Centrifugal Pump

[mycom47]

Dear All

is it necessary to have the impeller (solid) as part centrifugal pump analysis? 

I know it can be suppressed, but whether it can helo achieve good results?

Looling forward to your reply

Thanks

[Jon.Wilde]

Hi,

Assuming you are running with a Rotating Region (the best way), no you can suppress it. You can still use the edges of it to add surface refinement to the blade tips too.

Thanks,

Jon

[heath.houghton]

My experience has been that yes you can suppress that solid part from meshing, but the savings in memory are minor and you then have a harder time visualizing the results of pressure on the blades (have to use peel by surface). If you are not really experienced in this type of analysis, I'd leave the blades meshed.

[mycom47]

Thanks for your reply by sharing your experience.

My current results are still deviated from the performance curve of the pump.  Wondering whether you can share your experience to running such analysis with rotating region to achieve a better accuracy.

Thanks

[heath.houghton]

You'd have to upload your support share file to get truly detailed help. You can either do that here (publicly, some companies don't want that) or with our support team. On a more general advice level, I'd make sure your mesh is fine enough on the blades to capture that geometry accurately, use 3 degrees per time step, and make sure it runs enough revolutions to where you don't see the average torque changing. This varies depending on pump, but I'd say 8 to 10 revolutions. To make the process of running pump analyses and extracting results easier, I'd use the turbomachinery automation app on the app exchange store. This really simplifies the results processing portion (taking it to excel and graphing efficiency, pump curves, etc.)

[mycom47]

Thanks for yr pointer

One question that still puzzles me: why 3 degrees per time step (why not 2 deg)? Is it based on experience that, with such time step, it reaches equilibrium for the variables over 8 to 10 revolutions.

Thanks

[Jon.Wilde]

3deg/step is just what we have used and know to work well. You could also use 2deg/step 🙂

[mycom47]

yes.  I have been using either 2 or 3 deg per time step.

Like to ask about the recommended turbulence model to be used? K-epsilon model has been the model.  Is it the recommeneded one or others?

As for the meshing requirement, do you use mesh enhancement (with 1, 3, or more layers).

Thanks for yr advice.

[srhusain]

If you use the k-e model 3 layers work fine most of the time- but you should ensure that wall Y+ values stay above 12 as the k-e model works best for Y+ values in the range of 12 to 300. The SST model is immune to Y+ values falling under 12, so you can use more layers for better boundary layer resolution. This means more elements, but the SST is more stable and frequently reaches converged solution in steady state cases faster than the k-e model does.

Experiment with the mesh and the turbulence models to find your best combination.