I've been unable to get the above simulation to converge using CFX 11. Fluid model chosen was SST. the 0.2mm front and back leakage gap domains were meshed using the extruded 2D mesh with 5 layers while the other domains have tetrahedral mesh. Frozen rotor was specified for inlet-impeller interface and stage for impeller-volute interface. checking the results after 900 iterations gives an overprediction of the leakage flow of about 2.7 lpm compared to theoretical value of under 2 lpm. any suggestions on how i should go about modelling this case?

Hi,

If the simulation does not converge then the results are rubbish so the flowrate comparison is meaningless. Also make sure you are modelling the clearance which actually occurs when the impeller is spinning. Check the clearance is not taken up in part by expansion of the impeller due to centrifugal or fluid pressure forces.

Mesh quality is always an issue for simulations like this, make sure you have the best quality mesh you can do.

This link gives some pointers to help with convergence: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

Glenn Horrocks

Thanks for the reply.

I've managed to get my simulation to converge by increasing the physical timescale and refining the mesh but I'm getting a leakage flowrate that's higher than the theoretical value .

SST requires a y+ value of less than 2 but my model has a maximum y+ of 37 and is larger than 2 in regions of the volute and impeller according to cfx-post. I'm not able to further refine my mesh as the computer starts crashing (out of memory error) above 2 million elements for this simulation. I'm now switching to k-epsilon but i'd still like to return to SST. Is there a way i can do this?

Hi,

Get more memory or parallel licenses. Sounds like you need a bigger mesh.

Glenn Horrocks