Turbulence

T
b l
Modeling
M d li in
i
Computational
p
Fluid Dynamics
y
(CFD)
(
)
Jun Shao, Shanti Bhushan, Tao Xing
and Fred Stern

Outline
1.
2.
3.
4.
5.
6.
7
7.
8.

Characteristics of turbulence
Approaches to predicting turbulent flows
Reynolds averaging
RANS equations and unknowns
The Reynolds-Stress Equations
The closure problem of turbulence
Ch
Characteristics
t i ti off wall-bound
ll b
d turbulent
t b l t flows
fl
Turbulence models and ranges of applicability
7.1.
7
1 RANS
7.2. LES/DES
7.3. DNS

9 Example:
9.
E
l diffuser
diff
2

Characteristics of turbulence
Randomness and fluctuation: u = U + u

Nonlinearity: Reynolds stresses from the

nonlinear convective terms

Diffusion: enhanced diffusion of momentum,
energy etc.
Vorticity/eddies/energy cascade: vortex
stretching
3

Characteristics of turbulence

Dissipation: occurs at smallest scales

Three
Three-dimensional:
dimensional: fluctuations are always 3D
Coherent structures: responsible for a large part of
the mixing
A broad range of length and time scales: making
DNS very difficult

Approaches to predicting turbulent flows

AFD, EFD and CFD:

AFD: No analytical solutions exist

EFD:
EFD E
Expensive,
i
time-consuming,
ti
i
and
d sometimes
ti

impossible (e.g. fluctuating pressure within a flow)

CFD: Promising,
g, the need for turbulence modeling
g

Another classification scheme for the approaches

The use of correlations: CD = f ( Re )

Integral equations: reduce PDE to ODE for simple cases
One-point closure: RANS equations + turbulent models
Two-point
Two
point closure: rarely used,
used FFT of Two-point
Two point
equations
LES: solve for large eddies while model small eddies
DNS: solve NS equations directly without any model
5

Deeper insights on RANS/URANS/LES

RANS

URANS

LES

Reynolds Averaging

Time averaging: for stationary

turbulence
Spatial averaging: for homogenous
t b l
turbulence
Ensemble averaging: for any turbulence
Phase averaging: for turbulence with
periodic motion
p

RANS equations and unknowns

RANS equation

U i
U i
P

+ U j
=
+
2 S ji uj ui
t
x j
xi x j

RANS equation
q
in conservative form
U i

2 S ji )
+
+
U jU i + uj ui =
(
t
x j
xi x j

Numbers of unknowns and equations

Unknowns: 10 = P (1) + U (3) + ( uiuj ) (6)

Equations: 4 = Continuity (1) + Momentum (3)

The Reynolds-Stress Equation

Derivation:
D i i Taking
T ki moments off the
h NS equation.
i

Multiply the NS equation by a fluctuating property and time

average
g the pproduct. Usingg this pprocedure,, one can derive a
differential equation for the Reynolds-stress tensor.

ijj
t

+Uk

ijj
xk

= ik

U i
ui uj
jk
+ 2
xk
xk
xk xk

U j

ui p u j p ij

+
+

+ uiu j uk
+

x j xi xk xk

NEW Equations: 6 = 6 equation for the Reynolds stress tensor

NEW Unknowns: 22 = 6 + 6 + 10
ui p u j p
+

6 unkowns
x j xi

ui uj
6 unkowns
xk xk

uiu j uk 10 unkowns

The closure problem of turbulence

Because of the non-linearity of the Navier-Stokes

equation as we take higher and higher moments,
equation,
moments
we generate additional unknowns at each level.
In essence, Reynolds averaging is a brutal
simplification that loses much of the information
contained in the Navier-Stokes equation.
The function of turbulence modeling is to devise
approximations
i
i
for
f the
h unknown
k
correlations
l i
i
in
terms of flow properties that are known so that a
sufficient number of equations exist.
In making such approximations, we close the
system.

10

Characteristics of Wall-Bound Turbulent Flows

The turbulent boundary layer (zeropressure gradient) has universal

velocity distribution near the wall
(inner-layer) (Clauser 1951)
u
yu
= u+ = = y+ : y+ < 5
u

1
u+ =
ln( y + ) + 5 . 1 : 10 4 > y + 30
0 .4
k
4
+
=
0
.
3
:
10
>
y
30
2
u

The velocity
Th
l it defect
d f t when
h plotted
l tt d
vs. y/ collapse on a single curve
(outer-layer)
u u
y
= 9.6(1 )2
u

11

Turbulent models (RANS)

(
)

Boussinesq eddy-viscosity approximation

Algebraic (zero
(zero-equation)
equation) models
Mixing length
Cebeci-Smith Model
Baldwin-Lomax
Baldwin Lomax Model

One-equation models

Baldwin-Barth model
Spalart-Allmaras model

Two-equation models
k-
k model
k- model

Four-equation (v2f) models

Reynolds-stress (seven-equation) models
12

Turbulent models (RANS)

(
)

Boussinesq eddy-viscosity approximation

U i U j
uiu j = T
+
x
xi
j

2
ij k
3

Dimensional analysis shows: T = C qL , where q is a

turbulence
b l
velocity
l
scale
l and
d L is a turbulence
b l
length
l
h
1
scale. Usually q = 2k where k = uiui is the turbulent
2
kinetic energy.
gy Models that do not provide
p
a length
g
scale are called incomplete.

13

Turbulent models ((0-eqn

q RANS))

Mixing length
Assume

2
model: T = lmix

lmix = ( x )

dU
dy

for free shear flow, then

=0.180 for far wake

=0.071 for mixing layer
=0.098 for plane jet
=0.080 for round jet

Comments:

Reliable only for free shear flows with different values

Not applicable to wall-bounded flows
14

Turbulent models ((0-eqn

q RANS))
Cebeci-Smith Model (Two-layer model)
Ti , y ym
T =
To y > ym

ym iis the
Where
h
h smallest
ll value
l off y for
f T =T
i

Inner layer:

U 2 V 2
2
Ti = lmix
+

Outer layer:

T = U e v* FKleb ( y; )

12

lmix

= y 1 e y

A0+

0 0168
Cl
Closure
coefficients:
ffi i t = 0.40
0 40 = 0.0168

y
FKleb ( y; ) = 1 + 5.5

dP dx
A+ = 26 1 + y
U2

Comments:
Th key
Three
k modifications
difi ti
tto mixing
i i length
l th model;
d l
Applicable to wall-bounded flows;
Applicable to 2D flows only;
Not reliable for separated flows;
, v* , U e difficult to determine in some cases;

1 2

v* = (1 U U e ) dy
0

15

Turbulent models ((0-eqn

q RANS))
Baldwin-Lomax Model (Two-layer model)
Ti , y ym
T =
To y > ym

ym iis the
Where
h
h smallest
ll value
l off y for
f T =T
i

= y 1 e y

Inner layer: T = l lmix

Outer layer: T = Ccp Fwake FKleb ( y; ymax CKleb )
Closure coefficients: = 0.40 = 0.0168
2
mix

A0+

FKleb ( y; ymax

Comments:

y
CKleb ) = 1 + 5.5

ymax CKleb

A+ = 26 Ccp = 1.6 CKleb = 0.3 Cwk = 1

12

V U 2 W V 2 U W 2
=

+
+

x
y
y
z
z
x

Applicable
A
li bl to
t 3D flows;
fl
Not reliable for separated flows;
*
No need to determine , v , U e
16

Turbulent models ((1-eqn

q RANS))
Baldwin-Barth model
= C RT D1D2

Kinematic eddyy viscosityy : T

Turbulence Reynolds number :

RT
t

) + U ( R ) = ( C
x
T

2 f 2 C 1 )

RT P + ( + T )

2 RT
xk xk

Closure coefficients and auxiliary relations:

+

2 0 C = 0.09
0 09 A0+ = 26 A2 = 10
C 1 = 1.2
1 2 C 2 = 2.0

= ( C 1 C 2 )

U U j U 2 U U
y+
i
i
k
k
P = T
+

D1 = 1 e

xi x j 3 xk xk
j

C 1 C 1 1
1 D2 y +
+ 1
+

+
f2 =
D
D
D
D

+e
1 2
1 2
C 2 C 2 y +
D1 D2 A0

A0+

A0+

1 T RT
xk xk
C

= 0.41
0 41

D2 = 1 e
+
D
+ +1 e y
A2

A2+

y + A2+

17

Turbulent models ((1-eqn

q RANS))
Spalart-Allmaras model
Kinematic eddy viscosity : T =  f v1
Eddy viscosity equation:
2



 cb1  
 1

+U j
= cb1S cw1 f w +

(
)

+
t
x j
xk xk xk
d xk

Closure coefficients and auxiliary relations:

cb1 = 0.1355 cb 2 = 0.622 cv1 = 7.1 = 2 3

cw 2 = 2.0

cw 2 = 0.3
f v1 =

3
+c
3

3
v1


= 0.41 =

fv2 = 1
1 + f v1

cw1 =

cb1

(1 + cb 2 )

g = r + cw 2 ( r r )

1 + cw6 3
fw = g 6 6
g + cw3

16


S = S + 2 2 f v 2
d

r=


S 2 d 2

S = 2ij ij

18

Turbulent models ((1-eqn

q RANS))
Comments on one-equation models:
1.One-equation
q
models based on turbulence kinetic energy
gy are incomplete
p
as
they relate the turbulence length scales to some typical flow dimension. They
are rarely used.
2.One-equation models based on an equation for the eddy viscosity are
complete such as Baldwin-Barth model and Spalart-Allmaras model.
3.They circumvent the need to specify a dissipation length by expressing the
decay, or dissipation, of the eddy viscosity in terms of spatial gradients.
4 Spalart-Allmaras
S l All
model
d l can predicts
di better
b
results
l than
h Baldwin-Barth
B ld i B h
model, and much better results for separated flow than Baldwin-Barth model
and algebraic models.
5 Also most of DES simulations are based on the Spalart-Allmaras
Spalart Allmaras model.
model

19

Turbulent models ((2-eqn

q RANS))
T = C k 2

k- model:

U i
k
k

+U j
= ij
+
t
x j
x j
x j

k
( + T k )

U i
2
+U j
= C 1 ij
C 2
+
( + T )

t
x j
x j
k x j
k x j
,

0 09 k = 1.0
1 3 = ( C k ) l = C k 3 2
1 0 = 1.3
C 2 = 1.92
1 92 C = 0.09

C 1 = 1.44
1 44

k- model:

T = k

U i
k
k

+U j
= ij
* k +
t
x j
x j
x j
U i

+U j
= ij
2 +

=
=

13
25

= 0 f

ij jk Ski

( )
*
0

= * k

0* =

x j

* = 0* f
9
100

l = k1 2

f*

x j

1
2

k
*
( + T )

( + T )

x j

* =

1
2

k 0
1,

= 1 + 680 k2
1 + 400 2 k > 0
k

0 =

x j

9
125

k =

f =

1 + 70
1 + 80

1 k
3 x j x j
20

Turbulent models ((2-eqn

q RANS))
Comments on two-equation models:
1. Two-equation
1
Two equation models are complete;
2. k- and k- models are the most widely used two-

equation models and a lot of versions exist. For example, a

popular variant of k- model introduced by Menter has
been used in our research code CFDSHIP-IOWA. There are
also a lot of low-Reynolds-number
y
versions with different
damping functions.
3. k- model shows better results than k- model for
flows with adverse pressure gradient and separated flows
as well as better numerical stability.

21

Turbulent models ((4-eqn

q RANS))

v2f-k model: T

Dv 2
v2
= kf +
Dt
k x j

= C v 2T

P
C1 v 2 2
L f f = C2 k
T k 3
k

v2
( + t )

x j

t
D

t k
C
P
C
=

+
+

1
2
+

Dt
k
k
x
x

j
j

k x j

10
C1 = 0.4 C2 = 0.3 CL = 0.3 C = 70 C 2 = 1.9 = 1.0
3 14
k 3 2
4

2
k
L = max C
C
,
C 1 = 1.3 + 0.25 1 + ( CL d 2 L )

=
T
max
,
6

Dk

= P +
Dt
x j

v2f-k
v2f
k model:

T = C k v

Dv 2

= kf 6v 2 +
Dt
k x j

1 n

v2
( + t )

x j

U i
Dk

= ij
* k +
Dt
x j
x j

t k
+

k x j

v2 2
P
1
1 v2
L f = ( C1 1) 5
C2 k
k 3 T k
T
k

1 n

2
D
U i
2 v
= ij

k
x j
Dt
k

= 3 40 k = 1.0 = 1.5
14
k 3 2
3
, C
L = max CL

x j

t
+

x j

= 0.45977 * = 0.09
k

T = max , 6

22

Turbulent models ((4-eqn

q RANS))
Comments on four-equation models:

1. For two-equation models a major problem is that it is hard to

1
specify the proper conditions to be applied near walls.
2. Durbin suggested that the problem is that the Reynolds number is
low near a wall and that the impermeability
p
y condition ((zero normal
velocity) is far more important. That is the motivation for the equation
for the normal velocity fluctuation.
3. It was found that the model also required a damping function f,
hence the name v2f model.
4. They appear to give improved results at essentially the same cost
as the k- and k- models especially for separated flows. Hopefully the
v2f-k model can have better numerical stability than v2f-k model as
their counterparts behave in two-equation models.

23

Turbulent models ((7-eqn

q RANS))
Some of the most noteworthy types of applications for which models based
on the Boussinesq approximation fail are:
1. Flows with sudden changes in mean strain rate
2. Flow over curved surfaces
3. Flow in ducts with secondary motions
4 Flow
4.
Fl iin rotating
t ti fluids
fl id
5. Three-dimensional flows
6. Flows with boundary-layer separation
In Reynolds-stress models, the equations for the Reynolds stress
tensor are modeled and solved along with the -equation:
ij
t

+Uk

ij
xk

= ik

U i
ui uj ui p uj p ij

u
u
u
jk
+ 2
+
+
+
+

i j k
xk
xk
xk xk x j xi xk xk

U j

24

Turbulent models ((7-eqn

q RANS))
These are some versions of Reynolds stress models:
1. LRR rapid pressure-strain model
2. Lumley pressure-strain model
3. SSG pressure-strain model
4. Wilcox stress- model
Comments on Reynolds stress models:
1.Reynolds stress models require the solution of seven additional
PDEs and those equation are even harder to solve than the two-equation
models.
2. Although Reynolds stress models have greater potential to
represent turbulent flow more correctly, their success so far has been
moderate.
3. There is a lot of current research in this field, and new models
are often proposed. Which model is best for which kind of flow is not clear
due to the fact that in many attempts to answer this question numerical
errors were too large to allow clear conclusions to be reached.
25

Turbulent models (LES)

Large scale motions are generally much more energetic than

the small scale ones.
The size and strength of large scale motions make them to be
the most effective transporters of the conserved properties.
LES treats the large eddies more exactly than the small ones
may make sense
LES is 3D, time dependent and expensive but much less costly
than DNS.
DNS
26

Turbulent models (LES, filtering)

LES needs a velocity field that contains only the large scale
components of the total field, which is achieved by filtering
the velocity field (Leonard
(Leonard, 1974)

ui ( x ) = G ( x )ui ( )d

G(x-) is the filter kernel, is a localized function, which

includes a Gaussian, a box filter (a simple local average) and a
cutoff (a filter which eliminates all Fourier coefficients
belonging to wavenumbers above a cutoff)
Each filter has a length scale associated with it, .
Eddies of size large than are large eddies while those
smaller than are small eddies and need to be modeled.

12
x 2
if x

G (x ) =
2 G (x ) = 2 exp
2

0 otherwise

Box or top-hat filter

Gaussian filter

G (x ) =

sin (kc ( x ))
kc (x )

kc =

Cutoff filter
27

Turbulent models (LES,

g Equations)
q
)
Governing

Filtered Navier-Stokes equations (constant density,

incompressible):

(ui )
=0
xi

(ui ) (ui u j )
p

+
=
+
t
x j
xi x j

Note that:

ui u j ui u j

Introducing Subgrid-scale
Reynolds Stress

u u j

i +
x j xi

ijS = (ui u j ui u j )

The filter width > grid size h

The models used to approximate the SGS Reynolds
stress are called subgrid-scale (SGS) or subfilter-scale
models.
d l
28

Turbulent models (LES, Smagorinsky model)

The earliest and most commonly used subgrid scale

model is one proposed by Smagorinsky (1963), which is
an eddy viscosity model.
As the increased transport and dissipation are due to the
viscosity in laminar flow, it seems reasonable to assume
that

ui u j
1 S
= 2 t Sij
kk i , j = t
+

x
x

i
j

S ij Strain rate of the large scale or resolved field

2
2
t Eddy viscosity t = CS S
S
i, j

C S 0.2 Model constant

D
Drawbacks:
b k 1.
1 Cs
C is
i nott constant
t t
2. Changes of Cs are required in all shear flows
3. Need to be reduced near the wall.
4 Not accurate for complex and/or higher
4.
Reynolds number flows.
29

Turbulent models (LES, Scale-similarity model

and Dynamic
D namic model)
Dynamic model:
1. filtered LES solutions can be filtered again
g
using
g a filter
broader than the previous filter to obtain a very large scale
field.
2. An effective subgrid-scale field can be obtained by
subtraction of the two fields.
3. model parameter can then be computed.
Advantages: 1. model parameter computed at every spatial
grid point and every time step from the results of LES
2. Self-consistent
lf
subgrid-scale
b d
l model
d l
3. Automatically change the parameter near the wall and in
shear flows
Disadvantages: backscatter (eddy viscosity<0) may cause
instability.

30

Turbulent models (DES)

Massively separated flows at high Re usually involve

both large and small scale vortical structures and very
thin turbulent boundary layer near the wall
RANS approaches are efficient inside the boundary
layer but predict very excessive diffusion in the separated
regions
LES is accurate in the separated regions but is unaffordable
f resolving
for
l i thin
thi near-wallll turbulent
t b l t boundary
b
d
layers
l
att
industrial Reynolds numbers
Motivation for DES: combination of LES and RANS. RANS
i id attached
inside
tt h d boundary
b
d
layer
l
and
d LES in
i the
th separated
t d
regions

31

DES Formulation
Modification to RANS models was straightforward by
substituting the length scale d w , which is the distance to the
closest wall
wall, with the new DES length scale,
scale l defined as:

l = min(d w , CDES )

where
h C

= max( x , y , z )

iis the
th DES constant,
t t is
i the
th grid
id spacing
i andd is
i
based on the largest dimensions of the local grid cell, and x , y , z
are the grid spacing in x, y and z coordinates respectively
Applying the above modification will result in S-A Based,
standard k- (or k-) based and Menters SST based DES
models etc.
models,
etc
DES

k
DRANS
= *k = k 3 2 / lk
k
DES

32
k
=

l

lk = k 1 2 ( * )
l = min(lk , CDES )
32

Resolved/Modeled/Total Reynolds
stress (DES)
TKE
Modeled

Resolved
Total

Modeled Reynolds stress: u 'v ' = t u + v

y

33

Resolved/Modeled/Total Reynolds
stress (URANS)
Modeled

Total

Resolved

34

Deep insight into RANS/DES (Point oscillation on

free surface with power spectral analysis)
EFD

DES

URANS

-5/3

2HZ

35

Direct Numerical Simulation (DNS)

DNS is to solve the Navier-Stokes equation directly

without averaging or approximation other than numerical
discretizations whose errors can be estimated (V&V) and
controlled.
t ll d
The domain of DNS must be at least as large as the
physical domain or the largest turbulent eddy (scale L)
The size of the grid must be no larger than a viscously
determined scale, Kolmogoroff scale,
The number of grid points in each direction must be at
least L/
Th computational
The
t ti
l costt is
i proportional
ti
l to
t

Re 3L/ 4 (0.01 Re )

34

Provide detailed information on flow field

Due to the computational cost, DNS is more likely to be a
research tool, not a design tool.

36

Examples (Diffuser)
Asymmetric diffuser with
separation is a good test case
for turbulence models.
A inlet channel was added at
the diffuser inlet to generate
fully developed velocity profile
Boundary layer in the lower
diffuser wall will separate due
to the adverse pressure
gradient.
Results shown next include
comparisons
i
between
b t
V2f and
d
k-
LES simulation of this geometry
can be found in:
M. Fatica, H. J. Kaltenbach, and
R. Mittal, Validation of LES in a
Plain Asymmetric Diffuser,

center for turbulence research

research,
annual research briefs, 1997

37

Examples (Diffuser)

Mean velocity
predicted by V2f
agreed very well
with
ith EFD data,
d t
particular the
separation
region is
captured.
K- model fails
to predict the
separation
caused by
adverse
pressure
gradient.

38

Examples (Diffuser)
v2f

k-

TKE predicted by V2f agreed better with EFD

data than k- model, particular the
asymmetric distribution.
distribution
Right column is for the skin friction coefficient
on the lower wall, from which the separation
and reattachment point can be found.

2x/H
39

References
1. J. H. Ferziger, M. Peric, Computational Methods for Fluid
Dynamics, 3rd edition, Springer, 2002.
2 D.
2.
D C
C. Wilcox
Wilcox, Turbulence
Turbulence Modeling for CFD
CFD, 1998.
1998
3. S.B.Pope, Turbulent Flows, Cambridge, 2000
4. P. A. Durbin and B. A. Pettersson Reif, Statistical Theory and
Modeling for Turbulent Flows, John Wiley & Sons, LTD, 2001.
5. P.A. Davidson, Turbulence: An Introduction for Scientists
and Engineers,
g
, Oxford,, 2004.

40