CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   Separation Prediction with Menters SST k-omega (

Jonas Larsson March 11, 2002 10:12

Separation Prediction with Menters SST k-omega
Has anyone done any validation studies on how the new k-omega model that is implemented in Fluent 6 performs (Menter's SST model)?

I'm interested in predicting separation on curved surfaces in highly turbulent flows (3D contoured end-walls in axial turbomachinery).

Menter's SST k-omega models works quite well in our in-house codes and I do belive that it could improve separation predictions compared to the two-layer model that I have used previously in Fluent. Before I trust it I'd like to see some validation data though - how these models are implemented can make a big difference.

Jonas Larsson March 13, 2002 04:20

Re: Separation Prediction with Menters SST k-omega
Responding to myself here...

I've tried the SST model a bit now and it did affect separation predictions in some cases (not all) - the trend was a bit earlier separation in adverse pressure gradients than what is obtained with the two-layer Wolfstein model with the Realizable k-epsilon model.

However, the SST model predicts way too high losses - sometimes 50% higher than with the other models. This seems very strange and does not agree with my previous experience from this turbulence model. The standard k-w model in Fluent gives better loss-predictions. I hope that I have done something wrong in the SST simulation, otherwise I suspect that something is not fully correct in the SST implementation.

Anyone else compared loss-predictions with the SST model with other models or experimental data?

kim March 13, 2002 05:12

Re: Separation Prediction with Menters SST k-omega
Can I use the same grid resolution for normal k-w model (not SST) as for realizable-ke model ? Or do I need finer grid? Did you compare k-w (without SST) vs. realiazable-ke in terms of separation loss prediction ? Thanks.

Jonas Larsson March 13, 2002 05:51

Re: Separation Prediction with Menters SST k-omega
I haven't seen any recommendations from Fluent on this.

However, my experience from these models in our other in-house codes are that the Wolfstein's two-layer model is a bit less sensitive to the inner wall resolution - it often produces reasonable results with y+ up to around 2 or 3. The k-w models (both Wilcox's different models and Menter's SST model) usually require a bit better resolution, with y+ below 1. This is also true for the low-Re k-epsilon models available in Fluent's text interface. I ran all models on a very fine grid (y+ < 0.5, stretching < 1.25).

I am just now finishing a loss-prediction test with the std k-w model (not SST) in Fluent and losses seem to agree better with the two-layer model + realizable k-epsilon. This was in 3D (a 3D diffusive axial blade cascade).

I have also done some 2D validations and here the SST models gives good loss predictions. Strange! I still hope that I have made some misstakes in the 3D SST test.

Note that the differences in loss-predictions I'm referring to here are not primarily caused by different separation locations - then it would be natural. The overall flow looks very similar between the models (in most cases, as I mention the SST models predicts a bit earlier separation in some cases and then of course the losses should be different - nothing wrong with that).

Shahjalal March 16, 2002 13:34

Re: Separation Prediction with Menters SST k-omega
I am interested.

Amadou Sowe March 20, 2002 14:44

Re: Separation Prediction with Menters SST k-omega
There is a section on separated flows in "Turbulence Modeling for CFD" by David Wilcox that will be very useful for you to take a look at. He refers to three papers by Menters on the k-Omega turbulence model that will also be good to check. I would have loved to do the problem discussed in this section in Wilcox' book, but I simply do not have the time yet. Please keep us informed on any further verification of k-omega model you have made.

All times are GMT -4. The time now is 08:27.