Difference between domain motion and rotating wall
 

Hi everyone,

Can anyone tell me what is the difference between domain motion and rotating wall formulations? I know that it depends on the case, but what if I can use both on my case? I´m studying an internal flow in a vertical cilynder, something like a centrifugal pump, and I´m getting different results. Assuming a symmetric surface, I can´t see why this happens. I also have been looking on CFX help, but there isn´t such information.

I´ll be grateful if anyone can help me,
Willian

Moving mesh models arbitrary motion of the mesh with the internal mesh moved by mesh smoothing. Rotating domains only allows rotation and the internal mesh is moved by a simple rotation transform.

The result is that moving mesh adds significant CPU time and can make convergence harder, but rotating domains will take about the same time as a stationary mesh simulation and convergence will be similar (ie no extra simulation time or convergence issues).

If the motion is rotation then use a rotating frame of reference. If the motion requires more arbitrary motion then that use moving mesh.

Hi Glenn,

First of all, thanks a lot for your help. Until now I couldn't find any useful information about it. But when I use rotating domain the convergence becomes harder (I'm monitoring RMS and mass flow rates) and the mass flow rate at the outlet is smaller (difference of 16%). I didn't say before, but it's also a multiphase flow simulation.
I know there are many other factors that must be taken into consideration, but can I consider that in general the moving mesh model will provide better results?

Again, thanks for your help.
Willian

No. The moving mesh will not give better results. It is harder to get good results with the moving mesh model.

If the motion is rotation then use the rotating model. It is that simple.

Ok, that is what I wanted to know.

Thanks Glenn!

Your original question was asking about the rotating wall formulation - I take this to mean imposing a rotational velocity on a wall rather than using the moving mesh approach. If you are imposing a rotational wall velocity then it's only valid for motion that is tangential to the wall surface - i.e. you cannot model a blade using this approach.

Yes, in my case I have motion tangential to the wall surface, but like I said before I've decided to use domain motion in my simulations, especially because I'll start to simulate assymetric geometries and the velocity profiles on the wall can become too complex to specify.

But thanks, I appreciate your help.