[Kosdalak]

Hello,

After a few weeks of trouble, searching for (and often finding) some help here, I finally got my simulation to work. But it runs not very stable and also the results are not quite good. So maybe there are still some mistakes.

My case:
I got a ripped, square channel with cyclic BC with Re=50,000. I want to compare some RANS models, namely the RSM (SSG & LRR) with the k-w-SST and the v2f Model.

To keep this thread relativly short, I want to discuss here only the RSM. The BC at the walls for k, epsilon, nut and p are zeroGradient; for U noSlip and for R (0 0 0 0 0 0).
I started with a quick pre-simulation using the k-epsilon Model and afterwards calculated R. After a stable setting for starting my RSM Run, I switched to more accurate options. These "Control Files" are attached. Because my mesh got around 400 cells with a non-orthonognality of 77, I run some extra iterations. My mesh got in total around 2.2 million cells, so in total the mesh is in a good setting and my y+ for the "first" cell above each surface is around 1.
My timestep is pretty small but thats the only way to get mean Courant number less than 0.8 - even with a max. Courant number of around 9.


Also my residuals behave in a strange kind.



After 20,000 iterations the results for the velocity fields didn't changed even I'm calculating a transient (pisoFoam). And here is the reason for this thread:




--> What could be the reason for the strange velocity results?

• Could it be that my timestep is to small, so the Solver "creates" a steady-state case? (So far I'm not able to increase it, otherwise it crashes because of the Courant number)
• Are my control files "wrong"? Like my Schemes/Solver I'm using wrong? (My knowledge about the mathematics is not quite good. I found some post/information, telling me these are stable and accurate at the same time)
• Do you have some ideas or experience or just some hints for me?

The results for the velocity would be good if I had averaged them over some iterations. But as a result for just one timestep I think these are wrong.





I appreciate all kind of help and wish every reader a happy new year! If you need more information from me, don't hesitate to ask.
Greetings,
Jan

[Milad_ab]

Hi Jan
Thank you very much for sharing your experience.
I've started to learn OpenFOAM and simulated some models (k-epsilon, k-omega, RNG k- epsilon) in backward Facing Step flow. They were Ok.
Now I want to try with RSM model (LRR)
I have a problem with creating the R folder in LRR.
How can I create R in it?
Thanks
Milad

[dlahaye]

simpleFoam tutorial PitzDaily using Reynolds stress tensor (LRR RASModel)

[Kosdalak]

After checking with some colleagues, we figured out that my results don't represent the reality but with a RANS model (also with the RSM) you will not get better results and always these "averged" velocity fields.

Please keep in mind that the models were unchanged and used like implemented in OF v8!


If I get enough time, I will check some other models if they perform better for this Re and geometry.

[lpz456]

Originally Posted by Milad_ab

Hi Jan
Thank you very much for sharing your experience.
I've started to learn OpenFOAM and simulated some models (k-epsilon, k-omega, RNG k- epsilon) in backward Facing Step flow. They were Ok.
Now I want to try with RSM model (LRR)
I have a problem with creating the R folder in LRR.
How can I create R in it?
Thanks
Milad

Hi, I also do the simulation of backward Facing Step flow tutorial with k-epsilon model, but the process always diverged. Could you help me to share your experience？