[Ashkan Kashani]

Hello all,

I'm doing a 2D simulation of flow beneath a partially-submerged rectangular bluff body. The problem's geometry is shown in the figure attached. I'm using a fully structured hexahedral mesh with a y+ of below 1. When the draft (i.e. t in the figure) is small, the model predicts the reattachment length (i.e. Lr) quite accurately. However, it leads to excessive reattachment lengths for cases of large drafts.
As the draft increases, I wonder what turbulent mechanisms/characteristics become influential that SST fails to capture properly. Could it be the increasing curvature of the streamlines or probably turbulence anisotropy? I appreciate any insights.

Regards,
Armin

[LuckyTran]

I used to study channels with repeated ribs (instead of just 1 body like you shown) and turbulence models get more wrong than correct. Just one quick example is the eddy viscosity near the front face becomes negative due to the kinematic blocking effect occurring over such a short length scale that the energy cannot be dissipated or transferred in any way and all the energy gets shoved into the mean strain. That is, an eddy impinging onto the body must lose its energy but has nowhere to give it to, yet global energy must be conserved. None of the eddy viscosity models today predict negative turbulent viscosity ever.


Now I realize this phenomenon isn't directly related to the mechanism of separated shear layers reattaching that you are asking about. My point is, you should expect less and not more.

[Ashkan Kashani]

Dear LuckyTran,
Thank you for your comment.

Originally Posted by LuckyTran

My point is, you should expect less and not more.

I am not sure if I got it right. What do you mean by "expect less and not more"? Are you referring to the reattachment length? Or do you mean this problem is beyond the capabilities of SST?

Since this overprediction of the reattachment length appears to escalate with increasing the draft (i.e. the submergence t shown in the figure), I suspect that some sort of turbulence-damping mechanism should develop in such cases, which is missing from the SST model.
As you might be aware, there are several modifications to SST to account for mechanisms that SST could not otherwise capture. There is also this GeKo (Generalized K-) model with a number of tunable coefficients that might be helpful. However, I think I need some further insights to try them out in a sensible way.

I would appreciate your viewpoints.

Regards,
Armin

[LuckyTran]

kwsst already has like 12 coefficients. I'd still like to buy a negative turbulent viscosity. How many tunable coefficients is that gonna cost me?

Expect less from turbulence models when you apply them to cases that they are not calibrated for. If you know that they are over/underpredicting, then that is enough.