[artemus]

Dear all,
My issue today is the following : I am calculating a centrifugal pump in a cavitating regime, and I wish to take into consideration the influence of the perturbation of the intake channel which contains an elbow.
Downstream this elbow there is a flow separation that influences the distribution of cavitating clouds between blades of the pump.
For this purpose I decided that I make a separate calculation (steady one, on the mesh containing the elbow) of the intake channel with that elbow, then I put a probing plane on it, and I use the information as the inlet profile (total pressure and velocity directions in cartesian coordinates) for my second calculation (containing only pump itself in rotating motion).
According to the CFX tutorial, I thought that if I set the inlet frame type to stationary, this would azimuthally fix the position of the perturbation generated by the elbow, however, I realized while doing animation on transient results that the perturbation generated by the elbow is following the blade movement, which of course I did not want to simulate.
This is why I would like to ask if there is the correct option to use in my case?
I was thinking of creating either a small stationary domain relative to the piece of intake channel, or maybe to create a modified user-defined variable from the initialization of the inlet BC profile.

Thank you very much in advance for your ideas

[ghorrocks]

It sounds likely that the flow coming off the elbow is coupled to the pump impeller. If this is the case then you really need to run a model containing both so you can capture this coupling.

How close is the pump to the elbow?

[Opaque]

I assume you have a transient simulation, correct ?

Your modelled domain is in the rotating frame, correct ? and your inlet boundary condition profile is fixed in a stationary frame. Therefore, you would like to see the inlet profile moving backwards respect to the rotating frame, correct ?

It would be great if you read the documentation on how coordinate frame different from the global frame (stationary) behave when combined with rotating domains. To make it short, try the following:

Create a local coordinate frame, align it with the global frame (default anyways), and be certain is stationary. Use such frame at the inlet boundary condition you mentioned, run the calculation. You should see the profile fixed in space and not moving with the blades.

Please keep us posted.

[artemus]

I tried to put enough distance to avoid recirculation after the elbow, and also enough distance from the runner in order to avoid interference. To start I would like to limit the number of domains to keep the calculation simpler (to test different options etc.).
I tried the solution proposed by Opaque, but when I see the pressure profile movie of the imported inlet BC on transient result calculation, the elbow trace appears 'fading' with time, becoming totally homogenous after 200 iterations.
However I noticed a following message in CFX-Pre "the boundary 'inlet' references profile data defined by a user function which specifies a different coordinate frame to the one used by the boundary. The boundary coordinate frame will be used for calculating boundary condition values but the user function coordinate frame will be used for any visualization of the profile. As a result, the visualization will not provide an accurate representation of how the profile data is used on this boundary."
How do I check that the profile data is being used correctly?

[Opaque]

You must keep in mind that if the profile is moving in front of the domain (in the relative frame), the profile must include enough data in the circumferential direction to be interpolated. If your profile is of finite pitch, and the simulation runs long enough for the domain to not intersect such pitch you will be getting some "extrapolated values" which seem to be uniform. Summary: the software is very likely doing what is asked to do with the provided data.

How to fix this? If you go to CFX-Pre/Tools you should select Expand Profile Data. This tool allows you to replicate the profile as many time as you wish to cover enough of the rotation of the domain, say full rotation ?

You can then use the expanded profile at the inlet and run your simulation.

[artemus]

Thank you Opaque for your reply. However I am already using 360° data as profile inlet, is there any ideas to improve quality? Maybe to remesh the elbow calculation with much finer mesh?