Convergence problem with Stage, refined mesh
 

Hi all,

I am doing the simulation of a rotor-stator system with radial flow using the MFR-Stage model. Two RANS turbulence models was considered the standard k-epsilon (SKE) and SST model.

Convergence criteria *:

- MAX residual 1.0e-04
- Constant monitoring value at critical locations
- Imbalance <0.1 %

With SKE, yplus ~ 30-100, all convergence criteria are satisfied within 2k iterations.

With SST, the mesh was refined at walls so that the yplus value located in range of 1->3. I found out that when the MAX residual dropped around 1.0e-04, it fluctuates around this value. However, the monitoring variables do not show the oscillations, it change, but quite smooth. I reduced the time scale factor from 5 to 2 and the fluctuation disappeared. The MAX residual dropped out below 1.0e-04. All convergence criteria (*) satisfied well.


My questions are:

1) The final result with reduced time scale 5->2 to avoid the fluctuations may be considered as a converged solution?

2) Why the convergence with refined mesh always occurs with refined mesh ?


Thank you !

1) Yes. 1e-4 is a loose convergence criteria, but if you are happy with it then you are OK to continue.
2) This is unusual, usually convergence is harder to achieve with refined meshes. So you can count yourself lucky in this case.

Thanks for your help, ghorrocks