[**Anny**]

Hello everyone,

I hope my question is not to specific. Or not to unspecific asked (this is my first CFD project)

I am trying to do an analysis on a centrifugal pump under different load conditions. To verify my model I am comparing the performance curve of my cfd model with the real one.
In overload and optimum its about right but when I am getting in part load (q0.7 up to q0.4) the performance stays about the same which is absolutely not accurate.
I tried different hybrid meshes but it doesnt differ. (13< Y+ <65 for the rotor... it should be okay, right?)

My Domains consists of suction area,full rotor, side chamber, volute with outflow area
I read some articles and some took a transient model and some with a steady for the verification of the performance.

I am working with a steady model.
My boundaries:
pt for the inlet, massflow for the outlet and the SST model.
The interface for rot stat is frozen rotor. I already tried different rotor angles (10°,15°,30° tilted).
I once tried using a stage interface and got close to the right performance, but as far as I read and understand, the stage frame is no good for observing the secondary flow.

To verify the performance I use the pt massflowaves at inlet of the rotor and outlet of the volute as well as the massflow influenced pressures at these areas. Apparently the pt at the inlet is to high or the pt at the outlet is to low ...but why?

Has anyone experience with performance of part load or secondary flow in turbomachines using CFX and would like to share his or her experience?

Is it necessary to use a transient model as soon as secondary flow accures?

Thank you all anyways!

Best regards Anny

[ghorrocks]

13< Y+ <65 for the rotor... it should be okay, right?

It is not that simple to work out an appropriate mesh. You really need to do a mesh sensitivity study. And not just on the near wall resolution buy the whole mesh.

Frozen rotor just gives you are representative torque at a single position. You might need to model a number of positions to get a representative torque over the full rotation. But as soon as you have to do this (and you have said you have secondary flows) then I would be concerned that frozen rotor is not sophisticated enough. You probably need a more advanced frame change model.

[**Anny**]

Thanks for your reply and advise Glenn.
I have never done before a sensitivity study:
Here is what I tried already:
4 different grids (Rotor meshes were about 1.3 mil elements, 2.4, 3.1 and 4.2 mil; mesh quality all above 0.3).
From 3.1 to 4.2 the Post results looked alike, the performance (H,Q) got a little poorer. So i compared the visual effects, pressure and velocity at inlet and outlet and performance.
The meshes were improved in the leading edge, trailing edge and volute tongue areas aswell as the amount of prism layers (9 up to 15).
I tried already 4 different angles (0°,10°,15°,30°) in optimum Q1 and part load Q06. The performance varies about 5%.
So I will start reading about the rotating frames. This might effect the performance?

[ghorrocks]

I have no idea what Q1 and Q06 means. Please try not to use jargon.

Yes, the different rotating frames of reference interface conditions can make a big difference. You need to make sure you choose an appropriate model.

[**Anny**]

Nevermind the Q1 and Q06. I got what you were saying.
Thanks so much for your help!