You are on page 1of 40

Overview of

Turbulence Modeling
Paul Galpin
ANSYS Inc.
2011 ANSYS, Inc. October 23, 2012 2
Mixing
/combustion
Turbulence Modelling Challenges
Flow Separation
Vortical
Flows
Flow
Reattachment
Corner
Vortices
Transition
Unsteady Effects
2011 ANSYS, Inc. October 23, 2012 3
Turbulent Flow Simulation Methods
RANS
(Reynolds Averaged Navier-
Stokes Simulations)
SRS
(Scale Resolving Simulations)
DNS
(Direct Numerical Simulation)
Numerically solving the full unsteady
Navier-Stokes equations
No modeling is required
A research tool only far too much
information for industrial
applications
Not available in ANSYS CFD
Includes Large Eddy Simulation
(LES)
The motion of the largest eddies is
directly resolved in the calculation,
in at least a portion of the domain,
but eddies smaller than the mesh
are modeled
Inherently unsteady method
Solve Reynolds-averaged Navier-Stokes
equations (time-average)
Steady state solutions are possible
All turbulence is modeled. Larger
eddies are not resolved
RANS turbulence models are the only
modeling approach for steady state
simulation of turbulent flows
This is the most widely used approach
for industrial flows
2011 ANSYS, Inc. October 23, 2012 4
RANS models are (or will be) mostly for wall boundary layers
y
+
-insensitive wall treatment
Accurate separation prediction
Swirl flow (Curvature correction)
Corner flows
Laminar-turbulent transition
For free shear flows use Scale-Resolving Simulation (SRS)
SAS model
DDES models
Zonal/embedded models
LES/WMLES
Synthetic turbulence
Key Elements in Turbulence Modelling
2011 ANSYS, Inc. October 23, 2012 5
RANS
Advantages: For many applications, steady state
solutions are preferable, and for many applications a
good RANS model with a good quality grid will provide
all the required accuracy
Disadvantages: For some flows, challenges associated
with RANS modeling can limit the level of accuracy
that it is possible to attain
SRS
Advantages: Potential for improved accuracy when
the resolution of the largest eddies is important or
when unsteady data is needed
Disadvantages: computationally expensive
Higher grid resolution required
Unsteady simulation with small time steps generates long
run times and large volumes of data
Comparison of SRS and RANS
RANS
Modeling
2011 ANSYS, Inc. October 23, 2012 7
Integration Platform e-equation
e-equation
2-equation models
k-e, BSL, SST
Higher order models
EARSM e
SMC - e
Extensions
Stagnation point
Curvature correction
Rough walls
Reattachment correction
Wall Treatment
Automatic wall treatment
Transition Model
-Re
O
model
Unsteady models
SST-SAS
SST-DES
2011 ANSYS, Inc. October 23, 2012 8
ANSYS Models
Example: Solids suspension in an tall,
unbaffled tank. Reynolds stress model
together with Eulerian granular multiphase
model
Courtesy of the University of Bologna
It is not enough just to provide many choices
More importantly, for the models that are available, emphasis is placed on
Correct implementation
Models should be well understood and tested
Accurate and validated for some class(es) of applications
Robust performance on all mesh topologies
Interoperability with other physical models, e.g. multiphase, dynamic mesh,
.
Wall treatment
2011 ANSYS, Inc. October 23, 2012 9
Near Wall the e-
equation reduces to
an elliptic equation:
k- e Near Wall
Elliptic Relaxation
2
log
2
2
1
;
6
vis
u
y
C
u
y
t

t
e
v
k
e
| v
+
+
=
=
2
( )
( )
( )
j
t
k
j j j
U
P
t x k x x
e
e
e e e
o |e
o
(
c
c c c
+ = + +
(
c c c c
(

2
2
2
j
x
e
|e
c
=
c
Information about the wall presence is
transmitted by the Poisson equation +
boundary conditions (no damping
required)
Log layer:
Sublayer:
2011 ANSYS, Inc. October 23, 2012 10
Engine installation drag
AIAA Drag Prediction
Workshop 2003
WB
WBPN
Wide range of results
Mainly specialized aeronautics codes
Drag prediction is difficult!!
Cd - Drag
Cl
Lift
Cd - Drag Cd - Drag
Part of this work was supported by research grants from the European
Union under the FLOMANIA project
2011 ANSYS, Inc. October 23, 2012 11
Engine installation drag
ANSYS Results
Drag Polar
WB
WBPN
Grid refinement
Accurate prediction of lift and drag
Improved results under grid refinement
Web-page for AIAA Drag Prediction workshop
Cd - Drag
Cd - Drag
Cl
Lift
2011 ANSYS, Inc. October 23, 2012 12
Corner Flows
Early separation of linear Eddy
Viscosity Models in corners observed
Can be caused by lack of anisotropy in
the stress formulation(differences in
normal stresses near wall)
Anisotropy is cause of secondary flows
into the corner
Reynolds Stress Model (RSM) could
account for this but is often not
robust enough for complex flows
Explicit Algebraic RSM (EARSM) offer
an attractive alternative with reduced
numerical effort and increased
robustness
of cross section of square
duct. Secondary flow into
corner
2011 ANSYS, Inc. October 23, 2012 13
WJ-EARSM-BSL
Wallin-Johansson
2
3
ij i ij ij j
u u k a t o
| |
' ' = = +
|
\ .
1 1, 2 2, 3 3, 4 4, 6 6, ij ij ij ij ij ij
a T T T T T | | | | | = + + + +
1, 2, 3, 4,
6,
1 1
; ; ;
3 3
2
;
3
ij ij ij ik kj S ij ij ik kj ij ij ik kj ik kj
ij ik kl lj ik kl lj ij ij
T S T S S II T II T S S
T S S IV II S
o o
o
O
O
= = = O O = O O
= O O +O O
1 2 3 4 6
1 1
2 1
, 0, , , ,
N IV N
Q NQ Q Q
| | | | | = = = = =
c
+ ' =
k
P
C N
4
9
1
( )
1 1 1 1
9
1.2; 1 , 1.8
4
A C C C ' = = =
Non-linearity due to P
k
Linear part of Stress-Strain relation
2011 ANSYS, Inc. October 23, 2012 14
Stanford Diffuser
Flow Topology
Flow topology depends
strongly on turbulence
model
Stress anisotropy
necessary to obtain
correct behaviour
X/H=16
2011 ANSYS, Inc. October 23, 2012 15
Streamline Curvature

Streamline curvature and system rotation are typical for many turbulent flows
of practical interest
However, conventional eddy viscosity models often fail to capture important
flow features in such flows
This is partially due to the fact that linear eddy viscosity models do not have any
sensitivity to curvature or system rotation effects
2011 ANSYS, Inc. October 23, 2012 16
(
(

c
c
+
c
c
+ + =
c
c
+
c
c
j
t
j
r
t
k
j
j
x x
Cd D f
P
x
U
t
e
o

v
o
e
e
e
e e
) (
) (
) (
1
Rotation/Curvature function
for the SST turbulence model
The only difference between the modified, or SST-CC,
model accounting for the Rotation/Curvature effects and
the pure SST model is multiplier f
r1
in production terms
( ) ( ) 0 . 0 , 25 . 1 , min max
1 rotation r
f f =
(
(

c
c
+
c
c
+ =
c
c
+
c
c
j k
t
j
r k
j
j
x
k
x
k f P
x
k U
t
k
) (
) (
) (
*
1
o

e |

2011 ANSYS, Inc. October 23, 2012 17


Cross Flow & Axial Velocities
Cross Flow Velocity
Z
[
m
]
0 0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
1.2
Exp.
SST
SST-RC
Plane 1, X/C = 0.24 | 18 Jul 2007 |
Axial Velocity/U_inlet
Z
[
m
]
0.8 1 1.2 1.4 1.6 1.8
0.2
0.4
0.6
0.8
1
Exp.
SST
SST-RC
Plane 1, X/C = 0.24 | 18 Jul 2007 |
Cross Flow Velocity
Z
[
m
]
0 0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
1.2
Exp.
SST
SST-RC
Plane 1, X/C = 0.67 | 18 Jul 2007 |
Axial Velocity/U_inlet
Z
[
m
]
0.8 1 1.2 1.4 1.6 1.8
0.2
0.4
0.6
0.8
1
Exp.
SST
SST-RC
Plane 1, X/C = 0.67 | 18 Jul 2007 |
SST
SST-CC
2011 ANSYS, Inc. October 23, 2012 18
Tangential Velocity
Line 1
Line 2
Line 3
Line 4
Line 5
z
x [m]
T
a
n
g
e
n
t
i
o
n
a
l
V
e
l
o
c
i
t
y
[
m
/
s
]
-0.02 -0.01 0 0.01 0.02
-6
-4
-2
0
2
4
6
Exp 1
Exp 2
SST
SST-CC
RSM-SSG
Line 1, Z = - 20 mm | 03 Jul 2007 |
x [m]
T
a
n
g
e
n
t
i
o
n
a
l
V
e
l
o
c
i
t
y
[
m
/
s
]
-0.01 0 0.01
-6
-4
-2
0
2
4
6
Exp 1
Exp 2
SST
SST-CC
RSM-SSG
Line 3, Z = - 53 mm | 03 Jul 2007 |
x [m]
T
a
n
g
e
n
t
i
o
n
a
l
V
e
l
o
c
i
t
y
[
m
/
s
]
-0.005 0 0.005
-6
-4
-2
0
2
4
6
Exp 1
Exp 2
SST
SST-CC
RSM-SSG
Line 5, Z = - 117 mm | 03 Jul 2007 |
x [m]
T
a
n
g
e
n
t
i
o
n
a
l
V
e
l
o
c
i
t
y
[
m
/
s
]
-0.02 -0.01 0 0.01 0.02
-6
-4
-2
0
2
4
6
Exp 1
Exp 2
SST
SST-CC
RSM-SSG
Line 2, Z = - 32 mm | 03 Jul 2007 |
2011 ANSYS, Inc. October 23, 2012 19
Transition Model
Compatible with modern CFD code:
Unknown application
Complex geometries
Unknown grid topology
Unstructured meshes (no search directions)
Parallel codes domain decomposition
Requirements:
Absolutely no search algorithms
Absolutely no integration along lines
Local formulation
Different transition mechanisms
Robust
No excessive grid resolution
Laminar Flow
Transitional
Fully Turbulent
2011 ANSYS, Inc. October 23, 2012 20
Intermittency Equation
( )
( )
j
t
j j f j
U
P E
t x x x

o
(
c | |
c
c c
+ = + +
( |
|
c c c c
(
\ .

| |
0.5
1
(1 )
length onset
P F S F

=
F
onset
transition onset when:
F
onset
linked to exp. transition correlations
t u u
Re Re >
2011 ANSYS, Inc. October 23, 2012 21
Wall Shear @ Rotor 1
SST - Transition k- - Model SST - Model
2011 ANSYS, Inc. October 23, 2012 22
Cost of Transition Model
Eurocopter configuration
6 million nodes
Max y+ = 1
16 CPUs
Total Additional CPU: 17%
Discretization 12%
Linear Solution 5%
Fully Turbulent
Transitional
Drag reduced 5
% compared to
fully turbulent
Drag
Lift
SRS
Modeling
2011 ANSYS, Inc. October 23, 2012 24
Motivation for
Scale-Resolving Simulation (SRS)
Accuracy Improvements
Flows with large separation zones (stalled
airfoils/wings, flow past buildings, flows with
swirl instabilities, etc.)
Enriched Information
Acoustics - Information on acoustic
spectrum not reliable from RANS
Vortex cavitation low pressure inside
vortex causes cavitation resolution of
vortex required
Combustion
Fluid-Structure Interaction (FSI) unsteady
forces determine frequency response
2011 ANSYS, Inc. October 23, 2012 25
SRS by SAS Model
Model based in introduction of von
Karman Length Scale (L
vK
) into scale
equation (e-equation)
Model based on theory of Rotta using an
exact definition of the turbulent length
scale
SAS automatically detects attached flows (RANS)
and unstable separated flows (LES)
Least problematic hybrid RANS-LES model as no
explicit grid dependency in RANS portion
Requires sufficiently strong flow instability to
convert to LES mode
Suitable for numerous technical flows
Flows past bluff bodies
Combustion chambers
Jet in crossflow

URANS
SAS
2011 ANSYS, Inc. October 23, 2012 26
SAS 2-Equation Model (KSKL)
With:
L k = u
( )
( )
2
2
1 2 3
2
1
''
j
t
k t
j
U
P L U k
t x k y y
v
, , v ,
k o
u
c u
(
cu u c cu
| |
+ = +
( |
c c c c
\ .

2 2
'
' ; '' ;
''
i i i i
vK
j j j j k k
U U U U U
U U L
x x x x x x U
k
c c c c
= = =
c c c c c c
Relevant terms can also be transformed and included in other RANS
models (SST):
( )
( )
3/ 2
3/ 4
j
t
k
j j k j
U k
k
k k
P c
t x L x x

v
o
c | |
c
c c
+ = +
|
|
c c c c
\ .
1/ 4
t
c

v = u
2011 ANSYS, Inc. October 23, 2012 27
Types of highly unstable flows:
Flows with strong swirl instabilities
Bluff body flows, jet in crossflow
Massively separated flows
Physics
Resolved turbulence is generated quickly by flow instability
Resolved turbulence is not dependent on details of turbulence in
upstream RANS region (the RANS model can determine the
separation point but from there new turbulence is generated)
Models
SAS: Most easy to use as it converts quickly into LES mode, and
automatically covers the boundary layers in RANS. Has RANS
fallback solution in regions not resolved by LES standards (At, Ax)
DDES: Similar to SAS, but requires LES resolution for all free shear
flows (At, Ax) (jets etc.)
ELES: Not really required as RANS model can cover boundary
layers. Often difficult to place interfaces for synthetic turbulence.
Globally Unstable Flows
Green-recommended, Red=not recommended
2011 ANSYS, Inc. October 23, 2012 28
Types of moderately unstable flows:
Jet flows, Mixing layers
Physics
Flow instability is weak RANS/SAS models stay steady state.
Can typically be covered with reasonable accuracy by RANS
models.
DDES and LES models go unsteady due to the low eddy-viscosity
provided by the models. Only works on fine LES quality grids and
time steps. Otherwise undefined behavior.
Models
SAS: Stays in RANS mode. Covers upstream boundary layers in
RANS mode. Can be triggered into SRS mode by RANS-LES
interface.
DDES: Can be triggered to go into LES mode by fine grid and small
At. Careful grid generation required. Covers upstream boundary
layers in RANS mode.
ELES: LES mode on fine grid and small At. Careful grid generation
required. Upstream boundary layer (pipe flow) in expensive LES
mode. Alternative ELES with synthetic turbulence RANS-LES
interface.
Locally Unstable Flows
Green-recommended, Red=not recommended
BL Turbulence
ML Turbulence
y
x
z
2011 ANSYS, Inc. October 23, 2012 29
One SRS model for entire domain
SAS, DDES ideally suited
Steady boundary conditions
Wall b.l. treated in RANS mode
Separated zones in SRS mode
Globally unstable flow required
Requires strong flow instability
Generates unsteady resolved
turbulence
Easiest SRS model to set up & run
Global SRS Approaches
URANS
Global SRS Model
Courtesy: ETH Zurich
2011 ANSYS, Inc. October 23, 2012 30
Zone with high accuracy
demand within a larger
RANS domain
LES zone coupled to
RANS zone with
synthetic turbulence at
interfaces
LES zone requires
suitable (WM)LES
methods
Integrated or sequential
zonal approaches also
Arbitrarily large
computational savings
Zonal SRS Approaches
Courtesy: Benjamin Duda, Airbus Toulouse
2011 ANSYS, Inc. October 23, 2012 31
ELES: vortex structures
Q-criterion iso-surface colored by Velocity Magnitude
2011 ANSYS, Inc. October 23, 2012 32
Pressure Contours
URANS
ELES
2011 ANSYS, Inc. October 23, 2012 33
U velocity Profiles
x= -163mm
x= -223mm
x= -163mm
x= -223mm
2011 ANSYS, Inc. October 23, 2012 34
U velocity Profiles
x= -3mm
x= -123mm
x= -3mm
x= -123mm
2011 ANSYS, Inc. October 23, 2012 35
Sensors downstream the mirror
10 100 1000
Frequency [Hz]
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
S
P
L

[
d
B
]
Freestream Velocity = 140 km/h
Experimental data
SAS model
Sensor 121
10 100 1000
Frequency [Hz]
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
S
P
L

[
d
B
]
Freestream Velocity = 140 km/h
Experimental data
SAS model
Sensor 122
10 100 1000
Frequency [Hz]
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
S
P
L

[
d
B
]
Freestream Velocity = 140 km/h
Experimental data
SAS model
Sensor 123
Grid ~ 3 million nodes
SRS for Acoustics
2011 ANSYS, Inc. October 23, 2012 36
SRS for IC Engine Flows
Intake
Valve
Exp. RANS DES SAS
3 mm 1 0.95 0.985 0.996
9 mm 1 0.988 - 0.99
Mass flow Rates
Courtesy VW AG Wolfsburg: O. Imberdis, M. Hartmann, H. Bensler, L. Kapitza
VOLKSWAGEN AG, Research and Development, Wolfsburg, Germany
D. Thevenin University of Magdeburg
2011 ANSYS, Inc. October 23, 2012 37
DrivAer Validation Project:
RANS & SRS Modeling
http://www.aer.mw.tum.de
Courtesy by TU Munich, Inst. of Aerodynamics
2011 ANSYS, Inc. October 23, 2012 38
Comparison of the pressure distribution on the
symmetry plane with CFX & Fluent on Mesh3
ANSYS CFX, SST, steady-state, 1ms ANSYS Fluent, SST, steady-state, 1ms
RANS Simulations
Comparison of ANSYS CFX & Fluent
2011 ANSYS, Inc. October 23, 2012 39
SAS-SST Simulation:
10-15 days, 100 cores
2011 ANSYS, Inc. October 23, 2012 40
Summary
RANS modelling key to industrial CFD
Grid quality is key issue
Transition modelling important for many applications
External aeordynamics
Turbomachinery
Wind turbines

SRS is making its way into industrial CFD
Different types of model recommended for different types of
applications
Currently favored methods within ANSYS:
SAS globally unstable flows
DDES globally and locally unstable flows
ELES/WMLES marginally unstable flows

You might also like