I am doing simulation about cyclone separator by FLUENT, When I use RNGk-e turbulent model, convergence is easy to achieved, when I switched to RSM turbulent model, first residuals drop, velocity distribution looks like very good, then residuals rise abruptly and keep a horizontal level, velocity distribution deteriorates. All solving adopted PRESTO, QUICK scheme, and steady solver, and when using RSM, drop under-relaxtion facters a bit. How to keep RSM final convergence?

Thank you very much.

LCW

Often times, with cyclones, the RSM captures the precessing vortex core which is an unsteady phenomena. When the RSM is used with the steady solver, the solutions would not converge, typically showing a limit-cycle like behavior. Try the same problem with the unsteady solver. Chance is that you'll see the RSM captures the precession of the vortex core.