Turbulence model choice for automotive simulation?

cfdblender · December 7, 2015, 04:55

Hello everyone,

I'm doing a simulation (3-D, RANS, steady) of a ground vehicle in a windtunnel. The geometry is such that separation should definitely occur at the rear of the vehicle and I'm interested in good prediction of overall drag, the flow separation point, and the wake structure behind the vehicle.

What turbulence model would be best for this work? k-epsilon? or k-omega? SST k-omega?

fluid23 · December 10, 2015, 17:08

If you are looking for separation, go with SST k-omega. K-epsilon does not do well with adverse pressure gradients (the root cause of separation). K-omega models do a much better job with separated flow. If you have time I would recommend doing both Shear Stress Transport (SST) and standard k-omega and compare the differences. Sometimes standard k-omega will over predict the separation. If you don't have time for this then just stick with SST.

cfdblender · December 10, 2015, 17:14

Quote:

Originally Posted by MBdonCFD

If you are looking for separation, go with SST k-omega. K-epsilon does not do well with adverse pressure gradients (the root cause of separation). K-omega models do a much better job with separated flow. If you have time I would recommend doing both Shear Stress Transport (SST) and standard k-omega and compare the differences. Sometimes standard k-omega will over predict the separation. If you don't have time for this then just stick with SST.

Thank you MBdonCFD! Yes - the most important thing in my study is to describe the structure in the separated flow in the wake. I'll go with SST k-omega, then if time allows, k-omega, like you say.

Edeluc · December 18, 2015, 07:02

Dear cfdblender,

I agree with the previous comment. The k-epsilon model is not the best in terms of aerodynamic considerations (only if you need an accurate turbulent time scale, k-epsilon comes in handy).
K-w is a good start and also the v2-f model family might be a thing to look at. As all of these turbulence models are based on assuming an isotropic viscosity, the results are questionable.
If you have time and some experience with it, you could validate the results of k-w and SST-k-w with a Reynolds Stress Model. This also belongs to RANS simulations but does not assume isotropic conditions of the turbulent viscosity (because it models the Reynolds stresses directly).

cfdblender · December 18, 2015, 07:46

Quote:

Originally Posted by Edeluc

Dear cfdblender,

I agree with the previous comment. The k-epsilon model is not the best in terms of aerodynamic considerations (only if you need an accurate turbulent time scale, k-epsilon comes in handy).
K-w is a good start and also the v2-f model family might be a thing to look at. As all of these turbulence models are based on assuming an isotropic viscosity, the results are questionable.
If you have time and some experience with it, you could validate the results of k-w and SST-k-w with a Reynolds Stress Model. This also belongs to RANS simulations but does not assume isotropic conditions of the turbulent viscosity (because it models the Reynolds stresses directly).

Thank you Edeluc for your very comprehensive answer.
I'm working with SST K-omega.
As for RST, I've just looked it up in STAR CCM+ help and it looks like it would be computationally very heavy ! - so I'll try it later.

Edeluc · December 18, 2015, 08:03

Yes, you are right. The RSM approach has some big drawbacks:

1) Instead of 2 PDEs in a two equation turbulence model you have to solve 7 PDEs (6 for the symmetric Reynolds stress tensor and one for the turbulent dissipation).
2) They are more unstable in terms of convergence.
3) The RSM is not often used these days and if more accurate solutions are needed, people usually switch to LES.

I would advise to use it only if you really want to check your two-equation turbulence model for validity. Using it for optimization is just not possible due to the computational effort.

December 7, 2015, 04:55	Turbulence model choice for automotive simulation?	#1
cfdblender New Member Join Date: Jun 2015 Location: Great Britain Posts: 16 Rep Power: 10	Hello everyone, I'm doing a simulation (3-D, RANS, steady) of a ground vehicle in a windtunnel. The geometry is such that separation should definitely occur at the rear of the vehicle and I'm interested in good prediction of overall drag, the flow separation point, and the wake structure behind the vehicle. What turbulence model would be best for this work? k-epsilon? or k-omega? SST k-omega?

December 10, 2015, 17:08		#2
fluid23 Senior Member Matt Join Date: Aug 2014 Posts: 947 Rep Power: 17	If you are looking for separation, go with SST k-omega. K-epsilon does not do well with adverse pressure gradients (the root cause of separation). K-omega models do a much better job with separated flow. If you have time I would recommend doing both Shear Stress Transport (SST) and standard k-omega and compare the differences. Sometimes standard k-omega will over predict the separation. If you don't have time for this then just stick with SST. cfdblender likes this.

December 18, 2015, 07:02		#4
Edeluc New Member Lucas Join Date: Jun 2015 Posts: 20 Rep Power: 10	Dear cfdblender, I agree with the previous comment. The k-epsilon model is not the best in terms of aerodynamic considerations (only if you need an accurate turbulent time scale, k-epsilon comes in handy). K-w is a good start and also the v2-f model family might be a thing to look at. As all of these turbulence models are based on assuming an isotropic viscosity, the results are questionable. If you have time and some experience with it, you could validate the results of k-w and SST-k-w with a Reynolds Stress Model. This also belongs to RANS simulations but does not assume isotropic conditions of the turbulent viscosity (because it models the Reynolds stresses directly). cfdblender likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Turbulence Model in axisymmetric nozzle simulation	cric92	FLUENT	5	September 4, 2017 23:35
General question about SST turbulence model	Richard Renaud	CFD Freelancers	1	December 19, 2015 04:13
Simulation multiphase airlift reactor with Eulerian multiphase model question????	dilok.kumyoo	FLUENT	0	January 28, 2015 02:15
Waterwheel shaped turbine inside a pipe simulation problem	mshahed91	CFX	3	January 10, 2015 11:19
Water subcooled boiling	Attesz	CFX	7	January 5, 2013 03:32

December 18, 2015, 08:03		#6
Edeluc New Member Lucas Join Date: Jun 2015 Posts: 20 Rep Power: 10	Yes, you are right. The RSM approach has some big drawbacks: 1) Instead of 2 PDEs in a two equation turbulence model you have to solve 7 PDEs (6 for the symmetric Reynolds stress tensor and one for the turbulent dissipation). 2) They are more unstable in terms of convergence. 3) The RSM is not often used these days and if more accurate solutions are needed, people usually switch to LES. I would advise to use it only if you really want to check your two-equation turbulence model for validity. Using it for optimization is just not possible due to the computational effort.