[Ashkan Kashani]

Hi pals, I hope you are doing well.

I have a question regarding the application of the Generalized k- (GeKo) turbulence method, offered by Fluent, in my study case which is as follows:

The separating-reattaching flow over a sharp-edged rectangular bluff body, as shown in the sketch in the attachments.

I aim to tune the GeKo model for this problem to accurately predict the reattachment length, which the SST model overpredicts by far.

Based on my understanding of the Fluent user's guide and "Best Practice: Generalized k- (GEKO) Two-Equation Turbulence Modeling in Ansys CFD" (https://www.ansys.com/content/dam/am...-Ansys-CFD.pdf), the two free coefficients and are of relevance to separation behaviour. Either reducing or increasing leads to higher eddy viscosity levels in the free shear layer and, by doing so, reduces the reattachment length.

Since both coefficients seem to have the same effect, I am not sure which one should be tuned.

I would appreciate your comment, particularly if someone could provide further information on the underlying physics/mechanism associated with these two coefficients in the GeKo model.

[LoGaL]

So reading from the document you posted, CSep increases eddy viscosity in the boundary layer, Cmin only in undisturbed flow.

Indeed, both of them should affect the location of the reattachment zone. But I would expect Csep to also have something to do with where the flow separates. Either ways, you really should try to change both, making sure that you have more than one case to chjeck the model against. If you have only one case, the tuning makes no sense.

[Ashkan Kashani]

Thank you Lorenzo for your comment. Sorry for my delayed reply as I just noticed your post.

You are right about the fact that both and affect the location of the reattachment point. As a matter of fact, I tweaked them both, only to find pretty similar trends in the results with minor differences in the way the recirculation bubble is affected, at least in terms of the reattachment point.

Originally Posted by LoGaL

But I would expect Csep to also have something to do with where the flow separates.

Agree. However, in my study case, the anticipated influence of is probably prevented by the separation point being fixed at the sharp corner.

Originally Posted by LoGaL

If you have only one case, the tuning makes no sense.

Unfortunately, I am unable to do a thorough tuning due to limited experimental data.

What I am actually trying to achieve from the "tuned" GeKo model in this sense is to compare that to SST in order to understand what mechanisms the SST model is misrepresenting leading to the overprediction of the reattachment length.

This came as a surprise to me since I found SST quite successful in predicting the reattachment point in cases where the submergence t (see the sketch attached to my first post on this thread for definition) is small, in contrast to considerable overprediction in the cases with high t. I am very interested in plausible explanations.

I appreciate your comments and viewpoints.

[LoGaL]

so if i understand well, when this submergence increases, there's higher obstruction to the flow? In other words, the flow section decreases?.

[Ashkan Kashani]

Originally Posted by LoGaL

so if i understand well, when this submergence increases, there's higher obstruction to the flow? In other words, the flow section decreases?.

Correct.

I'm modelling a free surface flow (of definite depth h) intersecting the body. As the submergence t increases, it seems that SST begins to be affected by some false "turbulence-damping" mechanism, reducing turbulence mixing and thus leading to delayed reattachment.
Is this a common feature of SST? In your opinion, what could be behind this behaviour?

I appreciate your time and help.

[LoGaL]

did you check whether your y+ is still below one in the more submerged cases?