[Johan M]

Hi Everyone,

I have encountered divergence in CFX pump simulation involving a volute and impeller. Both geometries were automatically generated within Ansys Vista Centrifugal Pump (CPD).

Problem area:
The entrance of the volute consists of an InletHub, InletShroud and Inlet. InletHub and InletShroud are the lower and upper walled regions of the volute seen selected in image 0. Inlet is the fluid entrace of the volute, which would be interfaced to the exit of the impeller with a mixing plane interface.

The volute was meshing using Ansys meshing. The scoping method using automatic meshing did not include regions marked in image 1 which are the InletHub and InletShroud walls.

Without inflation layers on them, the solution converges. However when I manually insert inflation layers, the solution diverges with spiked residuals and unrealistic total outlet pressure monitors seen in image 1,2. The mesh quality in both cases are within the failure criteria prescribed in the manual.

Current workaround:
Since the addition of inflation layers on the wall was the issue, i adjusted the volute geometry until it only included the Inlet at the entrance - the InletHub and InletShroud were reduced to zero.

I am also doing a simulation where I mesh the volute with Fluent Meshing with Hexcore cells to check if I achieve a better solution. This sim is with the InletHub, InletShroud and Inlet configuration.

I would like to know what might cause the insertion of inflation layers to result in such a drastic divergence. Additionally, I would appreciate any other advice on a better workaround that doesn't involve such a drastic geometry change in the volute.

Kind regards,
Johan

[Opaque]

Mesh affects accuracy; however, a finer mesh may require a smaller timescale for the model to converge.

If a coarse mesh model, and coarse does not mean small mesh count, converges for a given timescale, it does not mean the finer mesh will converge for the same timescale.

Have you tried reducing your timescale?

The above assumes your inflation layer is "correct". That is, there is nothing wrong with the mesh.

[ghorrocks]

I would check the points covered in this FAQ before proceeding: https://www.cfd-online.com/Wiki/Ansy...gence_criteria

[zacko]

Can you share an image of your mesh?

[Johan M]

Originally Posted by Opaque

Mesh affects accuracy; however, a finer mesh may require a smaller timescale for the model to converge.

If a coarse mesh model, and coarse does not mean small mesh count, converges for a given timescale, it does not mean the finer mesh will converge for the same timescale.

Have you tried reducing your timescale?

The above assumes your inflation layer is "correct". That is, there is nothing wrong with the mesh.

Thank you Opaque for the reply. I reduced the time scale and my simulation is now running

[Johan M]

Originally Posted by ghorrocks

I would check the points covered in this FAQ before proceeding: https://www.cfd-online.com/Wiki/Ansy...gence_criteria

Thanks ghorrocks for the suggestion

[Johan M]

Originally Posted by zacko

Can you share an image of your mesh?

Hi Zacko, I shall do so when I have access to that pc again