Hi everybody!

I am dealing with a strongly swirled flow in a cylinder. Air is blown into the domain through slots on the circumference with high tangential velocity. The outlet is like in a cyclone through a outlet tube on the top. As I assume, k-e Model is not the best model for modelling this kind of flow field. However, it converges and gives me a flowfield dominated by a forced vortex.

I also tried to use Reynolds stress model (RSM), and I followed the suggestion from the manual to gradually increase the rotational velocity. At a certain point, it is not possible to further increase the inlet velocity (problems of convergence). The tangential velocity at this point is about a magnitude lower than what the actual velocity would be. The circumferential velocity forms a free vortex (which might be due of the changed inlet condiotions).

Questions:

1) From the inlet conditions, is it possible to estimate in a simple analytical consideration, what type of vortex can be expected, or - in case of an intermediate form of vortex - where (at what distance from the axis) the maximum velocity would occur, so that good starting values can be provided?

2) As far as convergence of the RSM goes, does anyone have experience to help me in this case? Is another approach to convergence more promising? (I tried to take the k-e solution as start values for RSM, and also some other ways of gradually approaching the desired inlet velocity conditions, not to talk about efforts on optimizing solution parameters such underrelaxation,... but so far everything failed )

Thanks in advance for every contribution.

Armin Hofstaedter

Dear Armin

I have also currently considered a flow with strong swirl. As you suggested it is well known that the standard isotropic k-e model only give reasonable agreement on a strong rankine vortex flow, but could give good result when the free vortex flow dominate the systems. In my flow problem, I have tested a RSM model and the turbulence model give encouranging result could show the measured rankine vortex, and show free-vortex dominate if compare to the standard isotropic k-e model.

Computation using RSM on single gid would cost you large compyational time. What about using a multigrid method, or using another powerful computer.

Dear Armin

For the inlet flow condition, as I would understand in rotating flow problem, if your folw rate ins more dominant compared to rotational effect, assuming that the mixing effect at the inlet is weak, the entering flow tend to conserved its angular momentum. So I think a free vortex assumption is probably reasonable.

If the rotational effect is dominat, it is possible that the mixing is significant, and thus there should be a correction on the entering swirl value of entering fluid.

Hope this could help