Internal 2D steady simulation results depend on turbulence models strongly.
 

Hello,

Internal 2D steady simulation results depend on turbulence models strongly. You can see the geometry in the attachment. Let's think the flow direction is in plus.
Flow is incompressible. y+ on the wall is between 1 and 2. The used solver is FLUENT and the residual targets are default values.

1. For case of RMS:
>> the velocity x-component in plus is dominant near the wall in Zone 1.
>> I can detect relatively huge detachment(Correct English word for Ablösung?) in Zone 2.

2. k-omega-SST:
>> the velocity x-component in plus on the wall is smaller than the case of RMS in zone 1.
>> Detachment in Zone 2 also smaller than in the case of RMS.

3. k-epsilon:
>> the velocity x-component in plus on the wall can be found only in the half length of zone 1.
>> Detachment in Zone 2 is barely to detected.

For each three models, I also tested different meshes (coarse, moderate, dense), but the results seem like that turbulence models have more effect on the different flow results.
In the case of k-omega-SST with more dense mesh, the results show bigger detachment in zone 2 than results with k-omega-SST and coarser mesh.

Which turbulence model will be best in this problem?

With kind regards
Breznak

Could you show the comparison of the results (for example the 1d velocity profiles) for all cases only for the finest mesh (number of final nodes in x and y?)?

If you are working without wall-modelled BCs you should use some more nodes at y+<1.

As far as I know, k-epsilon uses wall functions and requires much higher values of Y+ . Your Y+ is 1-2, so K-eps results cannot be trusted. Don't know about the Y+ required by the RMS. K-omega requires Y+<1.

Thank you so much for your answer, but unfortunately I cannot upload detailed geometry in public.