Guidelines for Setting up Laminar-

Turbulent Transition Cases in ANSYS CFD

Prepared and compiled by

Dr Aleksey Gerasimov
Version 2.0 European Technology Group
September, 2014
ANSYS UK Ltd
1 © 2013 ANSYS, Inc. October 6, 2014 Aleksey.Gerasimov@ansys.com
 Table of Contents
• Types of Available Transition Models 3
• Transition Model Detail 4
• Mesh Requirements 6
• Solver Methods 7
• Solver Controls 8
• Model Activation: γ ─ θ SST model (SST-T4) 9
• Model Activation: k – kl – ω 10
• Model Activation: γ SST model (SST-T3) 11
• Inlet Boundary Conditions 12
• Case Study: NACA0012 14
• Other Test Cases 22
• Conclusions 38
2 © 2013 ANSYS, Inc. October 6, 2014
 Types of available transition models

• Three models for transition prediction are available in ANSYS Fluent:

– the Transition SST model – 4 equation model (Gamma-Theta SST model)
– the Intermittency Transition model (available for SST, Scale-Adaptive
Simulation with SST, and DES with SST) – 3 equation model (Gamma model)
– the Transition k – kl – ω (only available in Fluent for Standard k – ω, not
available in CFX) – also known as Walters Model
• For many test cases the three models produce similar results
• Among the three models, only the Intermittency Transition model is
capable of accounting for crossflow instability
• Due to the above reasons and their combination with the SST model,
the Transition SST model and the Intermittency Transition model are
recommended over the Transition k – kl – ω model

! Please also refer to the Fluent User’s and Theory guide for more information
3 © 2013 ANSYS, Inc. October 6, 2014
 Transition model detail 1/2
• The γ ─ θ SST model (SST-T4) – 4 equation model (Gamma-Theta SST model)
– solves 4 turbulent equations: k, ω, γ, Reθ
– The transition SST model is based on the coupling of the SST – k ω, transport equations with
two other transport equations, one for the intermittency and one for the transition onset
criteria, in terms of momentum-thickness Reynolds number. An ANSYS empirical correlation
(Langtry and Menter) has been developed to cover standard bypass transition as well as flows
in low freestream turbulence environments.

• The γ SST model (SST-T3) – 3 equation model (Gamma SST model)

– solves 3 turbulent equations: k, ω, γ
– Most recent in-house development effort from ANSYS
– Has a special option to account for cross-flow transition
– The γ transition model has the following advantages over the γ-Reθ transition model: It reduces
the computational effort (by solving one transport equation instead of two). It avoids the
dependency of the equation on the velocity . This makes the transition model Galilean
invariant. It can therefore be applied to surfaces that move relative to the coordinate system
for which the velocity field is computed.
• Both models are:
– Built on SST k ─ ω turbulence closure. Based on empirical correlations to account for transition.
– Available in RANS as well as in scale-resolving mode (SAS-SST & DES-SST)
– Can be fine-tuned through correlations via UDF
4 © 2013 ANSYS, Inc. October 6, 2014
– Available in ANSYS Fluent and in ANSYS CFX
 Transition model detail 2/2
• Transition k – kl – ω model (also known as Walters)
– Solves 3 turbulent equations: k, kl , ω
– kl is the energy associated with the non-turbulent Tollmien-Schlichting instabilities that are present
in the transition region.
– The model aims to capture the physical transfer of energy from initial non-turbulent oscillations to
fully turbulent perturbations: kl → k.
– Strong reliance on empirical correlations still remains in k – kl – ω model too.
– The original formulation of the model was based on solution of k, kl & ɛ equations. The current
model was reformulated from ɛ to ω = ɛ/k and is now based on ω ─ equation to account better for
the effects caused by pressure gradients.
– Available in RANS mode only.
– Certain amount of tuning can be performed via constant tuning or source term
cancellation/modification via UDF
– Available in ANSYS Fluent only

• A number of examples is shown in this document and the choice of the model should be
based upon best performance in a specific type of flows.
– Gamma-Theta SST model (SST-T4) & Gamma Model SST (SST-T3) typically return very similar,
sometimes almost identical results, however, the 3-equation (SST-T3) model for some test cases
looked more robust and tended to converge faster.
– The k – kl – ω model generally shows similar transition location and converges nicely, however, the
5
dynamics of the transition
© 2013 ANSYS, Inc. October 6, 2014
process is often different.
 Mesh requirements
• It is absolutely crucial to have mesh resolution with y + ≤ 1 .
• Hexahedral meshes in the near-wall regions would be the most economical
approach, they will return the most accurate results and are preferred
• Gradual expansion of the mesh in wall normal directions, the expansion ratio
should ideally be 1.15 or less. Values of 1.2 can be used as absolute maximum.
• Aspect ratio for the near-wall cells can be high, however, it is advisable to keep
it below 200, if possible.
• Domain decomposition or use of the mesh inflation layers becomes necessary.

6 © 2013 ANSYS, Inc. October 6, 2014

 Solver Methods.
• Very often, the steady state solution can be achieved.
• It is difficult to converge transition models with the segregated solver and the
SST-T4 ( i.e. γ ─ θ) model can even clip to a wrong solution unless a pressure-
based coupled solver is used.
• Pressure-based coupled and, in particular, the pressure-based coupled pseudo-
transient solver is preferred.
• Green-Gauss Node-Based or Least Squares Cell-Based gradient reconstruction
would be needed.
• Use of higher order discretisation schemes (at least 2nd order) for mean
momentum and energy equations is a must.
• Use of the 2nd order discretisation scheme for turbulent quantities is generally
recommended, however, it has been found that no visible difference existed in
the results for NACA0012 airfoil when using either the 1st or 2nd order
discretisation for turbulent quantities.
• The convergence is always more challenging with higher order descretisation
for turbulent quantities. Use High Order Term Relaxation in such cases.
7 © 2013 ANSYS, Inc. October 6, 2014
 Solver Controls
• The default settings do not provide consistent convergence rates for all
equations. Therefore, some adjustments are needed to ensure convergence.
• The table below shows the best found choice of URF’s for all 3 models in
NACA0012 airfoil at moderate Reynolds numbers for a range of AoA whilst
using the pseudo-transient pressure-based solver.
• The pseudo-transient time step can be evaluated by dividing the cord length by
free-stream velocity.
• In some cases, but not always, the k – kl – ω model might allow more aggressive
time stepping and thus might offer a faster convergence (as for the NACA0012).
U, V P k kl ω γ θ Pseudo ∆t
(user-specified)

SST-T4 0.2 - 0.3 0.5 NA 0.5 0.5 - 0.5 - 0.75 0.0001

0.3 0.75
k - kl - ω 0.2 0.3 0.5 - 0.5 - 0.5 - NA NA 0.001 –
0.3 0.3 0.3 0.001 - 10-4
SST-T3 0.2 0.3 0.5 NA 0.5 0.6 - NA 0.0001
0.75
8
- Can be beneficial
© 2013 ANSYS, Inc.
in some cases.
October 6, 2014
 Model Activation: γ ─ θ SST model (SST-T4)
• Activation:
– Directly from the “Viscous Model” panel.
– In 3D solver, the model can be used with the
hybrid approaches such as SAS or DES-SST.
• Have a number of options available:
– Roughness Correlation: for modelling
transition on rough walls. Geometric
roughness K needs to be specified instead of
the sand-grain roughness Ks.
– Curvature Correction: an additional
correction that is designed to account for the
deficiencies of the SST closure to model flows
that involve significant amount of swirl.
– Production Limiter: limits excessive amount of turbulent kinetic energy production, Pk ,
by limiting the ratio of turbulent kinetic energy production to its dissipation rate to the
maximum default value of 10.
– Production Kato-Launder: a more elaborate way to limit excessive levels of Pk by
making it a function of, not only the strain rate S, but also a function of the vorticity
rate which is relatively small in stagnation regions due to low levels of local rotation.
9 © 2013 ANSYS, Inc. October 6, 2014
 Model Activation: k – kl – ω
• Activated directly from the main
“Viscous Model” panel
• This is the model with the smallest
amount of user inputs
– Does not include roughness correlations
– Does not need curvature correction
– Does not require production limiters

• The Viscous Heating option is available if the Energy Equation is enabled.

– This option should be enabled if either the viscous dissipation leads to a
considerable amount of heating in near-wall compressible flows or if the
rheological heating (in non-Newtonian fluids) takes place.
• The Compressibility Effects option becomes available if the ideal or real gas
options are chosen in the Materials panel.
– Designed to improve predictions in free shear layers, caution is needed in near-
wall regions and its effect on transition. Worth checking the sensitivity of the
10 © 2013 ANSYS, Inc. October 6, 2014 results with and without this option.
 Model Activation: γ SST model (SST-T3)

• This model does not have a

predefined dedicated radio button
in the Viscous Model panel, instead
it is activated as an option either
under k ─ ω SST model for RANS or
under SAS or DES-SST flavours of
hybrid Scale-Resolving Simulations.
• γ SST model has got a number of
options available that are similar to
γ ─ θ SST model and the main
differences in options are as follows:
– No Roughness Correlation.
– Cross-flow Transition Option: an
additional feature that allows one to
account for the effects of the cross-
flow instabilities in such flows as

11 © 2013 ANSYS, Inc. October 6, 2014

 Inlet Boundary Conditions 1/2
• The mechanism and location of the laminar-turbulent transition is very
sensitive to the free-stream conditions.
External disturbance – Bypass transition occurs if the
leading to instability external flow outside the laminar
boundary layer has a high level of
turbulence (> 1%).
– Example: Compressor or Turbine
blade, where upstream blades have
generated large disturbances traveling
Laminar Turbulent
Boundary Layer Boundary Layer
with the freestream.

• It is worth mentioning that the values for turbulent intensity, TU , and

turbulent viscosity ratio, μ t /μ, that are introduced at inlets to the domain
might change significantly by the time they reach a surface in question
inside the computational domain.
• The actual levels of turbulence in the vicinity of the object need to be
checked. If the calculated figures are incorrect then modifications to the
prescribed inlet values become necessary

12 © 2013 ANSYS, Inc. October 6, 2014

 Inlet Boundary Conditions 2/2
• It is desirable to have a relatively low Turbulent Viscosity Ratio (1 to
10) and estimate the inlet value of turbulence intensity such that at the
leading edge of the blade/airfoil, the turbulence intensity has decayed
to the desired value.
• The decay of turbulent kinetic energy can be assessed from the levels
of the inlet turbulence intensity, TU , and the free-stream velocity, U.
• The correlation between the inlet values and the values in the
proximity of a surface can be evaluated * :
0.5
  *  Example of decay of Turbulence
 2  3 U x  TU inlet  

2
 Intensity (TU) as a Function of
TU  TU inlet 1  
 2 ( t /  )inlet  Streamwise Distance (x)
   
 
where
x ─ streamwise distance from the inlet
μt /μ – turbulent to molecular viscosity ratio
ρ ─ density
  0.09,  *  0.0828

* Please see the accompanying Excel sheet where this formula is implemented and allows one to calculate the decay of TU and
µ t /µ and refer to Section “Specifying Inlet Turbulence Levels” in ANSYS Fluent “Theory Guide” under “Turbulence” and
“Transition SST Model”
13 © 2013 ANSYS, Inc. October 6, 2014
 Case Study: NACA0012 Problem
Definition & Test Matrix
Wind Turbines:
Rec is an order of
α magnitude lower than in
aeronautical applications
Inlet Values:
μt /μ = 5, γ = 1 or kl =1e-6
TU = 0.05% → reduced to 0.048% in the proximity of the airfoil (for all 3 models)

Re
α 0 4 5 7
SST γ-θ (4 eq.) Not SST γ-θ (4 eq.)
Re=6.0·105 Not
k-kl -ω (3 eq.) Tested k-kl -ω (3 eq.)
40m/s Tested
SST γ (3 eq.) SST γ (3 eq.)
Not SST γ-θ (4 eq.) Not SST γ-θ (4 eq.)
Re=1.05·106
Tested k-kl -ω (3 eq.) Tested k-kl -ω (3 eq.)
70 m/s SST γ (3 eq.) SST γ (3 eq.)
14 © 2013 ANSYS, Inc. October 6, 2014
 Convergence History
SST-T3
SST-T4

• Extensive number of solution controls k - kl - ω

have been tried. These are model-
specific (see Page 5).
• SST-T3 & k-kl -ω generally converge
faster than SST-T4.
• k-kl -ω shows a smoother behaviour of
residuals than SST-T3, but this is case-
dependent.

15 © 2013 ANSYS, Inc. October 6, 2014

 Postprocessing:
Separation Regions and Their Extent

Please Note: Fixed Length Vector Plots: used due to very low, near 0, velocity values

k-kl -ω is shown here as it has got the longest separation region at the suction side

17 © 2013 ANSYS, Inc. October 6, 2014

 Postprocessing: Turbulence Characteristics for
Intermittency Models
SST-T3: intermittency SST-T4: intermittency

SST-T3: Turbulent Kinetic Energy SST-T4: Turbulent Kinetic Energy

18 © 2013 ANSYS, Inc. October 6, 2014

 Identifying Transition with γ and kl models
SST-T3: intermittency k-kl -ω : Laminar Kinetic Energy

SST-T3: Turbulent Kinetic Energy k-kl -ω : Turbulent Kinetic Energy

19 © 2013 ANSYS, Inc. October 6, 2014

 Transition Onset, x / C, comparison with XFOIL
Angle of Attack 4ᵒ 4ᵒ 7ᵒ 7ᵒ
Oncoming Wind 40 70 40 70
Speed, m/s (Re) (600 000) (1 050 000) (600 000) (1 050 000)
Transition Onset
29%-35% 23%-31% 5.9%-7.2% 4.5%-5.3%
XFOIL (N = 9-15)

SST γ-θ (4 eq.) 42% 40% 8.2% 6.8%

k-kl -ω (3 eq.) 41% 38% 7.7% 6.1%

SST γ (3 eq.) 41% 37% 7.7% 6.1%

Situation Ncrit Ti % The e n method in XFOIL has the user-specified parameter

Sailplane 12-14 0.020-0.009 "Ncrit", which is the log of the amplification factor of the most-
amplified frequency which triggers transition.
Motorglider 11-13 0.030-0.013
A suitable value of this parameter depends on the ambient
Clean wind tunnel 10-12 0.046-0.020 disturbance level in which the airfoil operates, and mimics
the effect of such disturbances on transition. Here are some
Average wind tunnel 9 0.070 typical values of Ncrit for various situations.
Dirty wind tunnel 4-8 0.563-0.106 The location is predicted very similarly by all 3 models
and γ-based models return almost identical results
20 © 2013 ANSYS, Inc. October 6, 2014
 NACA0012: Mesh Sensitivity Results.
• All transitional models should be used with meshes of the maximum of y +
values not exceeding 1, y +<1.
• A few mesh sensitivity tests have been performed for NACA0012 airfoil with
different near-wall sizes (all cases with y +<1) and the expansion ratios of 1.2
and 1.07.
• The results for drag and lift coefficients remained almost identical in all cases
and for all 3 models with the maximum deviation of results being within 1.7%.

21 © 2013 ANSYS, Inc. October 6, 2014

 Other Test Cases
• List of test cases:

– Flat plate with zero dp/dx (T3A, T3B, T3A-)

– Flat plate with non-zero dp/dx (T3C2, T3C3 and T3C4)
– VPI turbine
– 3D RGW Compressor cascade
– Wind turbine airfoil
– S809 (wind turbine)
– A airfoil
– McDonell Douglas 3 element Flap
– Zierke & Deutsch Compressor Cascade
– 3D NREL Wind Turbine
The 3-equation γ SST model results are not shown here, and although one can
assume that they might not be very different from γ─θ SST results – this is something
that cannot be taken for granted.

22 © 2013 ANSYS, Inc. October 6, 2014

 Flat Plate Results: dp/dx=0
γ ─ θ SST & k ─ kl ─ ω (Walters) model
Zero Pressure Gradient and Moderate Levels of Free-Stream Turbulence Intensity
T3A
Free-StreamTurbulence Intensity (FSTI) = 3 %

γ ─ θ SST and Walters’ k ─ kl ─ ω giving similar results with SST being somewhat better
23 © 2013 ANSYS, Inc. October 6, 2014
 Flat Plate Results: dp/dx
γ ─ θ SST & k ─ kl ─ ω (Walters) model
T3C2 (transition near suction peak) T3C3 (transition in an adverse dp/dx)
FSTI = 2.5 % FSTI = 2.5 %

dp/dx < 0 dp/dx > 0

25 © 2013 ANSYS, Inc. October 6, 2014

 Separation Induced Transition
γ ─ θ SST & k ─ kl ─ ω (Walters) model

Laminar Turbulent
Separation Reattachment

• k ─ kl ─ ω delays the separation induced transition.

• γ ─ θ SST returns better predictions by allowing intermittency to increase above 1.0 in
laminar separation regions
26 © 2013 ANSYS, Inc. October 6, 2014
 VPI Turbine Setup
γ ─ θ SST & k ─ kl ─ ω (Walters) model
• Hybrid Mesh: 24386 cells
• Re = 23,000, Uin = 5.85m/s, Tin=20 °C, Lchord = 59.4cm
• Fluid is air, constant Cp and density
• Three Inlet Turbulent intensities: 0.6%, 10% and 19.5%

27 © 2013 ANSYS, Inc. October 6, 2014

 VPI turbine: Tu = 10.0% and 0.6%
γ ─ θ SST & k ─ kl ─ ω (Walters) model Note: Experiment authors indicated that the
instrumentation for the heat flux measurements was
inducing early transition in the heat transfer data

k ─ kl ─ ω returns better results

28 © 2013 ANSYS, Inc. October 6, 2014
 Test Cases: 3D RGW Compressor Cascade
γ ─ θ SST model

Hub
Vortex

Laminar
Separation
Bubble

Tip
Vortex

Transition

RGW Compressor (RWTH Aachen)

FSTI = 1.25 %
Rex = 430 000
Loss coefficient, (Yp) = 0.097
Yp = (poinlet - pooutlet)/pdynoutlet

29 © 2013 ANSYS, Inc. October 6, 2014

 Test Cases: 3D RGW Compressor Cascade
γ ─ θ SST model

Fully Turbulent Experimental Oil Flow Transition Model

Yp = 0.19 Yp = 0.097 Yp = 0.11

Transition γ ─ θ SST model results are in a

good agreement with the experiments

30 © 2013 ANSYS, Inc. October 6, 2014

 Wind Turbine Airfoil γ ─ θ SST model
Tu Contour
Transition

Transition

Transition

Results are only available with γ ─ θ SST model. These results are in a good agreement with
31
the experimental
© 2013 ANSYS, Inc.
and XFOIL results for a wide range of different angles of attack
October 6, 2014
 2D S809 Airfoil
k ─ kl ─ ω model

• Laminar Flow Airfoil for Wind

Turbine Applications
• Rex = 2 000 000,  = 0° to 20°
• Experiment:
• 2D: Tested in the low-
turbulence wind tunenel at Delft
University of Technology,
(Somers, 1989)
• 3D: Used as the profile for the
NREL Phase IV full wind turbine
experiment, (Simms, 2001)

32 © 2013 ANSYS, Inc. October 6, 2014

 S809: Transition Location xt/c
k ─ kl ─ ω model

Results are only available with k ─ kl ─ ω model. These results are in a good agreement with
33
the experimental
© 2013 ANSYS, Inc. October 6, 2014
results for a wide range of different angles of attack
 A-airfoil: Skin Friction Coefficient
γ ─ θ SST & k ─ kl ─ ω (Walters) model

Both models predict transition accurately.

γ ─ θ SST model shows better agreement
with the experimental data in the
turbulent region of the flow.

Experiment: transition location x/c=0.12

34 © 2013 ANSYS, Inc. October 6, 2014
 McDonnell Douglas 30P-30N 3-Element Flap
γ ─ θ SST model
Re = 9 million Tu Contour
Mach = 0.2
C = 0.5588 m
AoA = 8°

Exp. hot film

Main lower transition:
transition CFX = 0.587
location Exp. = 0.526
measured Error: 6.1 %
Flap transition:
as f(x/c)
CFX = 0.909
Exp. = 0.931
Main upper transition:
Error: 2.2 %
CFX = 0.068
Exp. = 0.057
Error: 1.1 %

Slat transition:
CFX = -0.056
Exp.= -0.057
Error: 0.1 %

Results are only available with γ ─ θ SST model. These results are in a good agreement with
35 © 2013 ANSYS, Inc. October 6, 2014
the experimental data
 Zierke & Deutsch Compressor Cascade
γ ─ θ SST model

Very challenging case, especially for the dynamics of the

Tu = 0.18% transition process. The Transition γ ─ θ SST, as well as
Incidence Angle = - 1.5% γ SST model*, results show good agreement with the
experiment for the location of the transition point.
36 © 2013 ANSYS, Inc. October 6, 2014 * ─ not shown in this plot
 3D NREL Wind Turbine
γ ─ θ SST model

At a wind speed of 20 m/s, the flow topology computed with the fully turbulent and the
transitional approaches are very different. This results in a 80% change in output torque. The
lower output torque appears to be the result of a laminar separation in the leading edge region
of the suction side of the blade. The transitional simulation is in much closer agreement with
the experimental data.

37 © 2013 ANSYS, Inc. October 6, 2014

 Conclusions
• State-of-the-art transition models available in ANSYS CFD software.
– Based strictly on local variables
– Applicable to unstructured massively parallelized codes
• Onset of transition is predicted completely automatically.
– User must specify correct values of inlet turbulence intensity and the length scale
• Validated for a wide range of 2-D and 3-D turbo-machinery and aeronautical
test cases.
• Computational effort is moderate and only increase comes from the need to
resolve boundary layer.
• For the NACA 0012 test case the 3-equation models looked more robust and
converged better than the 4-equation model.
• The 3-eqn. γ SST model has received a very considerable amount of
development in the recent years with thorough in-house knowledge of the
model. It can also account for the cross-flow transition explicitly.
• The intermittency-based 4-equation model is most tuneable and might appeal
to very experienced users.
38
! Please also refer
© 2013 ANSYS, Inc.
to the Fluent User’s and Theory Guide for more information !
October 6, 2014