Hello everybody,

We are trying for a simulation of cyclone separator used to separate dust from air using CFX. We created tetrahedral mesh (500 thousands) keeping y+ value 200. We are treating conditions as steady state. Inlet is velocity inlet, top outlet is pressure outlet having negative static pressure and bottom outlet is closed. Turbulence model used is Shear stress transport model & lagrangian approach used for dispersed phase.full coupling is used. Both coefficient of restitution for all wals are kept zero.

Now we found following problems: 1. solution is not converging rather it is very oscillating behavior. We tried expert parameters also but did hardly any miracle. 2. A virtual wall has been placed at outlet ( 18% area) even after elongation of outlet tube upto 2D length. 3. Values of Efficiency and pressure drop are not matching. If efficiency acheived then pressure drop goes too high and if drop is acheived then efficiency is too low. 4. Y+ value in Post are going upto 500.

We also tried std k-e and RNG models. We would apriciate your kind assistance.

Regards.

- Did you perform a simulation wihtout particles? Is that one unstable as well?

- I think you should use unrelaxation for the momentum of the particles. There are two options. The first option if for the first iteration, default 0.75 = OK. The second is 0.5. Perhaps this one is too high. Set it to 0.1 or 0.2. This implies you need more iterations but you might get a stable solution. Just be patience.....

Good luck, Gert-Jan

www.bunova.nl

Thanks Gert for your invaluable suggestions.

Yes We tried only singlphase simulation. It is extremely stable. We got our solution converged, residuals of order 10-4. Now also we run our simulation for some iterations on singlphase and then inject particle. It is particle tracking which is making the solution very oscilating.

The underrelaxation factor we used 0.55. Now we will try 0.1 or 0.2 too.

Thanks again.

Model which you have used are for isotropic flow, since flow in cyclone is anisotropic, have you tried with RSM model?