[marya.taher]

Hey everyone.
I doing LES on a 29 million mesh at Mach 0.85
Once my RANS calculations were converged, then i switched to transient and embedded LES. I am using k-omega SST.
My time scale is 8e-7 for LES.
Initially, my residuals were in good range from e-7 to e-15. But after crossing around 26k iterations and 4000 time steps, omega residual is increasing, it is in the range of e-1,e-2?
What are the possible causes of this? What is the solution for this ?

[swap]

Hi, I'm just a master's student and not a professional, but could you tell me what flow you're simulating? It seems like there might be an issue with the turbulent dissipation rate with the flow physics that you are simulating.

[Gerry Kan]

Originally Posted by marya.taher

Hey everyone.
I doing LES on a 29 million mesh at Mach 0.85
Once my RANS calculations were converged, then i switched to transient and embedded LES. I am using k-omega SST.
My time scale is 8e-7 for LES.
Initially, my residuals were in good range from e-7 to e-15. But after crossing around 26k iterations and 4000 time steps, omega residual is increasing, it is in the range of e-1,e-2?
What are the possible causes of this? What is the solution for this ?

Dear Marya:

It is not unusual to see a high omega residual, especially near the wall because omega = epsilon / tke, and you are bound to have very high omega near the wall was tke is very low. As long as all your other residuals are low, and the omega residual does not monotonically increase, you are fine.

In addition, since you are preconditioning your flow with a RANS run, there will be a transitional period where the solution field is trying to adjust itself. 4000 timesteps (32 ms, if my calculation is correct) is not a lot of time and you might need to let it run for longer before the flows reaches its (LES) equilibrium.

Hope that helps, Gerry.

[marya.taher]

Originally Posted by swap

Hi, I'm just a master's student and not a professional, but could you tell me what flow you're simulating? It seems like there might be an issue with the turbulent dissipation rate with the flow physics that you are simulating.

Hello
I am a master student as well.
I am simulating external flow. I guess there might be some issue with my mesh

[marya.taher]

Originally Posted by Gerry Kan

Dear Marya:

It is not unusual to see a high omega residual, especially near the wall because omega = epsilon / tke, and you are bound to have very high omega near the wall was tke is very low. As long as all your other residuals are low, and the omega residual does not monotonically increase, you are fine.

In addition, since you are preconditioning your flow with a RANS run, there will be a transitional period where the solution field is trying to adjust itself. 4000 timesteps (32 ms, if my calculation is correct) is not a lot of time and you might need to let it run for longer before the flows reaches its (LES) equilibrium.

Hope that helps, Gerry.

Hey Gerry, Thank you so much. I kept it running but the solution got diverged after almost 4200 time steps. There was a warning of temp limited to 100 in few cells. And i think the main issue is in my mesh and more specifically skewness.

I am doing LES analysis on a wing segment. I have divided the wing into three segments. Inner, outer and center. And i have to do LES on the center wing segment. The remaining wing and aircraft is made from tri mesh, but the center wing segment is of quad mesh.
I have made a 3D volume on that center wing segment in which i will be running LES. whereas the remaining domain will be in RANS.

I have a question regarding the interfaces? What boundary conditions should be given to the interfaces of the 3d volume? as they act as interface between RANS and LES.

Kindly shed some light on it.

[Gerry Kan]

Originally Posted by marya.taher

I have a question regarding the interfaces? What boundary conditions should be given to the interfaces of the 3d volume? as they act as interface between RANS and LES.

Dear Marya:

The transference of scalar quantities between the RANS and LES domain should be fairly straightforward. The only thing you need to worry about is introducing the perturbation. Since, for the most part, I have not done RANS / LES coupling like this (okay yes I have done this, but not the the way you are describing), I cannot really tell you which perturbation model is available for your code, and which is suitable.

But based on my limited knowledge on this, the one you choose should be capable of generating perturbation from a 3D field (i.e., your near wall RANS region) as opposed to 2D (i.e., through a boundary region). Most of these perturbation models should have been tuned for external flow so I (the unqualified self in this case) don't think they should make much of a difference.

Gerry.