < 3(_,)s,_
17

for small

enough

for non-negativity

of

In our implementation
conditions
are violated
positivity.

Keep

of the algorithm
we limit the

in mind

that

We conclude
this section
for the one-equation
model.
the

near-wall

Figure
to the
wall

region

3 and

Baldwin-Lomax
shear

functions.
which

which
This

This
meshes

pressure

model

(y+

in the

removes

from

the

which

must

be used

wall

flow

remark
variable

solver

This
over

require
limit

much

of the

plot the law of the wall for the flat plate


to demonstrate
the solution
independence

grid

(as

previously

shown

wall

spacing

comparable

for accurately

estimating

formulations

y+ in the
of the

y+ = 0.2 as mentioned

stiffness

in

layers

to determine

resulting

formulations

from

in

damping

k - e equations
in reference

the

of the k - e model.

using several mesh wall spacings


with wall spacing.
i

refinement.

a mesh

is required

near-wall

less than

near-wall

31

boundary

implementation

spacing

in the

is usually

concerning
grid-resolution
requirements
RT was designed
to behave
linearly

only

< 3.5).

present

improvement

mesh

however,

gradient

we typically

is used
require

remedy,

(18)),

is a drastic

typically

better

with a final
Because
the

for zero

equation

the

we strictly
enforce this condition.
Whenever
these
amount
of anti-diffusion
added so as to maintain

9.

extremely

fine

In figure

16 we

0.5 < Yw,,tt

< 3.1

251.._
I
o,.-_q..,_,,_.,I._
................
i._
/ I .... L.mi_,_.us_yer
I _
__
I

/1-:?._oo.,
/i
J"
I I r(..,,)._.,
l_J_ i

204-"|

y'(x,ndl)=l.2

i'_........
"a_";

I
i

...............
l

.................
................
................
-

"'__-,

"_F_'!

.................
..............
i
................
i...............
1

0 '"" :,:.2.....
,_ ........
_ ................
I0-I

I0

I01

102

lO_

I04

Figure

16.

Sensitivity

of boundary

layer

to wall

spacing.

CONCLUSIONS
A one-equation
finding

an algebraic

has predictive
and refinement.

and

turbulence
length
numerical

model

scale.

The

properties

has

been

preliminary
of sufficient

18

introduced
results

that

presented

interest

avoids

the

indicate

to merit

further

necessity
that

for

the model

investigation

APPENDIX

This
the field

In this

appendix

- SUMMARY

gives

equation

OF

a complete

THE

summary

ONE-EQUATION

of the

one-equation

MODEL

model.

We begin

for RT:

equation,

we use

the following

functions:

1 =(c,, - c,,)jc;/,d
0,"c

vt =%(VfiT)DID2
#t =pvt
D1 =1 - exp (-y+/A

+)

D_ =1

+)

-exp(-y+/A

P =r,t

\ Ozj

+ cgzi ] Oxj

For all calculations

we have

used

1
) D2 +--exp(-y+/A
A+

the

constants:

%=0.09,

1. Solid

2. Inflow
3. Outflow

Wails:
(V.n
(V-

the

following

Specify
< 0):

RT
Specify

n > 0):

+ D, D,)(J-_, D,

exp(_y+/A+

t =0.41,

We also recommend

c_,

f_(v+) =c,_
+(1 - _)(--y
d_ 2
y+
+ _(-AT

3 v' \_

following
c_

= 1.2,

A +=26,
boundary

c_

= 2.0

A2+ =10

conditions

for (35):

= 0.
RT = (RT)oo

Extrapolate

< 1.

RT from

19

interior

values.

+) D,))

with

REFERENCES

1. Baldwin,

B.S.;

Separated
2.

Cebeci,

Mass
3.

T.:

AIAA

Maksymiuk,

AIAA

Johnson,

5.

and

pp.

L.S.:
AIAA

6. Coakley,
87-0416,
7.

Pulliam,

Reynolds

vol.

Hopkins,

87-0415,

January

1987.

L.S.:

A Mathematically

Turbulent

Boundary

87-0418,

Numerical

and

Algebraic

Boundary-Layers

pp.

Model

for

1978.
with

Heat

and

1091-1097.

Transonic

Airfoil

Simple
Layers.

Workshop

Turbulence

AIAA

Results

Closure

Journal,

vol.

Model

23, no.

11,

Flows.

W.;

N.N.;

E.J.;

M.W.;

J.; and

8, 1989,
and

Flows

about

Airfoil

AIAA

Paper

Flows.

and

Stanford

Tennekes,

H.;
Mass.

Townsend,

University

Press,

15. Barth,

T.J.;

G.:

M.:

An

9, no.

Horstman,
Lilley,

eds.,

and

A.A.:

Lumley,
London,
The

1985.

Turbulence

Turbulence

Models

vol.

Near-Wall

23, no.

k -

Evaluation

on Flat

Plates

6, June

1971,

C.C.:

University,
and

Models

for Attached

for Near-Wall

and

9, 1985,

_ Turbulence

pp.

Lomax,

Low

1308-1319.

Modeling.

pp.

AIAA

Flow

on Complex

Thermo

J.L.:
England,

over

Turbulent

Sciences
California,

A First

for

Predicting

and

Turbu-

Hypersonic

Mach

Plate.

1980-

993-1003.

Supersonic

Stanford,

Structure

of Theories
at Supersonic

a Flat
Flows,

vol.

Division,

Mechanical

1981,

369-377.

Course

pp.

in Turbulence.

The
1, S.J.

H.L.:

Kline,

Engineering

The

MIT

Press,

1972.

of Turbulent

Shear

Flow.

Second

ed.,

Cambridge

1976.
and

for

1985.

Journal,

P.:

Conference

G.M.

Number

Computations

1068-1073.

Transfer
vol.

81 AFOSR-HTTM-Stanford

Cambridge,

for Transonic

Navier-Stokes
July

January

AIAA

Moin,

pp.

Inouye,

Heat

and

85-0375,

Scheurer,

T.J.:

85-1587,

Reynolds

A Review.

Kim,

27, no.

Paper
and

Flows:

Journal,

Department,

Models

Transonic

Barth,

Paper

of Low

AIAA

Rodi,

and

AIAA

A.: Evaluation

AIAA

Cantwell,

of Viscous

D.J.;

Airfoils.

and

Rubesin,

Closure

1987.

Simulation

Jespersen,

Friction

Numbers.

14.

Viscous

January

Number

Skin

13.

T.H.:

of Turbulence

V.C.;

11.

B.

Pulliam,

King,

Controlled

lent

12.

6, 1971,

A Comparsion

T.H.;

10. Mansour,
Journal,

January

1987.

Separated
Patel,

Approximation

Turbulent

9, no.

Paper

T.J.:

8. Sugavanum,

9.

vol.

Separated

January

Circulation

and

Layer
78-257,

1684-1692.

King,

Airfoils.

and

and

Thin
Paper

of Compressible

Paper

D.A.;

for Attached

H.L.:
AIAA

Journal,

C.M.;

Arc2D.

1985,

Lomax,
Flows.

Calculation

Transfer.

Using
4.

and

Turbulent

Algorithm

Development.

20

NASA

CP-2454,

1987.

Nalo_l!

Al,lOn_kll

S,l_Oe AdmtnlU

Report Documentation

v,d

Page

_io,',

2. Government Ac,oession No.

1. Report No.

NASA

3. Recipient's Catalog No.

TM- 102847
5. Report Date

4. Title and Subtitle

A One-Equation

Turbulence

High Reynolds

Number

Transport

Model

Wall-Bounded

August

for

Flows

Code

8. Performing Organization Report No.

7. Author(s)

Barrett

1990

6. Performing Organization

S. Baldwin

and Timothy

A-90231

J. Barth

10. Work Unit No.

505-60
9. Performing Organization Name and Address

Ames

Research

Moffett

Field,

11. Contract or Grant No.

Center
CA 94035
13. Type of Report and Period Covered

Technical

12. Sponsoring Agency Name and Address

National

Aeronautics

Washington,

DC

and Space Administration

Memorandum

t4. Sponsoring Agency Code

20546-0001

15. Supplementary Notes

Point of Contact:

Barrett

S. Baldwin,

(415) 604-5072

MS 202A-1,

Moffett

Field,

CA 94035-1000

or FTS 464-5072
=

16. Abstract

A one-equation

turbulence

model that avoids

the need for an algebraic

length

scale is

derived from a simplified form of the standard k -e


model equations. After calibration
based
on well established
properties of the flow over a flat plate, predictions
of several other flows are
compared
dictive

with experiment.

and numerical

ment. The one-equation

The preliminary

properties
model

results

of sufficient
is also analyzed

presented

interest

to merit

numerically

indicate

that the model

further

investigation

and robust

solution

has pre-

and refinemethods

are

presented.

18. Distribution Statement

17, Key Words (Suggested by Author(s))

Unclassified-Unlimited

Turbulence
Computational
fluid dynamics
Fluid mechanics
19, Security Classif. (of this report)

Unclassified

Subject
20. Security Classif. (of this page)

Unclassified

Category-34

21. No. of Pages

23

NASA FORM 1626 OCT_


For sale by the National Technical Information Service, Springfield, Virginia 22161

22. Price

A02

You might also like