[Engineer2k1]

Dear All,
Good Day.
I am using LES and Fluent sliding mesh technique to rotate a small zone.
The small rotating zone (SRZ) contains a rotor while the rest of the domain is stationary. For SRZ I specify a rotational velocity, rotation axis direction (1 0 0) and use "absolute" relative to cell zone specification. To rotate the rotor with SRZ; I use "Moving Wall" with no-slip, "Relative to adjacent cell zone" and "Rotational" as motion with rotation axis direction same as SRZ (1 0 0). The velocity formulation under "General" and under "Solution Initialization">Reference Frame is "Absolute".

Everything runs fine; Except the unsteady statistics; where e.g., Mean X-Velocity is correctly time-averaged in the stationary zone but wrongly averaged (some sort of Phase-Averaging is done instead) in SRZ.

The problem is exactly like, if you plot circumferential velocity (through turbo option) in CFD-Post. However, in CFD-Post I can correctly display by simply selecting the circumferential velocity in stationary Frame available in CFD-Post. Unfortunately, no such thing as Mean Velocity (under Unsteady Statistics) in Stn Frame is available in Fluent. The problem persists even if I export Unsteady statistics to CFD-Post as CFD-Post doesn't give any Stn Frame thing apart from the regular Mean Velocity. Even exporting the Mean-Velocity at every time-step on a plane and importing and then carrying out the mean gives the same problem. Only hack is to export Instantaneous Velocity on every Time-step and then carry out averaging RMSE etc. in Matlab. But why?
Why Fluent has this bug/issue?
Is there a solution?
Please find picture attached; comparing both correct X-Velocity and it's wrong mean in the rotating zone.
Thanks n kind regards,
Shakeel

[LuckyTran]

The unsteady statistics are done not by absolute position but by cell in the mesh. When the SRZ moves, the cell values moves with the moving mesh so that at the end of each time-step when you calculate the statistics, each new sample is done at the new cell positions. This isn't a bug, it's how moving meshes work.


If you wanted the mesh to not move and get averages by absolute position... then don't use a rotating zone / sliding mesh. You can move the rotor still the same way without using a sliding mesh.

[Engineer2k1]

Thanks Lucky, not just for clarifying this; but your valuable comments at CFD-online have always helped me a lot and for years.

Regarding your kind advice;

I tried rotating frame of reference for the SRZ with the same setup as sliding mesh (SM) but it gives same unsteady statistics problem as SM; worse it started eating away the vortical structures (as I ran from an earlier interpolated converged case) in the stationary wake zone region (pic attached).

I also tried rotating frame for the whole domain (SRZ + all the stationary zones) i.e., single reference frame (SRF); though just for a few time-steps as the initial results didn't look good. But now I am thinking to run it full for a couple of revolutions; and may be it converges eventually and solve the problem? If not, can you please suggest me the direction. Thanks.

[LuckyTran]

When you use a single rotating reference frame (as opposed to multiple), Fluent does use the relative velocity formulation of the N-S equations. This should allow you to get the statistics that you want.

Off the top of my head, I can't recall how the unsteady statistics are calculated when using a SRF and whether you need to do any additional post processing or not. I think you are on the right track though. If the statistics are calculated using the relative velocities (and I think they should be) you are in good shape. If not, you might have some trouble.

[Svetlana]

There is some discussion of the different models (single reference frame, multiple reference frame, sliding mesh - and the post processing) in these files:

https://www.afs.enea.it/fluent/Publi.../PDF/chp09.pdf
https://www.afs.enea.it/fluent/Publi.../PDF/chp26.pdf

Would suggest to run a simplified model with less than a million mesh points at first, as this may help to choose the appropriate model and the appropriate post-processing technique before using it for the full simulation.

[Svetlana]

Perhaps also this one.

http://depts.washington.edu/fluidlab...la_rapport.pdf