Hi there,

I have a problem calculating the flow in a radial fan, typically used in HVACs. When using the multiple reference frame of rotations I got wrong results at the 'tongue': a huge vortex disturbs the whole flow structure. When doing a tansient analysis with arbitrary sliding mesh motion this vortex disappears and the solution looks very good. I am very grateful for any suggestions how to get the right results with MRF.

Thanks, Thomas

Hi, Thomas

When you post-processing the results with rotating frames, the velocities are relative to that reference frame. I just wonder if you used in PROSTAR command RFRAME STATIC, before GETC ALL -this will load velocities in STATIC frame. Please let me know if this is the case.

Dear Luo,

yes, this is exactly the case. The velocities within the 'rotating' part look ok, but behind the interface there is a huge vertex, which I can't find in a transient solution. I tried to move the interface closer to the spiral housing and closer to the blades, with no improvements to the solution.

Regards,

thomas

I think what you are seeing is par for the course. Having performed a number of these analyses a while back I came to the conclusion that the 'phase difference' between the flow entering the blade on one side and exiting the blade on the other side is important to the operation of this kind of fan. This seemed slightly surprising given the relatively short blade involved. However in the near tongue region, if I remember correctly, the radial velcity is not that high and therefore the circumferential movement while in the blade passage was not insubstantial.

If you start the transient from the MFR solution it may be possible, by viewing the vortex decay, to decide what feature of the transient is suppressing this flow structure. It will most likely be of only academic interest as it will probably be impossible to replicate the necessary flow physics in an MFR environment.

Using coarse time steps at the beginning of the transient and dropping it once the flowfield is established is probably the most practical way of reducing the run time, which is the real disadvantage of the transient.

Dear Sir, I have a domain which has both rotating and Stationary parts. I would like know which solver ,relaxation, differencing scheme will give better convergence. I solved the same using STAR. The convergence was not good. Can you help me