© Fluent Inc.

5/16/2012 D1
Fluent Software Training
TRN-98-006
Modeling Turbulent Flows
© Fluent Inc. 5/16/2012 D2
Fluent Software Training
TRN-98-006
 Unsteady, aperiodic motion in which all three velocity components
fluctuate  mixing matter, momentum, and energy.
 Decompose velocity into mean and fluctuating parts:
U
i
(t) ÷ U
i
+ u
i
(t)







 Similar fluctuations for pressure, temperature, and species
concentration values.
What is Turbulence?
Time
U
i
(t)
U
i

u
i
(t)
© Fluent Inc. 5/16/2012 D3
Fluent Software Training
TRN-98-006
Why Model Turbulence?
 Direct numerical simulation of governing equations is only possible for
simple low-Re flows.
 Instead, we solve Reynolds Averaged Navier-Stokes (RANS)
equations:



where (Reynolds stresses)



















 Time-averaged statistics of turbulent velocity fluctuations are modeled
using functions containing empirical constants and information about
the mean flow.
 Large Eddy Simulation numerically resolves large eddies and models
small eddies.
(steady, incompressible flow
w/o body forces)
j i ij
u u R µ ÷ =
j
ij
j j
i
i k
i
k
x
R
x x
U
x
p
x
U
U
c
c
+
c c
c
+
c
c
÷ =
c
c
2
µ µ
© Fluent Inc. 5/16/2012 D4
Fluent Software Training
TRN-98-006
Is the Flow Turbulent?
External Flows
Internal Flows
Natural Convection
5
10 5× >
x
Re along a surface
around an obstacle
where
µ
µUL
Re
L
÷ where
Other factors such as free-stream
turbulence, surface conditions, and
disturbances may cause earlier
transition to turbulent flow.
L = x, D, D
h
, etc.
,300 2 >
h
D
Re
10 8
10 10 ÷ > Ra
µo
µ |
3
TL g
Ra
A
÷
20,000 >
D
Re
© Fluent Inc. 5/16/2012 D5
Fluent Software Training
TRN-98-006
How Complex is the Flow?
 Extra strain rates
 Streamline curvature
 Lateral divergence
 Acceleration or deceleration
 Swirl
 Recirculation (or separation)
 Secondary flow
 3D perturbations
 Transpiration (blowing/suction)
 Free-stream turbulence
 Interacting shear layers
© Fluent Inc. 5/16/2012 D6
Fluent Software Training
TRN-98-006
Choices to be Made
Turbulence Model
&
Near-Wall Treatment
Flow
Physics
Accuracy
Required
Computational
Resources
Turnaround
Time
Constraints
Computational
Grid
© Fluent Inc. 5/16/2012 D7
Fluent Software Training
TRN-98-006
Zero-Equation Models

One-Equation Models
Spalart-Allmaras
Two-Equation Models
Standard k-c
RNG k-c
Realizable k-c
Reynolds-Stress Model

Large-Eddy Simulation

Direct Numerical Simulation
Turbulence Modeling Approaches
Include
More
Physics
Increase
Computational
Cost
Per Iteration
Available
in FLUENT 5
RANS-based
models
© Fluent Inc. 5/16/2012 D8
Fluent Software Training
TRN-98-006
 RANS equations require closure for Reynolds stresses.


 Turbulent viscosity is indirectly solved for from single transport
equation of modified viscosity for One-Equation model.
 For Two-Equation models, turbulent viscosity correlated with turbulent
kinetic energy (TKE) and the dissipation rate of TKE.


 Transport equations for turbulent kinetic energy and dissipation rate are
solved so that turbulent viscosity can be computed for RANS equations.
Reynolds Stress Terms in RANS-based Models
Turbulent
Kinetic Energy:
Dissipation Rate of
Turbulent Kinetic Energy:
c
µ µ
µ
2
k
C
t
÷ Turbulent Viscosity:
Boussinesq Hypothesis:
(isotropic stresses)
|
|
.
|

\
|
c
c
+
c
c
+ ÷ = ÷ =
i
j
j
i
t ij j i ij
x
U
x
U
k u u R µ o µ µ
3
2
2 /
i i
u u k ÷
|
|
.
|

\
|
c
c
+
c
c
c
c
÷
i
j
j
i
j
i
x
u
x
u
x
u
v c
© Fluent Inc. 5/16/2012 D9
Fluent Software Training
TRN-98-006
 Turbulent viscosity is determined from:



 is determined from the modified viscosity transport equation:




 The additional variables are functions of the modified turbulent
viscosity and velocity gradients.
One Equation Model: Spalart-Allmaras
( )
2
1
2
2
~
1
~ ~ ~
~
1
~
~
~
d
f c
x
c
x x
S c
Dt
D
w w
j
b
j j
b
v
µ
v
µ
v
v µ µ
o
v µ
v
µ
v
÷
(
(
¸
(

¸

|
|
.
|

\
|
c
c
+
¦
)
¦
`
¹
¦
¹
¦
´
¦
c
c
+
c
c
+ =
( )
( )
(
¸
(

¸

+
=
3
1
3
3
/
~
/
~
~
v
v v
v v
v µ µ
c
t
v
~
Generation Diffusion
Destruction
© Fluent Inc. 5/16/2012 D10
Fluent Software Training
TRN-98-006
One-Equation Model: Spalart-Allmaras
 Designed specifically for aerospace applications involving wall-
bounded flows.
 Boundary layers with adverse pressure gradients
 turbomachinery
 Can use coarse or fine mesh at wall
 Designed to be used with fine mesh as a “low-Re” model, i.e., throughout
the viscous-affected region.
 Sufficiently robust for relatively crude simulations on coarse meshes.
© Fluent Inc. 5/16/2012 D11
Fluent Software Training
TRN-98-006
Two Equation Model: Standard k-c Model
Turbulent Kinetic Energy
Dissipation Rate
c c c
o o
2 1 ,
, , C C
k
are empirical constants
(equations written for steady, incompressible flow w/o body forces)
Convection
Generation Diffusion
Destruction

µc o µ µ µ ÷
)
`
¹
¹
´
¦
c
c
c
c
+
c
c
|
|
.
|

\
|
c
c
+
c
c
=
c
c
        
        
  
i
k t
i i
j
j
i
i
j
t
i
i
x
k
x x
U
x
U
x
U
x
k
U ) (
Destruction
Convection
Generation Diffusion
           
            
  
|
|
.
|

\
|
÷
)
`
¹
¹
´
¦
c
c
c
c
+
c
c
|
|
.
|

\
|
c
c
+
c
c
|
.
|

\
|
=
c
c
k
C
x x x
U
x
U
x
U
k
C
x
U
i
t
i i
j
j
i
i
j
t
i
i
2
2 1
) (
c
µ
c
o µ µ
c c
µ
c c c
© Fluent Inc. 5/16/2012 D12
Fluent Software Training
TRN-98-006
Two Equation Model: Standard k-c Model
 “Baseline model” (Two-equation)
 Most widely used model in industry
 Strength and weaknesses well documented
 Semi-empirical
 k equation derived by subtracting the instantaneous mechanical energy
equation from its time-averaged value
 c equation formed from physical reasoning
 Valid only for fully turbulent flows
 Reasonable accuracy for wide range of turbulent flows
 industrial flows
 heat transfer
© Fluent Inc. 5/16/2012 D13
Fluent Software Training
TRN-98-006
Two Equation Model: Realizable k-c
 Distinctions from Standard k-c model:
 Alternative formulation for turbulent viscosity
where is now variable

 (A
0
, A
s
, and U* are functions of velocity gradients)
 Ensures positivity of normal stresses;
 Ensures Schwarz’s inequality;

 New transport equation for dissipation rate, c:



c
µ µ
µ
2
k
C
t
÷
c
µ
k U
A A
C
s o
*
1
+
=
0 u
2
i
>
2
j
2
i
2
j i
u u ) u u ( s
b
j
t
j
G c
k
c
k
c S c
x x Dt
D
c c
c
c
vc
c
µ c µ
c
o
µ
µ
c
µ
3 1
2
2 1
+
+
÷ +
(
(
¸
(

¸

c
c
|
|
.
|

\
|
+
c
c
=
Generation Diffusion Destruction Buoyancy
© Fluent Inc. 5/16/2012 D14
Fluent Software Training
TRN-98-006
 Shares the same turbulent kinetic energy equation as Standard k-c
 Superior performance for flows involving:
 planar and round jets
 boundary layers under strong adverse pressure gradients, separation
 rotation, recirculation
 strong streamline curvature
Two Equation Model: Realizable k-c
© Fluent Inc. 5/16/2012 D15
Fluent Software Training
TRN-98-006
Two Equation Model: RNG k-c
Turbulent Kinetic Energy
Dissipation Rate
Convection Diffusion
Dissipation
 
µc µ o µ µ ÷
|
|
.
|

\
|
c
c
c
c
+ =
c
c
      
  
i
k
i
t
i
i
x
k
x
S
x
k
U
eff
2
Generation
|
|
.
|

\
|
c
c
+
c
c
÷ ÷
j
i
i
j
ij ij ij
x
U
x
U
S S S S
2
1
, 2
where
are derived using RNG theory
c c c
o o
2 1 ,
, , C C
k
(equations written for steady, incompressible flow w/o body forces)
Additional term
related to mean strain
& turbulence quantities
Convection Generation Diffusion Destruction

R
k
C
x x
S
k
C
x
U
i i
t
i
i
÷
|
|
.
|

\
|
÷
|
|
.
|

\
|
c
c
c
c
+
|
.
|

\
|
=
c
c
         
       
2
2 eff
2
1
c
µ
c
µ o µ
c c
µ
c c c
© Fluent Inc. 5/16/2012 D16
Fluent Software Training
TRN-98-006
Two Equation Model: RNG k-c
 k-c equations are derived from the application of a rigorous statistical
technique (Renormalization Group Method) to the instantaneous Navier-
Stokes equations.
 Similar in form to the standard k-c equations but includes:
 additional term in c equation that improves analysis of rapidly strained flows
 the effect of swirl on turbulence
 analytical formula for turbulent Prandtl number
 differential formula for effective viscosity
 Improved predictions for:
 high streamline curvature and strain rate
 transitional flows
 wall heat and mass transfer
© Fluent Inc. 5/16/2012 D17
Fluent Software Training
TRN-98-006
Reynolds Stress Model
k
ijk
ij ij ij
k
j i
k
x
J
P
x
u u
U
c
c
+ ÷ u + =
c
c
c µ
Generation
k
i
k j
k
j
k i ij
x
U
u u
x
U
u u P
c
c
+
c
c
÷
|
|
.
|

\
|
c
c
+
c
c
' ÷ ÷ u
i
j
j
i
ij
x
u
x
u
p
k
j
k
i
ij
x
u
x
u
c
c
c
c
÷ µ c 2
Pressure-Strain
Redistribution
Dissipation
Turbulent
Diffusion
(modeled)
(related to c)
(modeled)
(computed)
(equations written for steady, incompressible flow w/o body forces)
Reynolds Stress
Transport Eqns.
Pressure/velocity
fluctuations
Turbulent
transport
) (
j ik i jk k j i ijk
u u p u u u J o o + ' + =
© Fluent Inc. 5/16/2012 D18
Fluent Software Training
TRN-98-006
Reynolds Stress Model
 RSM closes the Reynolds-Averaged Navier-Stokes equations by
solving additional transport equations for the Reynolds stresses.
 Transport equations derived by Reynolds averaging the product of the
momentum equations with a fluctuating property
 Closure also requires one equation for turbulent dissipation
 Isotropic eddy viscosity assumption is avoided
 Resulting equations contain terms that need to be modeled.
 RSM has high potential for accurately predicting complex flows.
 Accounts for streamline curvature, swirl, rotation and high strain rates
 Cyclone flows, swirling combustor flows
 Rotating flow passages, secondary flows

© Fluent Inc. 5/16/2012 D19
Fluent Software Training
TRN-98-006
Large Eddy Simulation
 Large eddies:
 Mainly responsible for transport of momentum, energy, and other scalars,
directly affecting the mean fields.
 Anisotropic, subjected to history effects, and flow-dependent, i.e., strongly
dependent on flow configuration, boundary conditions, and flow parameters.
 Small eddies:
 Tend to be more isotropic and less flow-dependent
 More likely to be easier to model than large eddies.
 LES directly computes (resolves) large eddies and models only small
eddies (Subgrid-Scale Modeling).
 Large computational effort
 Number of grid points, N
LES
·
 Unsteady calculation
2
Re
t
u
© Fluent Inc. 5/16/2012 D20
Fluent Software Training
TRN-98-006
Comparison of RANS Turbulence Models
Model Strengths Weaknesses
Spalart-
Allmaras
Economical (1-eq.); good track record
for mildly complex B.L. type of flows
Not very widely tested yet; lack of
submodels (e.g. combustion,
buoyancy)
STD k-c
Robust, economical, reasonably
accurate; long accumulated
performance data
Mediocre results for complex flows
involving severe pressure gradients,
strong streamline curvature, swirl
and rotation
RNG k-c
Good for moderately complex
behavior like jet impingement,
separating flows, swirling flows, and
secondary flows
Subjected to limitations due to
isotropic eddy viscosity
assumption
Realizable
k-c
Offers largely the same benefits as
RNG; resolves round-jet anomaly
Subjected to limitations due to
isotropic eddy viscosity
assumption
Reynolds
Stress
Model
Physically most complete model
(history, transport, and anisotropy of
turbulent stresses are all accounted
for)
Requires more cpu effort (2-3x);
limited near-wall modeling options;
tightly coupled momentum and
turbulence equations
© Fluent Inc. 5/16/2012 D21
Fluent Software Training
TRN-98-006
Near-Wall Treatments
 Most k-c and RSM turbulence
models will not predict correct
near-wall behavior if integrated
down to the wall.
 Special near-wall treatment is
required.
 Standard wall functions
 Nonequilibrium wall functions
 Two-layer zonal model
Boundary layer structure
© Fluent Inc. 5/16/2012 D22
Fluent Software Training
TRN-98-006
Standard Wall Functions
µ t
µ
/
2 / 1 4 / 1
w
P P
k C U
U ÷
-
( )
¦
¹
¦
´
¦
>
(
¸
(

¸

+
<
=
-
-
-
) ( ln
1
Pr
) ( Pr
* *
* *
T t
T
y y P Ey
y y y
T
k
µ
µ
µ P P
y k C
y
2 / 1 4 / 1
÷
-
q
k C c T T
T
P p P w
' '
÷
÷

2 / 1 4 / 1
) (
*
µ
µ
Mean Velocity
Temperature
where
where
and P is a function of the fluid
and turbulent Prandtl numbers.
thermal sublayer thickness
( )
- -
= Ey U ln
1
k
© Fluent Inc. 5/16/2012 D23
Fluent Software Training
TRN-98-006
Nonequilibrium Wall Functions
 Log-law is sensitized to pressure gradient for
better prediction of adverse pressure gradient
flows and separation.
 Relaxed local equilibrium assumptions for
TKE in wall-neighboring cells.
 Thermal law-of-wall unchanged
|
|
.
|

\
|
=
µ
µ
k
µ t
µ µ
y
k C
E
k C
U
w
2 / 1 4 / 1 2 / 1 4 / 1
ln
1
/
~
(
¸
(

¸

+
÷
+
|
.
|

\
|
÷ =
- -
µ µk µk
y
k
y y
y
y
k
y
dx
dp
U U
v v
v
v
2
2 / 1 2 / 1
ln
2
1
~
where
© Fluent Inc. 5/16/2012 D24
Fluent Software Training
TRN-98-006
Two-Layer Zonal Model
 Used for low-Re flows or
flows with complex near-wall
phenomena.
 Zones distinguished by a wall-
distance-based turbulent
Reynolds number
 High-Re k-c models are used in the turbulent core region.
 Only k equation is solved in the viscosity-affected region.
 c is computed from the correlation for length scale.
 Zoning is dynamic and solution adaptive.
µ
µ y k
Re
y
÷
200 >
y
Re
200 <
y
Re
© Fluent Inc. 5/16/2012 D25
Fluent Software Training
TRN-98-006
Comparison of Near Wall Treatments
Strengths Weaknesses
Standard wall
Functions
Robust, economical,
reasonably accurate
Empirically based on simple
high-Re flows; poor for low-Re
effects, massive transpiration,
Vp, strong body forces, highly
3D flows
Nonequilibrium
wall functions
Accounts for Vp effects,
allows nonequilibrium:
-separation
-reattachment
-impingement
Poor for low-Re effects, massive
transpiration, severe Vp, strong
body forces, highly 3D flows
Two-layer zonal
model
Does not rely on empirical
law-of-the-wall relations,
good for complex flows,
applicable to low-Re flows
Requires finer mesh resolution
and therefore larger cpu and
memory resources
© Fluent Inc. 5/16/2012 D26
Fluent Software Training
TRN-98-006
Computational Grid Guidelines
Wall Function
Approach
Two-Layer Zonal
Model Approach
 First grid point in log-law region

 At least ten points in the BL.
 Better to use stretched quad/hex
cells for economy.
 First grid point at y
+
~ 1.
 At least ten grid points within
buffer & sublayers.
 Better to use stretched quad/hex
cells for economy.
500 50 s s
+
y
© Fluent Inc. 5/16/2012 D27
Fluent Software Training
TRN-98-006
Estimating Placement of First Grid Point
 Estimate the skin friction coefficient based on correlations either
approximate or empirical:

 Flat Plate-

 Pipe Flow-

 Compute the friction velocity:

 Back out required distance from wall:
 Wall functions • Two-layer model


 Use post-processing to confirm near-wall mesh resolution
2 . 0
Re 0359 . 0 2 /
÷
~
L f
c
2 . 0
Re 039 . 0 2 /
÷
~
D f
c
2 / /
f e w
c U u = ÷ µ t
t
y
1
= 250v/u
t
y
1
= v/ u
t
© Fluent Inc. 5/16/2012 D28
Fluent Software Training
TRN-98-006
Setting Boundary Conditions
 Characterize turbulence at inlets & outlets (potential backflow)
 k-c models require k and c
 Reynolds stress model requires R
ij
and c
 Several options allow input using more familiar parameters
 Turbulence intensity and length scale
 length scale is related to size of large eddies that contain most of energy.
 For boundary layer flows: l ~ 0.4o
99

 For flows downstream of grids /perforated plates: l ~ opening size
 Turbulence intensity and hydraulic diameter
 Ideally suited for duct and pipe flows
 Turbulence intensity and turbulent viscosity ratio
 For external flows:
 Input of k and c explicitly allowed (non-uniform profiles possible).
10 / 1 < <
µ µ
t
© Fluent Inc. 5/16/2012 D29
Fluent Software Training
TRN-98-006
GUI for Turbulence Models
Define  Models  Viscous...
Turbulence Model options
Near Wall Treatments
Inviscid, Laminar, or Turbulent
Additional Turbulence options
© Fluent Inc. 5/16/2012 D30
Fluent Software Training
TRN-98-006
Example: Channel Flow with Conjugate Heat
Transfer
adiabatic wall
cold air
V = 50 fpm
T = 0 °F
constant temperature wall T = 100 °F
insulation
1 ft
1 ft
10 ft
P
Predict the temperature at point P in the solid insulation
© Fluent Inc. 5/16/2012 D31
Fluent Software Training
TRN-98-006
Turbulence Modeling Approach
 Check if turbulent  Re
D
h
= 5,980
 Developing turbulent flow at relatively low Reynolds number and
BLs on walls will give pressure gradient  use RNG k-c with
nonequilibrium wall functions.
 Develop strategy for the grid
 Simple geometry  quadrilateral cells
 Expect large gradients in normal direction to horizontal walls 
fine mesh near walls with first cell in log-law region.
 Vary streamwise grid spacing so that BL growth is captured.
 Use solution-based grid adaption to further resolve temperature
gradients.
© Fluent Inc. 5/16/2012 D32
Fluent Software Training
TRN-98-006
Velocity
contours
Temperature
contours
BLs on upper & lower surfaces accelerate the core flow
Prediction of Momentum & Thermal
Boundary Layers
Important that thermal BL was accurately resolved as well
P
© Fluent Inc. 5/16/2012 D33
Fluent Software Training
TRN-98-006
Example: Flow Around a Cylinder
wall
wall
1 ft
2 ft
2 ft
air
V = 4 fps
Compute drag coefficient of the cylinder
5 ft 14.5 ft
© Fluent Inc. 5/16/2012 D34
Fluent Software Training
TRN-98-006
 Check if turbulent  Re
D
= 24,600

 Flow over an object, unsteady vortex shedding is expected,
difficult to predict separation on downstream side, and close
proximity of side walls may influence flow around cylinder
 use RNG k-c with 2-layer zonal model.

 Develop strategy for the grid
 Simple geometry & BLs  quadrilateral cells.
 Large gradients near surface of cylinder & 2-layer model
 fine mesh near surface & first cell at y
+
= 1.
 Use solution-based grid adaption to further resolve pressure
gradients.
Turbulence Modeling Approach
© Fluent Inc. 5/16/2012 D35
Fluent Software Training
TRN-98-006
Grid for Flow Over a Cylinder
© Fluent Inc. 5/16/2012 D36
Fluent Software Training
TRN-98-006
Prediction of Turbulent Vortex Shedding
Contours of effective viscosity µ
eff
= µ + µ
t
C
D
= 0.53 Strouhal Number = 0.297
U
D
St
t
÷ where
© Fluent Inc. 5/16/2012 D37
Fluent Software Training
TRN-98-006
Summary: Turbulence Modeling Guidelines
 Successful turbulence modeling requires engineering judgement of:
 Flow physics
 Computer resources available
 Project requirements
 Accuracy
 Turnaround time
 Turbulence models & near-wall treatments that are available
 Begin with standard k-c and change to RNG or Realizable k-c if
needed.
 Use RSM for highly swirling flows.
 Use wall functions unless low-Re flow and/or complex near-wall
physics are present.

Fluent Software Training TRN-98-006

What is Turbulence?
 

Unsteady, aperiodic motion in which all three velocity components fluctuate  mixing matter, momentum, and energy. Decompose velocity into mean and fluctuating parts: Ui(t)  Ui + ui(t)
ui(t) U i (t) Ui

Time

Similar fluctuations for pressure, temperature, and species concentration values.
D2 © Fluent Inc. 5/16/2012

Fluent Software Training TRN-98-006

Why Model Turbulence?

Direct numerical simulation of governing equations is only possible for simple low-Re flows. Instead, we solve Reynolds Averaged Navier-Stokes (RANS) equations: R (steady, incompressible flow Ui p  2Ui U k    ij w/o body forces) xk xi x j x j x j where

Rij    ui u j

(Reynolds stresses)

D3

© Fluent Inc. 5/16/2012

and disturbances may cause earlier transition to turbulent flow. ReD  20. 5/16/2012 . surface conditions. Dh. etc. D.300 Other factors such as free-stream turbulence.000 Internal Flows Re Dh  2.Fluent Software Training TRN-98-006 Is the Flow Turbulent? External Flows where Re x  5  10 5 along a surface around an obstacle UL ReL   L = x. Natural Convection Ra  108  1010 where gTL3  Ra   D4 © Fluent Inc.

Fluent Software Training TRN-98-006 How Complex is the Flow?  Extra strain rates       Streamline curvature Lateral divergence Acceleration or deceleration Swirl Recirculation (or separation) Secondary flow     3D perturbations Transpiration (blowing/suction) Free-stream turbulence Interacting shear layers D5 © Fluent Inc. 5/16/2012 .

5/16/2012 .Fluent Software Training TRN-98-006 Choices to be Made Flow Physics Computational Resources Turbulence Model & Near-Wall Treatment Computational Grid Accuracy Required Turnaround Time Constraints D6 © Fluent Inc.

Fluent Software Training TRN-98-006 Turbulence Modeling Approaches Zero-Equation Models One-Equation Models Spalart-Allmaras Include More Physics RANS-based models Two-Equation Models Standard k-e RNG k-e Realizable k-e Reynolds-Stress Model Large-Eddy Simulation Direct Numerical Simulation Available in FLUENT 5 Increase Computational Cost Per Iteration D7 © Fluent Inc. 5/16/2012 .

 U U  2 Rij    ui u j    k ij  t  i  j  Boussinesq Hypothesis:  x  3 (isotropic stresses)  j xi    Turbulent viscosity is indirectly solved for from single transport equation of modified viscosity for One-Equation model. Turbulent Viscosity: t  C k2 e  Transport equations for turbulent kinetic energy and dissipation rate are solved so that turbulent viscosity can be computed for RANS equations. Turbulent Kinetic Energy: k  uiui / 2 ui  ui u j  Dissipation Rate of e     Turbulent Kinetic Energy: x j  x j xi    D8 © Fluent Inc. For Two-Equation models. 5/16/2012 .Fluent Software Training TRN-98-006 Reynolds Stress Terms in RANS-based Models  RANS equations require closure for Reynolds stresses. turbulent viscosity correlated with turbulent kinetic energy (TKE) and the dissipation rate of TKE.

D9 © Fluent Inc. 5/16/2012 .Fluent Software Training TRN-98-006 One Equation Model: Spalart-Allmaras  Turbulent viscosity is determined from: ~   / 3  ~ t    ~ 3  /   c 13     ~  is determined from the modified viscosity transport equation: ~~ 1      cb1S   Dt  ~  x j  ~ D Generation ~ ~ 2 ~        ~     c     cw1 f w 2     b2   x j  d     x j    Diffusion Destruction  The additional variables are functions of the modified turbulent viscosity and velocity gradients.

. throughout the viscous-affected region.Fluent Software Training TRN-98-006 One-Equation Model: Spalart-Allmaras  Designed specifically for aerospace applications involving wallbounded flows. Sufficiently robust for relatively crude simulations on coarse meshes.  Can use coarse or fine mesh at wall   D10 © Fluent Inc. 5/16/2012 .   Boundary layers with adverse pressure gradients turbomachinery Designed to be used with fine mesh as a “low-Re” model.e. i.

5/16/2012 . e . incompressible flow w/o body forces) D11 © Fluent Inc. C1e .Fluent Software Training TRN-98-006 Two Equation Model: Standard k-e Model Turbulent Kinetic Energy  U j U i  U j k   k    t    (t  k ) e     x  x x x j  xi  xi  i     i    Destruction  i      U i Convection Generation Diffusion Dissipation Rate e2  e   e   e   U j U i  U j   U i  C1e    t   (  t  e )   C 2e     k   x  x xi x j  i xi  xi  k  i                  Convection Generation Diffusion Destruction  k . C2e are empirical constants (equations written for steady.

5/16/2012 .Fluent Software Training TRN-98-006 Two Equation Model: Standard k-e Model  “Baseline model” (Two-equation)   Most widely used model in industry Strength and weaknesses well documented k equation derived by subtracting the instantaneous mechanical energy equation from its time-averaged value e equation formed from physical reasoning  Semi-empirical     Valid only for fully turbulent flows Reasonable accuracy for wide range of turbulent flows   industrial flows heat transfer D12 © Fluent Inc.

Fluent Software Training TRN-98-006 Two Equation Model: Realizable k-e  Distinctions from Standard k-e model:  Alternative formulation for turbulent viscosity t  C    k2 e where C  1 Ao  As U *k is now variable e (A0. ( ui u j )2  ui2 u 2 j  New transport equation for dissipation rate. 5/16/2012 . e: De    Dt x j     t  e   Diffusion  e  e2 e   c1Se  c2  c1e c3e Gb  x  k k  e  j  Generation D13 Destruction Buoyancy © Fluent Inc. ui2  0 Ensures Schwarz’s inequality. and U* are functions of velocity gradients) Ensures positivity of normal stresses. As.

separation rotation. recirculation strong streamline curvature D14 © Fluent Inc. 5/16/2012 .Fluent Software Training TRN-98-006 Two Equation Model: Realizable k-e   Shares the same turbulent kinetic energy equation as Standard k-e Superior performance for flows involving:     planar and round jets boundary layers under strong adverse pressure gradients.

C1e . incompressible flow w/o body forces) D15 © Fluent Inc.Fluent Software Training TRN-98-006 Two Equation Model: RNG k-e Turbulent Kinetic Energy k   k  2     e U i  t S   x  k eff x   x i   Generation  i  Dissipation  i   Convection Diffusion 1  U j U i   S  2Sij Sij . Sij     x x  2 i j  where Dissipation Rate e2  e   e  e  2  e  eff   C2e      U i  C1e   t S   k  R   x k    xi   x    i     Additional term  i    Convection Generation Diffusion Destruction related to mean strain & turbulence quantities  k . 5/16/2012 .e . C2e are derived using RNG theory (equations written for steady.

5/16/2012 . Similar in form to the standard k-e equations but includes:     additional term in e equation that improves analysis of rapidly strained flows the effect of swirl on turbulence analytical formula for turbulent Prandtl number differential formula for effective viscosity high streamline curvature and strain rate transitional flows wall heat and mass transfer  Improved predictions for:    D16 © Fluent Inc.Fluent Software Training TRN-98-006 Two Equation Model: RNG k-e   k-e equations are derived from the application of a rigorous statistical technique (Renormalization Group Method) to the instantaneous NavierStokes equations.

Fluent Software Training TRN-98-006 Reynolds Stress Model ui u j J ijk Reynolds Stress U k  Pij  ij  e ij  Transport Eqns. 5/16/2012 . incompressible flow w/o body forces) D17 © Fluent Inc. xk xk Generation Pressure-Strain Redistribution Dissipation Turbulent Diffusion Pij  ui uk U j xk  u j uk U i xk (computed)  u u j   ij   p i   x   j xi  (modeled) (related to e) (modeled) ui u j e ij  2  xk xk J ijk  uiu juk  p( jk ui   ik u j ) Turbulent transport Pressure/velocity fluctuations (equations written for steady.

swirl.    Transport equations derived by Reynolds averaging the product of the momentum equations with a fluctuating property Closure also requires one equation for turbulent dissipation Isotropic eddy viscosity assumption is avoided   Resulting equations contain terms that need to be modeled.Fluent Software Training TRN-98-006 Reynolds Stress Model  RSM closes the Reynolds-Averaged Navier-Stokes equations by solving additional transport equations for the Reynolds stresses. rotation and high strain rates   Cyclone flows.  Accounts for streamline curvature. 5/16/2012 . swirling combustor flows Rotating flow passages. secondary flows D18 © Fluent Inc. RSM has high potential for accurately predicting complex flows.

directly affecting the mean fields. and other scalars. boundary conditions.Fluent Software Training TRN-98-006 Large Eddy Simulation  Large eddies:   Mainly responsible for transport of momentum. i. Anisotropic. subjected to history effects. and flow parameters.  Small eddies:     LES directly computes (resolves) large eddies and models only small eddies (Subgrid-Scale Modeling).. strongly dependent on flow configuration. energy. 5/16/2012 . NLES  Reu Unsteady calculation D19 © Fluent Inc.e. Large computational effort   2 Number of grid points. and flow-dependent. Tend to be more isotropic and less flow-dependent More likely to be easier to model than large eddies.

tightly coupled momentum and turbulence equations © Fluent Inc.L. limited near-wall modeling options.). type of flows Robust. combustion. buoyancy) Mediocre results for complex flows involving severe pressure gradients.Fluent Software Training TRN-98-006 Comparison of RANS Turbulence Models Model SpalartAllmaras STD k-e RNG k-e Realizable k-e Reynolds Stress Model Strengths Economical (1-eq. transport. and anisotropy of turbulent stresses are all accounted for) D20 Weaknesses Not very widely tested yet. swirl and rotation Subjected to limitations due to isotropic eddy viscosity assumption Subjected to limitations due to isotropic eddy viscosity assumption Requires more cpu effort (2-3x). swirling flows. reasonably accurate. and secondary flows Offers largely the same benefits as RNG. strong streamline curvature. economical. lack of submodels (e. good track record for mildly complex B. long accumulated performance data Good for moderately complex behavior like jet impingement.g. separating flows. 5/16/2012 . resolves round-jet anomaly Physically most complete model (history.

Special near-wall treatment is required. 5/16/2012 .Fluent Software Training TRN-98-006 Near-Wall Treatments   Most k-e and RSM turbulence models will not predict correct near-wall behavior if integrated down to the wall.    Standard wall functions Nonequilibrium wall functions Two-layer zonal model Boundary layer structure D21 © Fluent Inc.

D22 © Fluent Inc.Fluent Software Training TRN-98-006 Standard Wall Functions Mean Velocity U   1 ln Ey    Temperature where U   U P C 1/ 4k 1/ 2  P w /   C 1/ 4k 1/ 2 yP  P y    Pr y   T    1  Prt  ln Ey  P        * ( y *  yT ) * ( y *  yT ) thermal sublayer thickness (Tw  TP )  c pC 1/ 4 k 1/ 2  P where T *   q and P is a function of the fluid and turbulent Prandtl numbers. 5/16/2012 .

Fluent Software Training TRN-98-006 Nonequilibrium Wall Functions    Log-law is sensitized to pressure gradient for better prediction of adverse pressure gradient flows and separation. 5/16/2012 . Relaxed local equilibrium assumptions for TKE in wall-neighboring cells. Thermal law-of-wall unchanged ~ 1/ 4 1/ 2   C1/ 4 k1/ 2 y  U C k    1 ln  E     w /    2 y  yv yv  ~ dp  yv  y where U  U  1 ln     2 dx   k 1/ 2  yv    k1/ 2     D23 © Fluent Inc.

Zoning is dynamic and solution adaptive. 5/16/2012 . e is computed from the correlation for length scale. Zones distinguished by a walldistance-based turbulent Reynolds number Rey  200 Re y  200 Re y       ky  High-Re k-e models are used in the turbulent core region. D24 © Fluent Inc. Only k equation is solved in the viscosity-affected region.Fluent Software Training TRN-98-006 Two-Layer Zonal Model   Used for low-Re flows or flows with complex near-wall phenomena.

allows nonequilibrium: wall functions -separation -reattachment -impingement Two-layer zonal Does not rely on empirical law-of-the-wall relations. massive transpiration. model good for complex flows.Fluent Software Training TRN-98-006 Comparison of Near Wall Treatments Strengths Standard wall Functions Robust. poor for low-Re effects. reasonably accurate Weaknesses Empirically based on simple high-Re flows. economical. strong body forces. applicable to low-Re flows Requires finer mesh resolution and therefore larger cpu and memory resources D25 © Fluent Inc. p. severe p. highly 3D flows Nonequilibrium Accounts for p effects. massive transpiration. strong body forces. highly 3D flows Poor for low-Re effects. 5/16/2012 .

D26  .Fluent Software Training TRN-98-006 Computational Grid Guidelines Wall Function Approach Two-Layer Zonal Model Approach  First grid point in log-law region   First grid point at y+  1. 5/16/2012 50  y   500   At least ten points in the BL. Better to use stretched quad/hex cells for economy. Better to use stretched quad/hex cells for economy. © Fluent Inc. At least ten grid points within buffer & sublayers.

2  0.2  Compute the friction velocity: u   w /   Ue c f / 2  Back out required distance from wall:  Wall functions y1 = 250/u • Two-layer model y1 = / u  Use post-processing to confirm near-wall mesh resolution D27 © Fluent Inc.0359 ReL c f / 2  0.039 ReD 0. 5/16/2012 .Fluent Software Training TRN-98-006 Estimating Placement of First Grid Point  Estimate the skin friction coefficient based on correlations either approximate or empirical:  Flat PlatePipe Flow- c f / 2  0.

For boundary layer flows: l  0.499 For flows downstream of grids /perforated plates: l  opening size Ideally suited for duct and pipe flows For external flows: 1  Turbulence intensity and hydraulic diameter   Turbulence intensity and turbulent viscosity ratio   / t  10  Input of k and e explicitly allowed (non-uniform profiles possible). D28 © Fluent Inc. 5/16/2012 .Fluent Software Training TRN-98-006 Setting Boundary Conditions  Characterize turbulence at inlets & outlets (potential backflow)   k-e models require k and e Reynolds stress model requires Rij and e Turbulence intensity and length scale     Several options allow input using more familiar parameters  length scale is related to size of large eddies that contain most of energy.

5/16/2012 . or Turbulent Turbulence Model options Near Wall Treatments Additional Turbulence options D29 © Fluent Inc... Laminar. Inviscid.Fluent Software Training TRN-98-006 GUI for Turbulence Models Define  Models  Viscous.

5/16/2012 .Fluent Software Training TRN-98-006 Example: Channel Flow with Conjugate Heat Transfer adiabatic wall cold air V = 50 fpm T = 0 °F 1 ft insulation constant temperature wall T = 100 °F P 1 ft 10 ft Predict the temperature at point P in the solid insulation D30 © Fluent Inc.

980 Developing turbulent flow at relatively low Reynolds number and BLs on walls will give pressure gradient  use RNG k-e with nonequilibrium wall functions. 5/16/2012 . Vary streamwise grid spacing so that BL growth is captured. Use solution-based grid adaption to further resolve temperature gradients. Develop strategy for the grid     Simple geometry  quadrilateral cells Expect large gradients in normal direction to horizontal walls  fine mesh near walls with first cell in log-law region. D31 © Fluent Inc.Fluent Software Training TRN-98-006 Turbulence Modeling Approach    Check if turbulent  ReDh= 5.

Fluent Software Training TRN-98-006 Prediction of Momentum & Thermal Boundary Layers Velocity contours BLs on upper & lower surfaces accelerate the core flow Temperature contours Important that thermal BL was accurately resolved as well P D32 © Fluent Inc. 5/16/2012 .

5 ft Compute drag coefficient of the cylinder D33 © Fluent Inc.Fluent Software Training TRN-98-006 Example: Flow Around a Cylinder wall 2 ft air V = 4 fps 1 ft 2 ft wall 5 ft 14. 5/16/2012 .

600 Flow over an object. D34 © Fluent Inc. Large gradients near surface of cylinder & 2-layer model  fine mesh near surface & first cell at y+ = 1. 5/16/2012 . and close proximity of side walls may influence flow around cylinder  use RNG k-e with 2-layer zonal model.   Develop strategy for the grid    Simple geometry & BLs  quadrilateral cells. difficult to predict separation on downstream side.Fluent Software Training TRN-98-006 Turbulence Modeling Approach  Check if turbulent  ReD = 24. unsteady vortex shedding is expected. Use solution-based grid adaption to further resolve pressure gradients.

Fluent Software Training TRN-98-006 Grid for Flow Over a Cylinder D35 © Fluent Inc. 5/16/2012 .

Fluent Software Training TRN-98-006 Prediction of Turbulent Vortex Shedding Contours of effective viscosity eff =  + t CD = 0. 5/16/2012 .53 Strouhal Number = 0.297 D where St  U D36 © Fluent Inc.

Use wall functions unless low-Re flow and/or complex near-wall physics are present. 5/16/2012 . Use RSM for highly swirling flows.Fluent Software Training TRN-98-006 Summary: Turbulence Modeling Guidelines  Successful turbulence modeling requires engineering judgement of:    Flow physics Computer resources available Project requirements   Accuracy Turnaround time  Turbulence models & near-wall treatments that are available    Begin with standard k-e and change to RNG or Realizable k-e if needed. D37 © Fluent Inc.

Sign up to vote on this title
UsefulNot useful