You are on page 1of 56

NASA

Technical

Memorandum

ICOMP-97-08;

CMOTT-97-03

Modeling

Tsan-Hsing
Institute

Shih,

of Turbulent

Jiang

for Computational

and Center for Modeling


Cleveland,

Ohio

Kuo-Huey

Chen

113112

Zhu,

and William
Mechanics

of Turbulence

Swirling

Flows

Liou

in Propulsion
and Transition

University
of Toledo
Toldeo, Ohio
Nan-Suey
Liu
Lewis Research
Cleveland,

Center

Ohio

John L. Lumley
CornelI University
Ithaca,

August

New York

1997

ICOMP)_)
\%_.,..,.= _%(_ /
InstRute

NationalAeronauticsand
SpaceAdministration

Modeling

of turbulent

Tsan-Hsing
Center

for

Modeling

Shih,

of Turbulence

and

Cedar

Jiang

Zhu

Transition,

Point

Rd.,

swirling

NASA

Park,

OH $_I$P

Lewis

University

Research

of Toledo,

Lewis

Research

Center,

22800

Ohio

Liu

Center,

John

Liou

Chen

Nan-Suey
NASA

William

ICOMP,

Brook

Kuo-Huey
The

and

flows

Cleveland,

OH _135

L. Lumley

Cornell

University,

generally

involve

Ithaca,

New

York

Abstract
Aircraft
mixing

engine
and

flame

axe unable

of these

in the

flows,

modeling

of many
and

This

a large

swirling
model,

amount
is devoted

flows,
one

the

general
for model
flow with

which

and

other

and

to the

resources

by
two

for the

dissipation

evaluation.
swirl and

stress-strain

the

work
e. The

relationship

One is a fully
recirculation.

CFD

of a cubic

developed

(Shill

and

rotating

model

developed

Lumley,
pipe

at

1993).
flow,

and

models

will

Using

turbulent

scalar

scheme

will

for turbulent
this

kinetic
report

flows
other

solution

flow.

UMIST.

the

the

ability

other

this

model

in this

for

of its

require

Also,

Two

that,

plus

swirling

for the

turbulence

stresses

stress-strain

group
one

fuel-air

because

scheme

combustor

Reynolds

equations,
cubic

this

Reynolds

to enhance

suggested

implementations.

of Launder's

turbulence

rate

(six

for a general

development

inspired

for any

viscosity
been

be considered

However,

equations

in order

eddy
it has

should

effects.

flows

that

Therefore,

scheme

transport
task

swirling

recognized

flows.

closure

is a difficult

of computer

turbulent

been

curvature

moment

to solve

Reynolds

was

order

needs

only

long

swirling

a second

second

variances),

report

It has
model

of rotational

complicated

require

and

stabilization.

to appropriately

modeling

fluxes

combustors

have

type

is based
been

is a more

of

energy
on
chosen
complex

k
a

Co_e_s

Introduction

Cubic

Modeling

Reynolds

stress

of turbulent

swirling

3.1

Rotating

pipe

3.2

Complex

swirling

Conclusion

and

model

flows

flow
flow with

recirculation

discussion

8
8

References
A

Appendix:

Development

Appendix:

Equations

of a Cubic
in a General

B.1

Equations

in tensorial

B.2

Equations

in a general

B.3

Another

form

Appendix:

Equations

C.1

Mean

equations

C.2

Nonlinear

C.3

Another

Appendix:

ciated

14

System

22
22

form
coodinate
cubic

23

system

model

in Cylindrical

26
28

Coordinates

28
of turbulent

stresses

31

rij

42

of rij

Equations

fuel-air

with

in engine
eddy

ing length
eddy

form

Coordinate

Model

for Axisymmetric

46

Flows

Introduction

For better

are

part

of the

Turbulent

the

mixing

flows.

combustor
viscosity
models

viscosity

and

design.
models
and

model

flame

Therefore,

Common

which

appropriate

in a combustor,

modeling
turbulence

include

k-e models).
is not

stabilization

accurate

of turbulent
models

zero-equation

However,

it has

for predicting

and
long

used

a swirl
swirling

swirling

flows.

calculations

models

recognized

that

In fact,

asso-

is important

in engineering

two-equation
been

is generally
flows

(e.g.,
this

the

mix-

type

deficiency

of

of eddy

viscosity

a fully

models

developed

approximately
produce

rotating

avoid

has

been

this

kind

of mean

rotation

second

moment

transport

and

variances.
second

in combustor
Recent

with

development
and modern
The

model

which
of mean

two

formulation
undetermined

stresses

the
and

in this report

for many
of swirling

from

the coefficients

(Imao,
data

on

In this report,
model

the

pipe

Harada,

(Roback
mean

cubic

computer

Scalar

resources

implemented

Appendix

models

B gives

relationship

of the mean

with

a general

This

gradient

with

we need

a quadratic

constraints

works

However,
(1995)

be rectified

and

of Reynolds

model

our

appropriately

Launder

to trun-

formulation

quadratic

is not

not

relation.

velocity

application,

separation.

at UMIST.

polynomial

quite
recent

predicted,

pointed

out

that

by introducing

Reynolds

stress-stain

used in Shih

et al's

just

formulation

quadratic

model

flows. Modeled
k-s equations
are used together
for mean flow calculations.
The first test flow is

about
test
These

its own axial


two flows

In Appendix
equations

axis with various

flow is a more

comparisons

the

development

Cayley-Haanilton

velocity

method

CFD

of a tensorial

This

can

1993).

stress-strain

in terms

realizability

will be reported

appendices.

Reynolds

the

flows

et

to both

analysis.

swirl

a similar

The

Further

suggested

from

second

components.

al,

(Craft

(1995)

group

1983).

swirling

interest

power

viscosity
model
relation."

The

method

turbulent

practical

the

terms

a practical

for any

including
that

by using

Johnson,

from

of great

by using

flow rotating

there are four

is described.

other

bring

equations

stresses

theory

Launder's

1996).

and

velocity
results

the

in simulating

a generalized

Lumley

from rotating
pipe
stress-strain
model

developed

Itoh and

computational

show

of the linear eddy


to the stress-strain

and

recirculation

and

flows

flows

we retain

of fully

scheme

complicated

plus

successfully

models

potential

up to the sixth

coefficients

complex

and the measured


data
with the cubic Reynolds

from

rapid distortion

the finding

determine

rotation.

simulate

many

of the large

varies
models

closure
to

stresses

not been

on a general

Obviously,

Zhu

In this report,

that

is based

terms

relevant
from

calculations
verifies

have

transport

It is derived

Shih,

three

"the weaknesses
quadratic
terms

because

stress-strain

of their

for the Reynolds

coefficients.

successfully
which

Reynolds

of these models are


combustor
design.

contains

a result

and

pipe

body

ability

solving

modeling

viscosity

order

of its

requires

in the

a solid

a second

six Reynolds

models

because

expression

this polynomial.

determined

equation

turbulence

gradients.

eleven
cate

transport

this

by

eddy

describes

because

involve

velocity

however,

models,

However,
which

demonstrated

swirl

(r2),
which

flows

of this complexity

modeled

developed

general

velocity,
viscosity

curvature.

equations,

and evaluation
aircraft
engine

velocity

Measured
radius

of swirling

in nonlinear

is an explicit

1995).

of eddy

flow calculations

only

be analytically

flows.

developments

flows

and

moment

can

of the swirl

modeling

Because

swirling

for combustion

flow (Fu,

profile

for

flows

of the normalized

of deficiency

suggested

required,

pipe

linear

effects
fluxes

swirling

as the square

an exact

To

for

both

between

complex
have
the

rotation

flow with
detailed

rates

swirl

and

experimental

experimental

data

and

in detail.
A, the derivation
in a general

of the proposed

coordinate

system,

cubic
which

will

be

useful

for studying

axisymmetric
Therefore,

flows

Cubic

A cubic

list

both

In terms

forms

stress

--pUiU

model,

used

A. The

for convenience

j =

--

velocity

-_pk

The

",j" means
i.e., Skk.

The

_q

(Uk.iUkj

invarints

coefficients

1] 2

C_=

strain

and

cubic

model

system.
flow and

derivative

and

rotation

for Reynolds

with

Sij and

fhj.

flows,
of mean

Here,

stresses

we

is

1"I3

'ij

(1)

_--1"I2_iJ_]3
}
respect

II3 (which

As

2113

to j. Is is the first
appear

Ui,kUi,pUp,

in Eq.(A.1))

principal
are

defined

invariant

of

as foUows

(2)

are

or _ = pG/.
'

rates,

swirling
in terms

_ii)

Uk,jUk,pUp,i

+ Ui.kUj.k

Ui,jUi,j

of turbulent

can be expressed

k(k+ v_)
kU*

4.0+

coordinate
for a general

Ui.,U,.,)

II1, II2 and

#T, A3

k2
.T = pGY.T,
1

model

- 3

a tensorial

three

the

[Uk,iUk,pUp,j

-1-Is2

example,

system.

for modeling

cubic

Ui,jUj,i

equations

For

applications.

I.tT

-_Is

1"_1 =

systems.

in a cylindrical
the

coordinate

study

of mean

gradients,

pk4
+ A 5--_1

Sii,

in this

resultant

of their

+ A3-_
pk32 (Uk.iUkd-

where

studied

D, we write

in a cylindrical

Ui,i, or in terms

of mean

coordinate

stress model

in Appendix

gradients,

curvilinear

conveniently

Appendix

flow respectively

Reynolds

velocity

C and

Reynolds

is developed

in various

be most

in Appendix

an axisymmetric

will

flows

f_,=Eq.(22),

or

Eq.(26)

(3)
(4)

As_
E

1-}C_

(5)

A3 =
3k2-*

0.5+ _a

S*

1.6 #T
A 5 =

(s)

pk' 7(s*)2+ (fl*)2


3

in which

As=v_cos,

=_arccos(v_W),

W*-

(7)

(S.)S

(8)
The

model

coefficient

C t, is also constrained

Cj, < y

where

IIs

is defined

In terms

of mean

-puiuj

=-

and

C,

<_

by the following

As

conditions:

(9)

+-_IIsAs

in Eq. (12).

strain

2pk6ij

and

rotation

+ p,T2S_

rates,

+ As pks

Eq. (1) can be written

(Si_flkj

as

- _ikSki)
1
(I0)

where

(11)

IIs

is the

IIs

Note
SspS_

second

invariant

of S/./ defined

as

(12)

that

in the

in which

It should
form

principal

of #r

also

above
each

equations,
term

contains

be mentioned

= pC.

k2 for high

that

Skk means

Su + $22 + Saa and

a summation
the eddy

turbulent

viscosity

Reynolds

operator

#T in Eq.
number

S_k means

on the

StvS_,l + S2pSp2 +

subscript

(3) will become


k2
flows (_T >> 1).

"p".
the

standard

Modeling

of turbulent

The

model

proposed

in the

this

study.

The

flow

1996).

This

first

flow was

rotating

rates

swirling

flow with

encountered

3.1

A fully
lence

developed

combustor.

pipe

flow provides

for modehng

pipe

flow

however,

for the

separation

swirling

model.

The
and

of swirling

(Imao,

a pipe

(Roback

fail to predict

a very

this flow. In fact,

of a pipe

for a fixed

mass

for a fixed

standard

eddy

In a fully

developed

derivatives,

0/00,

components

are

flow

second

flows

Itoh

and

with

various

1983),

in

Harada,

flow is a more

Johnson,

axial

complex

which

are

at

Orpk

Orpe
--d -=

axial
drop,

viscosity

models

zero,

axial

pipe
and

velocity

velocity.

Equations

-- _-+_
Or2pW

the

is often

U and

_0

any

eddy

radial

axial

the axial
reduce

viscosity

will increase
changes

velocity

total

"+_I_

drop.
mass

In
flux.

O/Ox,

and

= 0. The

(or swirl)

the

non-zero

velocity

azimuthal
velocity

= rf_, where

(13)
Or'Far

. OrW1 _ 20__

#T

For

at all.

gradients,

tangential

model

rate.

pressure

the

eddy

experimental
demonstrate

rotation

the

turbu-

used

for this flow are

- 0r L(_ + _T)r--o7-_
]

[(

rotation
no such

all the
the

with

will strongly

axial

so is the

that

the

commonly

(_+_)_-_;+

0 [

significantly

show

flow,

case for checking

previously,

one can show

rotation
the

test

for a rotating
pipe flow, while
rotation.
Experiments
further

flow changes

pressure

turbulent

the

angular

flux,

clean

flows. As mentioned

a solution
of solid body rotation
that the flow is not a solid body

words,

f_ is the

case

and

to model

the characteristics

However,

test

rotating

development;

engine

rotating

models

example,
other

a critical

ability

will produce
data shows
that

developed

for model

recirculation

will be used

pipe flow

model's

viscosity

section

is a fully

in an aircraft

Rotating

previous

used

is still

swirling flows

0r [(_+ _)rw]

0--'-7-

(14)

(15)

r'_r

+ rPk

- rpe

r_r

+ c, lfl rP_fi - c_I,_ --F- + _o

rp 2

[Z#T r

( OS_

\ or /

(16)

or

OrpE_
Ot

c9
Or

[(

I_ + P/r

r-_r

4" Clfl

rpS

E -- C2f2

rpe 2
k + _

##r r (os_
+ -- p

\0r/

2
(17)

whereS= _=,/(v_2 V ka_] -t- (aw

w)

and

Eq. (1) or Eq.

to,,

from

the proposed

T=_ --

model,

nonlinear
(10),

parts

of turbulent

stresses,

T=_

are

To,

The

cubic

2. The

(18)

s--j- w\o,. /

production

rate

(19)

of turbulent

kinetic

_3 T _

energy

Pk is

(20)

L_aT/

where

k 2

_r = pC. /.-;,,f.

[1

exp(-alRk

s_= 1,
and
used

(21)

particular

aaR_

(22)
(23)

""s'11/2
- as_k_]

s_ = 1- 0.22_p (-R+2/36)

al = 1.7 10 -3,
in this

report

as = 1 10 -9, al = 5 10 -1,

are

modeled

standard:

k-e

different
formulations
1995) is used together

Rk = px/ky/#.

ak = 1, a_ = 1.3, Cel = 1.44 and

equations,
proposed
with

the

model

by various

coefficients
researchers.

and

Other

damping

For example,

model

C_2 = 1.92.
function
if Eq.

constants

Depending

(17)

on

f_, may

have

(Shih

et al,

(24)

then

Cl=max

and

_,

0.43,

5+r/

'

C2=1.9,

77

fl are

.f_, =1

- exp {-

.fl = 1-exp{-

(air

(a;R

+ al2R 2 +ataR

a2 R2

-4- a3.R 3 -b aaR

-]-a_R

4 -'k
4

asRS)

+ a;RS)}

(26)

(27)

.f2 = Eq.(23)

(28)

where

k 1/2(k + v/_) 3/2

R =

(29)

V_

and

a1=3.3.10

-3 ,

a4 = -3.6
F

a t=2.53,10
a 4' =

From

Eq.

solid

body

is the

(14),

it is easy

rotation,

does.

Equations

1 - 3 show
are

also

stress

increases

the

kinetic

3.2

A confined
are ejected
the

duct,

often

into

viscosity

and

model

The

and

body

swirl

pipe
axial

profile,

pipe

and

The

flow.
axial
and

model

code.

Figures

compared

k-s

eddy

with

viscosity

N is defined

Reynolds

number

viscosity

based

model

the

present

rotation

on the

pipe

profile

towards

it reduces

R
will

Therefore,

In contrast,

velocity

and

viscosity

parameter

k-e eddy

of

models

flow.

(17)

standard

rotation

of the pipe.

of the

velocity

(15)

the standard

effects

of the wall

by a parabolic

the

the

a solution

stress

like an eddy

Eqs.

from

figures,

the

rotating

solved

on the

changes

velocity
Reynolds

just

with

velocity

all the

will produce

swirl

develed

results

figures,

rotations

velocity

is the

rotation,

In the

in these

model

any quadratic

accurately

cubic

swirling

the

flow

coaxial
general

an enlarged

with

has
cubic

flow:

it

a parabolic

the relative

turbulent

jet was

experimentally

features

of the flow. At the

duct.

recirculation

in an aircraft
plane
model.

from

recirculation

Besides
bubble

engine

combustor.

the cubic

model

Solutions

5 compares

the

were obtained

calculations

studied

an annular
is created

(Liu et al, 1996) and the other is FAST-2D


used in this calculation.
Numerical
results
Figure

body

et al (1996).

capture

W,_jz

that

TO, for a fully

be easily

of axial
can

viscosity

where

of solid

present

non-solid

an axisymmetric
eddy

-7 ,

k /U 2.

a central

observed

eddy

U,_ is the average

effect

swirling

4 shows

= r/R,

As shown

centerline

Complex

Figure

where

it maintains

any

for comparison.

model

energy

can

by Imao,

the

a 3=6.55,10

(31)

component

of the

included

missed

that

a solution

R is 20000.

Reynolds
shape,

to the

results

as N = W_au/Um,
on U,_ and

to show

(13)-(17)

measurements

(30)
!

It can also be shown

produce

the

-7,

a s' = 8.3 10 -12

pipe.

will also

a3=6.6.10

-5 ,

i.e., W/Ww,u

of the

no contributions

totally

a 2----5.7,10

-9 ,

they

model

-3.6.10

radius

-5 ,

as = 8.4 10 -12

-3 ,

have

the

a2=-6.10

* 10 -9,

and

an inner

jet and

inlet,
separation

by the

due

swirhng

In this

figure,

is compared

with

Johnson

This

codes.

jet

expansion

of

flow feature

is

velocity

one from

(1983).

an annular

to sudden

flow.

calculated
the

by two Navier-Stokes

(Zhu,
from

of the

by Roback

vectors

the standard
One

in
k-_

is CORSAIR

1991). Eq. (16) and Eq. (17) are respectively


the two codes are quite close to each other.

centerline

velocity

using

a standard

k-_

eddy

viscosity model (SKE) and the


velocity

indicates

of central
much

recirculation

better

bubble

the central

than

from

quite

does

the cubic

the comparison

present

cubic

well,

the SKE
model

but

of calculated

and

model
rotation

region.

predicts
which

This study

shows

tions

the

have

CFD

discussion

that

nonlinear

potential

The

applications

further evaluations
the model's
validity
The

proposed

tions,

yet

best

velocity

the

cubic

Reynolds

stress

combination

point

the recirculation

SKE model.

Figure

at x=51mm.

6 shows

Both

models

scatter.
However,
significant
models have been found in the

velocity

the cubic

swirling

report

engine

model

needs

stagnation

in Fig. 4 that

stress-strain

turbulent

aircraft

negative

the strength

profile

model

at x=305mm.

shows

a non-solid

SKE
body

observation.

Reynolds

in this

swirl

whereas

the rear

profiles

The

predict

models

flows

appears

encountered

simple

industry

with modeled

and

can

further

is particularly

be combined

studies

and

with

k-e equa-

in aircraft

numerically
the

existing

k-e

engine
robust

interested.

against
other flows axe needed in order to determine
and to seek possible
further improvements.

cubic

the

the

7 shows

experimental

proposed

in which

is also reflected

data.

models

predicts

axe within
experimental
profile between
the two

to simulate

model

both

model

mean

rotation,

with

and

This

the experimental

that

than that from the standard

Fig.

body

is consistent

Conclusion

combustors.

solid

present

measured

For example,

a nearly

with

It is seen

the

model.

is larger

give reasonably
good profiles
which
differences
in the tangential
velocity
downstream

model

recirculation.

in

However,

flow range

model

of

equa-

evaluations.

The proposed
cubic model appears
however it requires
about 15% more

the simplest
among other cubic or higher order models;
CPU time than does a linear k-_ eddy viscosity
model for

a general

flow. We expect

2D axisymmetric

or fifth

order)

increase

and

is used,
the

swirling

then

model

the

may

that

CPU

time

for calculating

become

very

costly

for the

if a higher
Reynolds
calculation

order

model

stresses

(e.g.,

fourth

will significantly

of a general

3D swirling

flOW.

Acknowledgements
This

work

Professor

was supported
Theo

G. Keith,

by ICOMP

TASK

Jr. for his advice

YOM5120.

and

useful

The

authors

would

like

to thank

discussions.

References

[1] Craft T.J., Launder B.E. and Suga K., 1993, "Extending
models
through
the use of deformation
invariants
and

the applicability
of eddy viscosity
non-linear
elements."
5th IAHR

Conference
1993.

on Refined-Flow

Modelling

and Turbulence

[2] Fu S. and Wang C., 1995, "Second-Moment


Frame."
Proceedings
of the Mathematical
Daignji

and Yutaka

Miyake,

Japan Society

Closure
Modelling

Measurement,

[4] Launder,
B.E., 1995, "An introduction
to single-point
ERCOFTAC/IUTAM
Summer School, in Stockholm,

7-10 September

Modelling of Turbulence
in Non-Inertia
of Turbulent Flows. Edited
by I-Iisaaki

of Computational

[3] Imao, S. and Itoh, M., 1996, "Turbulent


characteristics
Int. J. Heat and Fluid Flow 17: 444-451.

Paris,

Fluid Dynamics,

95-110.

of the flow in an axially

closure methodology,"
12-20 June, 1995.

rotating

Proceedings

pipe,"

of the

[5] Liu, N.-S. Quealy, A., Kundu,


K., Brankovic
A., Ryder, R. and Van Dyke, K., 1996, "MultiDisciplinary
Combustor
Design System and Emissions
Modeling,"
NASA 1996 CAS workshop
Proceedings,

Ames

Research

Center,

NASA,

Aug.

13-17, 1996.

[6] Reynolds,
W. C., 1987, "Fhndamentals
of turbulence
for turbulence
Lecture Notes for Von Karman Institute,
AGARD-CP-93,
NATO.
[7] Roback, R. and Johnson,
B.V., 1983, "Mass and momentum
with confined swirling coaxial jets," NASA CR 168252.
[8] Shih,

T.-H.

Comput.

and

Modelling

[9] Shih, T.-H.,

Lumhy,

J. L., 1993,

"Remarks

on turbulent

modeling

turbulent

and simulation."

transport

constitutive

experiments

relations."

Math.

18, 9-6.
equation

model."

[10] Shih, T.-H., Liou, W.W., Shabbir,


A., Yang, Z. and Zhu, J., 1995, "A new k-6 eddy
model for high Reynolds number turbulent
flows," Comput. Fluids 24 (3) 227-238.

viscosity

Comput.

Zhu, J. and Lumhy,

Methods

Appl.

Mech.

J. L., 1995, "A new Reynolds


Engrg.

[11] Zhu, J, 1991, "A computer


program
flows with
complex
boundaries,"
Karlaruhe.

125 (1995)

stress

algebraic

287-302.

for numerical simulation


of two-dimensional
Rep. 690, Institute
for Hydromechanics,

incompressible
University
of

1.5

1.0

0.5

0.0
0.0

0.2

0.4

0.6

0.8

1.0

dR
Fig. 1. Axial velocityprofilein a rotatingpipe

1.0
[] N=0.5, exp. Re=200()0
0 N=l.0,exp
- N=0.5, Cubic model
- N=I.0, Cubic model
....

/,']
./'_')/"
.fJ f
,x..
_
_

- N=0.5, 1.0 eddyviscositymod,eJf-"


// _
N=0.5, 1.0 Laminarflow /_-/
X /

I
I
I

E 0.5

._/"

0.0
0.0

0.5

dR
Fig.2. Tangentialvelocityprofile
in a rotatingpipe

10

1.0

0.040
............N=O.O,Cubic model
-- -- --

0.030

N=0.5,
N=1.0,

Cubic model
Cubic model

N=O.O, 0.5, 1.0, std eddy viscosity

mod_

//
o:" /A

..-."//
,."

E 0.020

--" //
..,ooO
._*

jj

Js

J4

:T
0.010

......
-'--

0.000

0.0

-"

0.2

0.4

0.6

0.8

1.0

r/R
/c
Fig. 3. Effect of rotation

Fig. 4. Velocity

on turbulent

x=25

x=51

x=102

x=152

(unit: ram)

x=25

x=51

x=102

x=152

(unit: mm)

vectors

in an axisymmetric

plane.

11

kinetic

(a) from present

energy

model,

u_

(b) from SKE model.

Centerline

Velocity

1.5

X
lt

Present

Exp.

0.7 1_
0.3

---

model

data

sr_,,o_l

-0.1
-tl.5
0

10

20

30

40

x (cm)
Fig. 5. Centerline

1.o

in Roback

l-o

-f

velocity

and 3ohson

flow

l-o
I
i

o.8

IP

o.8

O.8
P

0.6

o.6

0,4

0.6

oA

J
o

0.4

01
I

x_imm

-i

o.2

o.2

O.2
x=$1 _

x=$t mm

o.o
-1.0

O-O
IJ
(,n/s)

2.0

Fig. 6. Mean

o.o
4.4

:
0.0

velocity

'
0.4
v {m/J)

profiles

12

'
0.8

0.0
4.4

J-

O.8
,e,(mu's)

at z = 51ram

1.0

--

x=30.5cm

r_ o.5

,0

-0.5

0.0

w (m/s)
Fig. 7. Tangential

velocity

13

profile at x=305mm

Appendix:

Development

A truncated
general
be written
as

-p_j

cubic

of a Cubic

turbulent

Turbulent

stress-strain

Model

relation

from

Shih and

Lumley

(1993)

can

= - _pk _ + C_pT
pk' (Ui,,_Uk,t+ ut,W_,i _2-_nl ,hi)
+ c 1--_+

pk3

2--_-

(U,,kUt,k

_II21 ,q)

- ]lII

"pk3(u',iU'j

_it)

+ C'Pk' (Ui'kU_"U"k+ Ui"V"'U"k-- _ IIs *i')7

+ c _7Ok'
The six model
we consider

coefficients

This

the

rotating
write

was

the
model

pipe

model

flow.

t = -2pk'it

2-fi-to o

+
+
+

Cs by using

strain

pk3

,, t_(2*)

_-fi-toit

f_!2..)

- --,j

shear

flow, and

energy

Shih

and

and

rotation

pure

shear

rates,

+f_,,

(2,)

second
data

flows,

which

Schwarz'

will allow

us

is to deter-

of a fully

it is more

First,

reMizabil-

and

procedure

developed

convenient

to

as ill the following:

_!2.)

- --,t + s;k_,t - niks?,t)


+

.-, o,_.,
a*ikkj
c_

C(2*)n

--

flikSkm_'_mt

0(2.)

l3 a_,s,=n_6i t )
1

where

S_ = &t - gS_ _t,


s!._')
, = -

apply

components

et al (1995),

(73. The

procedure.

- S_kflk_+ fhkS_j)

2c4ak_
(s}] *, -- o(_')..
g
ik
aLkJ
Pk4(

strain

+ Ct-fi--.,-.i

following

the experimental

P k3 2(,q(2.)

+ C._2S_

+ C pk3 ,,.,(2.)

and

of C_, C1, 6'2 and

the pure

of mean

positive

(1987)

(A.1)

by the

flow and a pure

to ensure

by Reynolds
C4 and

To anMize

in terms

strain

stresses

coefficients

2 'it)
- -glI_

will be determined

a pure

suggested

coefficients

Eq.(A.1)

--pulu

cases:

on the Reynolds

to determine
mine

in Eq.(A.1)

two extreme

ity constraints
inequality.

+ u_,_vk,, v,,,

5'}_ *)=

_2

S_

SilSlt

--

l-S_k_ij

'3

--itf_(2*) =

_-_i2t _. _il_tt

14

nit2

--

X a2kk_it3

(A.2)

1 U
S_j

"-

SimSmlSlj

Note that

._,(2.)

Si_

(2*)

5'ij , -'ij

7(

i,j

._,(3.)

, _ij

and -ij

--_ Uj,i)

1
_'_ij = -_(U,j

are all traceless

tensors.

-- Uj,i)

(A.3)

Using Cay]ey-Hami]ton

relation,

S3 - IsS_ + IIsSij - IIIs6ij = 0

.q(3,)
"ij
can be expressed

in terms

(A.4)

of quadratic

and

linear

terms

as

(A.5)

S_ *) = _so_jr
c,(2.)
_ IIsS_

where

Is,

IIs

and

Is = Sii ,

Using

IIIs

IIs

Eq.(A.5),

are the three

invariants

of Sij:

3s.sj5 2s )

= -_

we may

write

Eq.(A.2)

2 k 6_i + C_,_2S_
= --_p

-pu_uj

principal

as

Pk3"_(2*)
+ 2A2-_-_zii

-"
Pk3_(2*)
+ 2A1--_--aii

+A,p___k23
,
-2A5_

(_'_ikSk_n_'_rrtj

(A.6)

._34i'o(2,),-,
.,,,,-

l _'_klSlm_'_mk'iJ3 IIsS_

(2,))

(A.7)

- IsS!-2")_tz
]

where

1
A2=_(2C1-C2-C3)

A1 =

(2C1+C2+C3),

A3=63-62,

A4

A result

from

bulence

should

way

a rapid

(C5-

distortion

not be affected

is to eliminate

the

pure

theory

analysis

by a pure

mean

rotation

(A.8)

As'-(C4+Cs)

C4),

term

(Reynolds,1987)
rotation.

states

that

this

result,

To satisfy

in Eq.(A.7),

i.e.,

A2 = 0, which

isotropic
the

tur-

simplest

indicates

that

2C1 = C2 + C3.
To determine
Under

this

puiuj

In principal

the model

coefficients,

let us first

consider

a pure

strain

flow, in which

f_ij = 0.

situation,

= 2pk6,j

axes

(C_, -

k2rr
A _Pk22c*
_
kr A I Pk3o(2.)
-_--s_s]--_"_ij
(A1 + 7-s._sl-fi-,.,-,,j

of S_j, we may

write

(see Shih,

15

Zhu and

Lumley,

1995)

(A.9)

=
S;j

where

[10 i}
_L+__
02

0
0

a and

S_1,

_
2_

b can take

on arbitrary

-ij

values.

[1:2/
0
0

Then,

_n
0
c(2")
_ 1-__bb
2

x+_

one may

(A.10)

write

(A.11)

If we define

S*=_,

from

Eq.(A.10),

-_2

(A.le)

we obtain

s* = ISx,I
Therefore,

/o(2,) _,(2,)
= VOii
o_i

S(2.)

-_ , s(=*) IS_

Eq.(A.11)

may

2 k
(C_= _p
_

be written

+ b=

(A.13)

as

k--22IIsAs)Pk22s*_f
c
c

2
V3 + a 2

+ k

Pk3

2. /

(A_ -_sA_)--_-2S( )W--V-_(A.14)

Since

u 2 k 0, we must

k2
-_IIsAs)

1 - (C_ -

require

the following

k S* / 3 +18 a 2

(At

inequality

for any

large

+ k[IsA5)_ k2./18
s(= )

S* and

_3 + b2 -> 0

S (2.)

(A.15)

If we write

k2II

(C.-_

then

we must

k . [ 18
s)_S V3+a2-a,

k
-kS
(AI+_IsAs)--_S

(2.) [ 18
V3+b2-/3

require

a +/3 _< 1

while

(A.16)

(A.17)

we write
k 2

( C_, -

-_ IlsAs ) =

kS*

a
_

'

(A_ + kIsAs)c

-_V3-_

_ _
k2S (2.)[
18

_
16

V3+_

(A.18)

Following
which

Shih,

Zhu

indicates

and

Lumley

(1995),

C2 + C3 = -k-IsAs.

for simplicity

Then,

a must

we set/3

be less than

= 0, i.e.,
unity,

A1

= -klsA5,

i.e.,

k_
(C_-

-_IIsAs)

<_-

(A.19)

kS*
As_
g

where

As

(1995),

and

-------_
3 +18 a

equals

or Reynolds

be

can

using

calculated

A,= v cos,=
Eq.(A.19),

C_, <
-

relations

[see Shih

et al

(i987)]:
,

From

following

the

C_ can

1
kS*

(A.20)

arccos(V W*),
W*=

be written

as

k2
+ -_IIsAs

(A.21)

As_
g

Now,

let us consider

a pure

shear

flow, in which

there

is only one non-zero

component,

U1,2,

i.e.,
Ui,j =

In this

case,

Sn

1U
= ill2 = _ i,2. Under

0
0

this

0
0

0
0

situation,

we obtain

from

Eq.(A.7)

-2
,_ A Pk3 pu 2 = -_pk + ,_3-_-5"i2_i2

(A.22)

2
P_22 = -_pk -

(A.23)

1u2

2As

S12_12

= -2c.Pk2sl2

(A.24)

Note

that

in Eqs.

that

A3 must

(A.22)-(A.24),

be positive

since

the condition
the

shear

(Al+'2IsAs)
g

UL2 will make

17

= 0 has been
u_ increase

and

used,

and

u_ decrease.

note

also

Applying

Schwarz'inequality, (u-_) 2 _<u_

A3 <

u 2, to the

above

equations,

we obtain

a constraint

for A3:

1 - 9C_--_$12S12

k2

(A.25)

3_-$12_212

Noting

that

expression

(S*) 2 =

2S12S12

for A3 may

and

fFS*

be written

= 21212S1_ for the

pure

shear

flow,

a generalized

as

A3 -__

3k 2

c02+ 5_za s

C0>_0

value

of A3,

(A.26)

where

To

ensure

following

a positive

real

condition

for any large

values

the

coefficient

Cv

must

be

also

restricted

by

of S*:

Cv < y

The

(A.28)

formulations

turbulent
to the

for C_, and

stresses.

However,

coefficients

To determine
In this

case,

appear

in the

_"=-

C4 and

A4

A3,

and

As,

two

components

only
mean

the

(A.21)

and

(A.26),

will

ensure

realizability

determined,

which

developed

rotating

are

Cs, let

of the

us study

non-linear

i.e., Eqs.

(13)

a fully

of

related

part
and

of turbulent

(14),

which

stresses,

pipe

Eq. (14)

flow.

_-,, and

_'a,,

axe

(A.29)

_
_

_--_- Wt, o_ ] +
integrate

Eqs.

As axe left to be further

or C4 and

flow equations,

7-7-

i.e.,

A4 and

Cs by Eq.(A.8).

TOT
=--'_-2_-" _

Now

the

for the

_
velocity

W)]r

18

component

(A.30)

at a steady

state

to obtain

0
Experimental

data

of the pipe,
obtain,

show

W_u

for high

that

Wwat-----_t
"_"

is the wall
turbulent

swirl

for a large

velocity.

Reynolds

Insert

(A.31)

range

this relation

of W_u.
into the

Here, R is the radius


above

equation,

we

numbers,

(A.32)

If we write

then

Pk4_

47

from

T 2

= a'

ur2

,.-R_,, _o,,

Eq.(A.32),

L"R,"""+ _0r/ j

pc,

we must

require

a'+;3'

.._ 1. The

coefficients

= fl' Pc

C4 and

(A.33)

Cs can be expressed

as

a' pc
C4"--

Pk42

-_3
In a fully

R4

Cs=

R4

developed,

r2W 2

fl'

rotating,

--2
_3 [
pipe

flow,

(A.34)

Pc

+ "--''\
0r ] J

we find that

we obtain

expressions

for C4 and

relations

hold,

(A.36)

Cs as follows

a'Pc
C4 =

following

(A.35)

_ _" + t 0r/ j = _ (Ts;Ss,5


+
Finally,

the

(A.37)

pk 4 1 (S.)_ _ ,'_._21
z3 2
/3' _T

(A.38)

05= pk'_(7(s'y+
(a')2)
e3 4
19

From

the

calculation

appropriate

(i.e.,

of rotating

pipe

flows,

we find

that

the

following

coefficients

seem

we set a' = 0, j3' = 1.6):

1
C_, =

kU*

(A.39)

4.0 + As_
g

C, =0

(A.40)

1.6/_T
Cs = pk 4 7(S.)2
+ (a.)2

(A.41)

e3

where
(A.42)

Equations
Now,

(A.40)

we may

and

(A.8)

summarize

the

suggest
cubic

that
model

A4 = As = C5.
and

its coefficients

as follows:

(A.43)
where

#r=

Eq.(3)

(A.44)
1

C_, =

kU*

(A.45)

4.0 + As_

A_=

+/

1-2C_(k"
_
ak_ , .
0.s + _za s
1.6pT

As =

In Eq. (A.43),
=

(A.47)

pk'7(s*) _+ (n*)_
z3

S_2*)n
k
a_kj

(AA6/

we have used

_ikS2_.

S2kf/kj.

the fact
In addition,

that

*
S_kl'lk j

c.

must

2O

*
12ikS_ i = Siki-lki
also

be constrained

- fli_Skj
by the

and

c_ o(2*)
_,.ikoki
-

conditions

from

Eqs. (A.21) and (A.28), i.e.,

C_, <_

The
(A.1).

cubic

As

model

The

can

corresponding

+ -_IIsA5

and

be directly

expressed

coefficients

C_ < _

(A.48)

in terms

of mean

C2=-_

A3 +

(A.51)
(A.52)
(A.53)

C4=0
Cs = As
the

cubic

Eq.

(A._0)

IsA5

C3 = _

then

i.e.,

(A.49)

IsAs

2_

gradients,

are

lk
C1 -

velocity

model,

Eq.(A.43),

becomes

2U

_o/
/

+ _- _--7[
2II
AsP k4
+
7
[u_,,u_,pup_ + u_,juk,pup,_ - -_ 3 _j

(A.54)

21

Appendix:

Equations

In this appendix,
general
form.

a set of mean

coordinate

a curvilineax

system.

coordinate

In Appendix

example

in a General

flow equations

This

appendix

system.

how

to write

with
with

these

the

equations

the equations

and

System

a general

will be found

We start

C, we will write

to show

Coordinate

cubic

useful

model

for studying

governing

equations

in a cylindrical
models

will be written
turbulent
in general

coordinate

for a specific

curvilineax

in a

flows in
tensorial

system

as an

coordinate

system.

B.1

Equations

in tensorial

p,t + t

form

(B.:)

_
0
,,_pUJ)
,J

(ok)..+ (ov'k)..=

. +_

k_ + e, - 0_

(i061, t -1- (pui_),

]A --[- -_

6 ]

,, ..[- Celfl_P

(B.2)
(S.3)

k -

Ce2f2P--

(B.4)

-_Ce3 _]_T gj, S j S,r

P
where

Pk= # (-_,-_)<'_,
The

turbulent

stress

where

the

subscript

of a coordinate
cubic

model,

zii =

is written

2 k

s=

in the

following

(v.. +

",i" denotes

system,

(B.5)

which

a tensorial
axe defined

pk3
17

(Vi,kV_

derivative,
in Eq.(B.16).

TO, is

form:

2
gij)
+ vj,k<,k - 5II,

+ c Ok_('U'''U''_-W
- -g:n_ ,,,)
+ c,--fi-Pk"
(vk,,V_- -g:
:I, ,,,)
22

(B.6)

g'J and
The

gii are

nonlinear

the two metric


part

of the

tensors
general

gij)
(B.7)

where,
IIt = U,_i
i k ,
In addition,
Eq.(A.20)

the

II2 = g klUi,kU,i,z ,

often

used

can be written

(s.) =

parameters

S* and

(B.8)

i2" defined

in Eq.(A.42)

and

W*

in

as

1 (gi.iU,,iUt}

(a')

scalar

II3 = g-ktUi,kUi,,,,U_
,z

i2

(B.9)
(B.10)

_ U_U{)

(B.11)

The

nonlinear

part

strain

and

rotation

B.2

Equations

of turbulent
rates

Sij and

in a general

Let x i represent

stress

_'ij, Eq. (B.7),

f_ij which

coodinate

can also be expressed

will be listed

in terms

of mean

in Eq. (B.33).

system

a general

curvilinear
coordinate
system,
then the corresponding
contravaridx i
ant velocity
is defined
as U i = m
and the covariant
velocity
is defined
as Ui = glj U j. To
dt
write Eqs. (B.1)-(B.10)
in this general coordinate
system,
we need the following
expressions
for various
tensorial
derivatives:

OAi
Ai,j = OxJ
A_-

A_,k-

OAi
OAij
0x k

_ OA_
Ai, k - _xk
where

P_jAq

+ r',jA
q
FikAq1 -

q i
_
q
FikAq + FqkAi

F_.k is a Cristoffel

(B.1)-(B.10)

Ps +

can

q
FikAi q

be written

symbol

(B.12)

defined

in Eq.(B.17).

With

the

above

formulations,

Eqs.

as follows

OPUJ
J
Ox---T + Fj_ pU _ = 0

(B.13)

23

r_pu=u_ + r_=pu,u = -

(pU,),,+ -- OxJ

-g'rf'('+_)\oxp+
+ gJ"

Ox"

oP

oxJ erjpu.-_ogj_
(B.14)

Fi'%i - Fi"wi_

where
OU k

o=<_=

yF + rLv-

(B.15)

and
Ox _ Oxi
g'J- OXk OXk

here

X k denotes

system.

The

0,"

The equations
can be written

(pk),t

the

symbol

OXJ

OX _'OX I'
,

gli -

Cartesian
F}i,, called

Ox i Oxi

coordinte
the

(B.16)

system

Christoffel

while

symbol,

x _ represents
is defined

a general

\ ax k )

for the turbulent


as

(B.17)

kinetic

Ox-----7- + ri_pkU =

energy,

Ox-; _ +

giJ_,,U

i ---UiU i,

and

- g_rj,

its dissipation

_+

rate

ox,,

+ Pk - Pe

(pc),,

coordinate

as

(B.18)

OpUic
+ -c

24

##T2,0S

OS

(B.19)

where
(B.20)

Pr = -gr;pu--_ kOzr + r_u"

2
-puiu

i = --_pkgij

If we decompose
to the

Pr into

nonlinear

Pr-

part,

p_l)+

OU_
Oz_ + --

+ Pc tOxJ

two

then

parts,

one

,,
2uU,_

due

to the

2
-_Og o

linear

+ T#

part

(B.21)

of -puiuj

and

the

other

due

we may write

pk(2)

(B.22)

where

= --_(pk

The

nonlinear

(ou+

+ I_TO)O

\ o.k + rkw m .

of the

cubic

model,

Eq.

(B.7),

in a general

coordinate

system

c pk3(v' our ov, _


our p kv#q
_'J= 17 ko_r o.J + -y_rqjvq- r,_vq0%7- r,kr_j
OUj
OU r
0X r OX i

"{"

OUj
r
_-_-xkrqiuq

\b-_ o_'
C

pk3

(OUr

q
- riruq

_r_,u.

OU r

OUkFr

OUk
_

U q -

ppir-qi,_p,_
r,k rr rrq

,r "b-_

F q U, oUr

2II
"_

--

/
/

r,_u,u.u
Fp _k

lgii

rr rrq

- gn_gij
lII

)
/

\o_r o..,

o_z

+rfru_v.u.OU'-

OUi

Oxl

,n

-_xkrjrnUqrrlV

OUt OUi
-_

"q- OX r

OUIOU"

-_kU_,_z_

OZ k _m

ox_ Ox_r_.,u_ - irU_,_x_

ou_ov, ou"

OX I

(B.23)
(B.24)

\ oxk
part

oxi

+ 9kJlgr _OxJ

OX rn

ouj ov _ .

OT, 1

OZ

vpvq rrrrOU"
+,.jr-i_,_,_q
,.

OUi'_

,,., ,.

OUi vq

+ Ozr _z_F,._

rr r,,_rr,

2II
+r_rr,kr,"%u,u" ) - -_
_,_]

25

OUi

"_

+ ,rr_._r.,u.u_u"

rrr,,,_rr-

Oxk_im_q*rl

OT, 1 rirnUq

OU_OUj _ ,

Oz_

OXm_.ik_,q.trlV

_rlU

--

Ox---_F_kUq_rlU

,n

,.

Oz t

is

c s--_pk, (ov_
ov_av'
\ Ozi Oz z ozJ

ov_
, av' r_v_OV_
au'
Ozk r_"uP-6_-iz_+
oz z ozJ

vk vp rr rrq OUl
--J" qi_" kl_P

'J

OU k OUk OU z

"Jff

k
OUkOU l
+FqiUq Ozz Ozk
OUk p I

OX j

OU t

rklVp

OUkpk U UqU _
+ _

OUk OUkF_ U,

+ Ox-/ _

OU _ p

vkvprrrrq OUz OU_OUkrl U"


-qj_kz'-'p'-" _
+ Ozk _
,.j
,

Ox i FmF,iUpU

OX 10X

qi ,i

OUk p t

"

k p t

q ,

- FqiFmF,iUpU

Oxi PmP'iUpU

OUkpk U UqU"

+ _

q-/ ,i

217

(B.25)

where

(or,

II_ = \ Oxk + r'pkUv) \ Ozi +

(B.26)

ii2 = 9k, (koz_+


OUk r;_u.
.
) (koz.
OU, _ Fit

(B.27)

. ) \ Oz"

\ Ozj' + r'pkuP
The

scalar

strain

and

rotation

rates

) ik Oz t

__ piqUq

+ r.,u

(B.28)

are

(or' .
] +ko.i +r;y')

l [9ii(\ OUk
(s')_ = -_
O=k rLU,) (\ U_+

(ov,
,
"]-rlkVl)]
O_ ,i

(B.29)

--_ \ Oz_ + r_u'


l [gii ( OUk

k O=j

B.3

W*=

Eq.(B.39)

Another

form

In terms

--puiu

of strain

( aUk

r;iuq]

+ r_v')
(B.30)

OT'i

(B.31)

of the

and

2
i -- ---_p

r_iUq)

cubic

rotation

model

rates,

k gq "b cPk22(
_T

the

cubic

Reynolds

1
Ski -- -_E)
gii ) + "rk./

26

stress

model

can

be written

as

(B.32)

where the

nonlinear

"rij: 2Al_2

part,

Tij, is

(g_SipSqj

pk3 g pq(Si,_qj
+ A3--_

_S(2)glj)

+ 2A2_2

_ _ipSq,)

(gPq_ipflqj

_F 2A4p-'_gN
k4

l _(2)glj)

g "" (SipSqr_-_sj

__ _.-_ipSqrSsj)

(B.33)
where

1 (OUi

OUj

2F_jUk)

0 = .q'WS_ , S (2) = g_g'S;,,.S,q


_S_2

= g_'g

,t

_ij

=1

(OUi

(B.34)

OUj)

(B.35)

, FL(2) = g_g"'_p,.f_,q

,_a , Is=O,

IIs

= _

02-S

(B.36)

(2)

and

(S*) 2 = gkVgzqSkzS_ -- _e 2

(B.37)

(_.)2

(B.38)

W*

= g_, glq_kZ_ _
= g_i gkl g,,,,, S_*_S_* S_i

(B.39)

(s'p

where,
1

Note

that

convenience

Eq.

(B.33)

for the

appears
CFD

to be more

compact

implementation.

27

than

Eq.

(B.25)

and

may

bring

some

Appendix:

Equations

C.1

Mean

Now,

let us write

this,

we need

cylindrical
two

in Cylindrical

Coordinates

equations

all equations

to

calculate

coordinates.

systems

Let

in a cylindrical
the

metric

coordinate

tensors

gii,

X i = (x, y, z) be the

system:

g#

and

cartesian

x i = (x, r, 0). To accomplish

the

Christoffel

system.

The

symbol

relation

F_k for

between

the

is

x = x ,

y = r cosO ,

z = r sinO

(C.1)

x--x,

r--

O=azctan(z/y)

(C.2)

or

We may

easily

Ozl

calculate

[1 0 0} xi10 o}
0

x--V=
The metric

_/y2 + z 2 ,

cosO

tensors

gij and

[lo0]

g'J=

and

the

sinO

o -sinOlr

1/r 2

Christoffel

symbol

eosOlr

'

90 can then

be obtained

gij=

0
i
0

cosO

-r

0 sinO

1
0
0

sinO

according

to Eq. (B.16):

0
0]
r2

F_k can be obtained

(C.4)

from

Eq.(B.17)

[000}[00o} 1000)

r_k= 0 0 0 , rh=
0

The

contravariant

velocity

0 0
0

in the

cyhndrical

, r_,=

0 0 1/_
0

coordinates

1/r

U and

covariant

is

(c.6)

V are the axial

velocity

(c.5)

v _= (v, v, _)
where

(C.3)

_cosO

and radial velocities,

can be obtained

_2 is the angular

velocity.

The

corresponding

from

Ui= g_jUj = (U,V, r2S_)

(C.7)

28

With Eqs.(C.1)-(C.7),
become
Continuity

the

equations

for turbulent

flows

in a cylindrical

coordinate

system

equation

op opu opv op_ pv


_ +--_- +--_- +-_- +--T = 0
Momentum

(c.s)

equations

OpU

OpU 2

Ot

cOx

OpUV

OpUfl

cOT

pUV

O-P

cOX

cOO

av)]

+_-_
+_-z_o0(.+.r)
cO_-_

+_

-_+

cOT_,

+_

0x ]J+; ("+"r)

i cOTzS

N +_ov)

(c.9)

+ Tz cO--_+ -n,T

(c._0)
Opr2f_U

Opv2fl
cOt

Jr

copr2fW

cOx

+ _o

2(. +.r)

+T("+"_)\
OTO_

+_

cOr

Opr_22
+

cO----O--+ rpV_

COTS,

1 0_-_

cOO

T_o + ;_ [(. +.r)v]

cOo

cOT+_

+_

O-P

TOT
1

(C.11)

+ T-_cO-_+ -Tno

where
cOU

cOV

cO_

(c._2)

e= cO-_+-b-T+ N +-T
29

k-s

equations

in Cylindrical

(c.t3)

coordinates

(C.14)

(c.ts)
where

(c._o)

p_ = p(kt)+ p(k_}

(c.tT)

(C.18)

3O

C.2

Nonlinear

After

gii,

(B.25)

part

gij and

l_i_ for the

or Eq. (B.33)

program.
are zero,

of turbulent

stresses

cylindrical

to calculate

However,
therefore

Tij

coordinate

system

all the turbulent

stresses

are calculated,
automatically

in the cylindrical
coordinate
system,
most components
it is possiple
to manually
write down all the turbulent

many unnecessary
form for the cubic

null operations
model, so that

their applications.
Ai, or vice versa.

Note

that

we may
through

in the computer
code. We write them here in a general
model users can use their particular
model coefficients
for

with

Eq.

(A.8),

the

coefficients

Ci can be easily

27

_'z1"

C4 pk4

Cs pk*

C pka

7
7

17

Tf_ + TIT +...

+ T_" + T_V -

iV:

+ Ts*_

+T_

Ox Ox

C pka

_-

+ ...

ou ou ou oy
_+_+_+_7

OU OV
+

OV OV
+

ov off

1
Oz ]

OU
-O--'-r +
Or OV

OU

T_* o"

r l 2 OUOV
O0 O0

C3 7
pk3

[OUOU
_Or

-[-

C4 _-3

[T;;"

+...

+ T_" + T_']

C s_

[T;_" + T[7" +...

+ T_*0" + T_']

C pk3

[OU OU

7"=O

O0
+rVo_
+

C pk_
2-7

ov

ov

Ox
L-g_--+

g2ii 3]

+ T:? - _n3]

Or Ox + O00z
----

[OU OV

Eq.

of gij gli and Fi_


stresses
to avoid

C. pk3

use

a computer

OV

OUOV
Or oo

+ r 20fl
_
( OU

O_

OU_]
O0
r20_

0_

+ OUOft
o-_oW+ OUV
oo r

+_+
O_)

rt2(OU-57+-F;
OV)

Or=_
O,_]aV

OU O_
[ OU O_
OU Or2_
Lr2_--'
Ox Ox +----Oo O0 + Or Or'-

31

rft OU

V OU ]

-_-r+r-O-OJ

obtained

from

OU OU

+
+

,r_

OV OV

-_ _

Cs pk4

+ 0--7o-_+

r2 _Oft Oft + rV Oft

o_ oe

r_t OV]

_-

-_j

[r_:+rg +... +r;: +r;:]


[V:+r;_+... +r_: +r_:]

--.

+
_2

3ii2

C4-_3 4 "T_'19+ T_2_ +...

+T_3_ +T_'_

_IIa ]

C 1-W
pka
OV
OU
a---_
o--_+ OV
0--7OV
a--_+ OV
ae O_
ae

2m

+ r2 Oft OU + Or2ft OV + r= 0f_ Oft


Ox Or
Or Or
08 Or
----+

rV 0_2 ]

pk4
_7

_3 + T_ +... + _

+ T_';- _2rl

"

]
V OV

T_O

T_

_j

_-_b_x+ or & + oo oo
Ca pka

[OUOU
or o_ + OVOV
o---_o--_+ &2_O_
Or O_

[_+_+...+_+_]

[_g+_+... +rg+_]

_-r + _

2\

Ox 08

2r_

_vO_
Or _ m (Ov
_r

+ -_

Or

08

08-_

+ 2r208

+ 2_y_-_+ 2v=+ 2_n =- S_ n,

C pka

2-_-

+2rV-_-_-

\ _]
-

r2_2

+\

o_ ] +

V2

\_]

-2m o--7-

_r2yI2]
2

+ 2rV O_

+ -_ 00 + av

C1 pka

+ --r 08

Pk a

--"

_r

=7 [koe]

+ koe]

-_

32

OO

+-_o_) + _y

--2r_ _Oe
V + V2
pk 4

r2_2

- _r
1 2ri2]

(c._9)

The

terms

T_...

The

terms

in r

T_i_ in the

above

equations

are listed

below:

ov ( o__

Tf: =_ _

ovov ovon_ ov ovov

a_ + 0-70-7+ _]

+ _ (_

OU OV

OU Oft ]

ev {_ o_ e_ov ov e_,_
T_g = O
T_'=O
T_" = O
T_" = Tf:

T27=0
2&'=O
T_%"
=0

T&:= O
T_'=O
T;g = O
T3%"
=0
T_: = O
TA" = O

r:: = o_ _,-gg=
(ov_o_ + _+
ovov _oe_]+__o
o__ ov ( ov_ + ovov ov o_
T:f=2rnO_
T_" = O
T&"= O

VOQ

(Oft)"

Tgf= T:_
T&'=O

33

T_3"= 0
TL'=0
T:s"- 0
T::=0
T:-0
T::-0
T:_=0
T_o"-- 0
T_'_= 0
_The terms

in %,:

OV OV
OVO_]
T_;--_-_ OU(OVOU
-_ Ox + OVOV
O_.._.O_.__
+____x/

+_.rOU(OVOU
-_ Or

Or

Or

+ r-ZO-'g
1 ov (or
-_z _ov + ov ov + ov on]

+ ov
+ r 2 ov
( ov on
Oz
Or on
Or
O0 on)
T_(= O
T_2"= 0
OU OV
Or

Or

OV 0_)

34

36

:F_.."
OV fort _ _,.
19 _ -'--UV

r_

r;t, -

-_- .._

_,-,/'ovo_

. ----mr

I _--__..

_" =w _

T"

- V/OV\

r 2

_,=__ov

--?" oo
OVOD

Or

oo ooJ
/

o, oo oooo)

OqO _l

O_
+ _. ( O_UO_U

uv OV

---..._

o,

Or 0::
OV ov

_-Or 0o:)

_to,. o,.+_o"7 +,_,.---_--)


" "_'

--.

_9 =_VY?.L _ _OV
T."
...__ Or _'r

Or OeJ

_ov
Or OOJ

37

_7: 00

T47= .___v

38

r 0800
Or
_V oV

T_3_ =

O-O
V z OV
-20V

T_ = _.n__ _V 2
r

Tg= g _T; _ + 0-7o-7+

_ _ ] + -bg k-g; _

o_ {or ov ov ov +o_o__
__ _ + g _ )
T_ = _ (r_ O_ av
2"_

ar2_ ov

+_

o_o_ _ o_/

0-7oW
r_ a_ o_ ]

- V OFt _ t2 0V

=_

-dg

O_'l (OU OU

OV OV

r20_'l O_ _

O0 Or

Or ]

_= r:

: [+ o_ov o+_ov +oao_)


_ _Y;+-_

= __V aV
Or + r_O_ _+

Or

Yg_]

V _.0_
Or

= v (o_ o_ ovov +o_o_

_:= T::
" k_-+
-_

(o_ ov ovov o_

--7--g+ _-)-"nkY;;_-_
Or

_1 = n ( o_ov ov ov +onon
Tg= T;g
_= _

39

+_/-_

+_

o,. )

40

_, _ v {at; ov av ov ,an an
- 7k-b-g_-g+ a-_o-_+ _g}
-,-n k_
TI:= _

+ o--g
o-7+gg o_ }

\ o_ +_
T_4Te
_

_+
Or

O0

V_
_
+ 2r_2V
r

TZ = T_?
Other

scalar

II1=

uantities:

2 (

r_ O_t

[OU

_2

_ OV

V O_ _

IOV'2

V2
(C.20)

+1

(c.22)
_}

+_ko_

+_

t_;

_+_}
+_

+_\-y;_
+ _1

OV OV

OV Ogt]

+ 0-70-7+ _}

+rv _

(c.23)

_,_,= o_{o_o_
o-7k-o;-YF+ ovov
o-7o-7+ ovo__
_g_} + _ov (ovo_
_-_-_; + ovo.
_
+ ovo_
o_ ov
{ ou ov o_ o_ yg _ }
+_ k_
+_
+ _g_} +ovov _oaon)
{_)
_
{onov
_
t_
+
o-_ + -+_vt_ } +2vo_+_n
onov orion]
o_Yg
+n_ + vn _
t_
+ o-70--7+ _g_}
T

n_,,_ ov (ovov
=g t_-+

ovov ovoo] ov {ov ov ovov _ on) <c.2_)


-7--g+_}+_
41

_;_+_-_

+ oo_}

(e') == _

W*=

C.3

Another

In terms

+" _

+ \_7]

\_]

Eq.(C.41)

form

of Sq and

of 7q

fq,

the

components

of _'ii can be written

as

(C.29)

42

pk 3 1 _
_-._.=2.,41-_23

ISnS12-{-S12S22+_-2S13S32/-I-2A2-_'_-_

13

[_12(SII- S22)+-_1 (S13_32 - _lzSn)

+A3_
+2A4_-3

[Sn (Snf212 +-_

+l

s13 (S3_12

S13f232) + Sn_

+ _S3_32)

32

S23f_32

- _12 (S22S22 + _S_3S32)

+
(c.3o)

Tz 0 --

_.cL17

$11S13

+A3-_2 3

Sl1_'_13-

+2A

S12S23

--_

_'_12S23

Jc

S13S33

7"2

23 (S12-

$32)]

s_ (s_a_ + s_2a_) + Sl,(S21a_ + s_,a,_)

_-_S13S32_23-_12($21S13Jf

-- 2A5

--_ ..%_127_'_12_'_23

($32S23

_34

- S22S23_S23S33)

.Jc

[_'_ 12 (,21__13

-_- _22_'_23)

-_- _2

_13

('31_'_13

-_- $32_'_23)

(C.31)

_-IIsS13-1sISllS13-]-$12S23-]--_S13S33)]

- o_

T1.v -- _.c'1 7pk3

($21S12 +

@-2A4_-3

[2S21

_34

-l_'_S_-_-}-,,s

[_"_12

-_- _823S32-_S

(2))

(S11_12-{--_S13_32>

+_ 2 S 23 (S31fhe
-- 2A5

$22 $22

q- 2S22

(S21_12

_- _

S23_32

>

+ _S33f_n)]

($11_"_12

($22-

--_- 1S13

1(_)>

_'_32)

-Is

-_- _2

_'_23

(S21S12-_-$22S22

43

($31_'_12

--_- _

$33_'_32)

nt- _$23S32

_S(2))]

(c.32)

pk 3 _

T,O=2A 17pk3 ($21S13+

$22S23 +1 r._$23S33)+

+APk3(s,l_133
7

$22['_23

__21S13

z_27_21f113
_'_23S33)

1.4

--'_--2.A4-_3

= [S21

($11_'_13

-_- _'_ $23S31__13

--_fl23
-

S12_"_23

_'_21

) -{.- $22($21_'_,13-.}-

(,11,,,+
S12S23

2Al_2

$22_"_23)

T_-_$13S33

(S31S13 + -_ S3_$33) ]

J1'-././S,,.,_23

Tso"-

--

(S31S13

(_'21

,.._13 -J- ,,._22 $23

-Jc- S32S23

Jr-

--_.- _

-._S33,.,_33
1

_'23 _33)

(C.33)

2S(2))

E*

2A

pk3

IA-

-.{.-2A4.._3=

[2S31

(Sl1_'_13

-._- $12_'_23

) +

2S32

($21__13

$22_'_23

1
7"2

2A5--_3

S33 (S13_'_31

[_-`_3_ ( _

-'{-././S

The

scalars

S (2)

that

$11

--

"_

SllSll

($33-

appear

$22

$23fl32)I

-_-

_3

-_- c_2_.-_23 )

1'F20)--1,.,

in the above

-_- _._32 ( S2_1_13

-.}- S22_.`_23 )

($31S13--[-S32S23

equations

-1T2-_---_

-.{- _$33S33

are

(C.34)

_-1-2S(2))]

as follows

(c.35)

"J" _2_33

i
-Jc ,..,,c22S22 -Jr- _-_'S'33S33

-Jr--2,-,c12,-,c12

-+-

(S'13S13

$23S23

(c.36)

(c.3z)
_"_S_'_

-"

['_12_'_21(Sll

"Jr S22

) +

__12a31_--_(S23

-.}- $32

) +

_"_21_'_32

(Sis

-Jr S31

Is=O,

Iis=1(02-S

(c.3s)
(c.39)

(2))

44

Two

other

scalars

(S*) 2 and

(s.)_ = s(_)_ 30_,

(fl,)2

can be expressed

(c.40)

(a.)_= ac_

' [.(....

,_shsh )

(s,) 3 s,, s,_sl_ + s,,s,1 +


+S_2

as

* *
* * + _shs;,)
(s_,s,2+s_@,,

,.(
+S_1

_
*

"

+S_2 (S2,S,2
* * + S22S22+-_S_,S;2)
* *

,_(..
--,_
)
,.(_)shsh + s;_s;, + s;,sh
+T_s_,
'.(_)S;lSh+ s;_sh+ s;_sh
+;zs,_
1.(
,_ )]
+

$31S12 + S_2S_2 +

S_3S_2

(c.4_)

where

Sn. = Sn _ 1-0
3 '

$22=$22-3

S_2

S13 ,

Finally,
_0

--

the

= -flJi)

S12,

S_ 3 =

six components

"

S;3

10 '
--

of Sij and

$33=$33-

1 2O,
_r

(c.42)

S23

the

three

components

of flij

(note

that

S 0 = Sji and

are

(c.43)

45

Appendix:

Continuity

Equations

for

Axisymmetric

Flows

equation

Op

cOpU

OpV

pV

N+-y_-_ +-_
Momentum

(D.I)

+--_ -

equations

OpU 2
OpU
--+--+--+
Ot
Ox

cOpUV

pUV

Or

O-P

Ox

(auav)
OpUV
0-'"7

OpV
--+Ot

+_

OpV _
Or

o[

(.+.T)

Or, z
+ _

Or,_
+

Or

(D.2)

o,=o,..1
pV 2
rp_2

+-r

O-P
=---Or

-8-_--=+_--+_

)]

1
(D.3)

r3 r_ + -Z,,r

Opr2_
&

+ Opr212U
o= +

Opr212V
o_

+_0 [ (,+_r)

i)r2_]
Or

+ _pv_ = _0

[ (_ +

.0r2_2]
_)--_-]

Or2_
o,.

j+7 1 (_+_)

1
2o
_)r'_]
+ _O're= + _0_'_ +-_.,
rOr[0,+
,"

(D.4)

where
OU

OV

V
(D.5)

e= o-_-+-av +-r

(D.6)

k-E equations

46

(D.7)
(_k

0---2-+

7"

Or

PP
(D.8)

[\g) + \_1 1
where
(D.9)
pk = p_l ) + p(2)

p(k_) = __(pk + _rO)O

ov

oa '_

_(ov' ov (or
+ 2v71
o,,)
(ov'f
_
(o_'f
oi_(u
+
OU

OV

Off

OU

(n.lO)

OV

(D._i)
+_'o, N+

-r.. in axis

mmetn'c

+-#

flows

,, _ [ o_(orov ovov_ _ov (orov ouov'__


,. _, hov (orov ovov_

ov (orov ovov'_
(D.12)

47

(D.13)

48

Or

Or2_

OV

OV Or2_

Or J

2 _2 L Oz Oz + "_r --_--r - r. _7 + .V]


e ['
"-'4

Or-

Or +_vl

o _.---_ 12_2Z2_ at/


x

_,azaz
Or

Ox Or

va;

Or_m _,,,

+_'----

0,- Or

-r_

- rf2

avav

---_._

-..-...

av

--.

_U

01.2

__

OV

arJ

Or ar ) - r_ k az ar +

Or

+ 2V2 + 2r2f_2 -

49

3 r I'I_]

Or

(D.16)

where
V_

_,----

-2n_+ -_

o_
O0 Or_

'
+ _

(eh'

+ 2e_

5O

(D.24)

1 [(_V_ 2

r2/Of_'_2

(OU_

(D.25)

51

Form Approved

REPORT

DOCUMENTATION

PAGE

OMBNo.0704-0188

Public repo(t_g burden for this collection of informationis estimated to average 1 hour per response, including the time for reviewinginstructions,searching existing data sources,
gathenng and maintainingthe data needed, and completing and reviewingthe collectio_of information. Send commenlsregardingthis burden estimateor any other _
of this
collection of information, includingsuggestionsfor reducingthis burden, to WashingtonHeadquartersServices. Directoratefor InformationOperationsand Reports, 1215 Jefferson
Davis Highway. Suite 1204, Arlington. VA 22202-4302, and to the Office of Management and Budget, PaperworkReductionProject (0704-0188), Washington,DC 20503.
1. AGENCY

USE ONLY (Leave

bianlO

2. REPORT

3. REPORT

DATE

TYPE

4. TITLE

AND

AND

SUBTITLE

Modeling

DATES

Technical

August 1997

COVERED

Memorandum

5. FUNDING

of Turbulent

Swirling

NUMBERS

Flows
WU-522-31-23-00

S. AUTHOR(S)
Tsan-Hsing Shih, Jiang Zhu, William
and John L. Lumley
7. PERFORMING

ORGANIZATION

NAME(S)

Liou, Kuo-Huey

Chen, Nan-Suey

Liu,

AND ADDRESS(ES)

8. PERFORMING
ORGANIZATION
REPORT NUMBER

National Aeronautics
and Space Administration
Lewis Research Center
Cleveland, Ohio 44135-3191

E-I0868

9. SPONSORING/MONITORING
AGENCYNAME(S)ANDADDRESS(ES)

10.

National Aeronautics
and Space Administration
Washington, DC 20546-0001

SPONSORING/MONITORING
AGENCY REPORT NUMBER

NASATM-113112
ICOMP-97-08;

11. SUPPLEMENTARY

CMOTT-97--03

NOTES

Tsan-Hsing Shih, Jiang Zhu, and William Liou, Institute for Computational Mechanics in Propulsion and Center for
Modeling of Turbulence and Transition, Cleveland, Ohio (work performed under NASA Cooperative Agreement
NCC3-520);
Kuo-Huey Chert, University of Toledo, Toldeo, Ohio 43606; Nan-Suey Liu, NASA Lewis Research Center;
John L. Lumley, Cornell University, Ithaca, New York 14853. ICOMP Program Director, Louis A. Povinelli, organization
code 5800, (216) 433-5818.
12a.

DISTRIBUTION/AVAILABILITY

Unclassified
Subject

STATEMENT

121).

DISTRIBUTION

CODE

- Unlimited

Category

34

This publication is available from the NASA Center forAeroSpace Information, (301) 621--0390.
13. ABSTRACT

(Maximum

200 words)

Aircraft engine combustors generally involve turbulent swirling flows in order to enhance fuel-air mixing and flame
stabilization.
It has long been recognized that eddy viscosity turbulence models are unable to appropriately
model
swirling flows. Therefore,
it has been suggested that, for the modeling of these flows, a second order closure scheme
should be considered
because of its ability in the modeling of rotational and curvature effects. However, this scheme will
require solution of many complicated
second moment transport equations (six Reynolds stresses plus other scalar fluxes
and variances), which is a difficult task for any CFD implementations.
Also, this scheme will require a large amount of
computer resources for a general combustor swirling flow. This report is devoted to the development of a cubic Reynolds
stress-strain model for turbulent swirling flows, and was inspired by the work of Launder's group at UMIST. Using this
type of model, one only needs to solve two turbulence equations, one for the turbulent kinetic energy k and the other for
the dissipation rate & The cubic model developed in this report is based on a general Reynolds stress-strain relationship.
Two flows have been chosen for model evaluation.
One is a fully developed rotating pipe flow, and the other is a more
complex flow with swirl and recirculation.

14.

SUBJECT

TERMS

15. NUMBER

OF PAGES

54
Turbulent

model; Swirling

flows; Combustors

16. PRICE

CODE

A04
17.

SECURITY
CLASSIRCATION
OF REPORT

Unclassified
NSN

7540-01-280-5500

18.

SECURITY
CLASSIFICATION
OF THIS PAGE

Unclassified

19. S ECU RITY CLASSIFICATION


OF ABSTRACT

20.

UMITATION

OF ABSTRACT

Unclassified
Standard
Form 298 (Rev. 2-89)
Prescribed by ANSI Std. Z39-18
298-102

You might also like