[kandelabr]

Hello.

Please see the attached image. It is a single channel of a centrifugal impeller, simplified to have infinitely thin blades. There are cyclic boundary conditions on wedge sides on inlet (two blue blocks) and outlet (red block) and MRF zone between inlet and outlet (yellow block).

I (again) have huge problems with mesh generation:

blockMesh:

• That would be the preferred and fastest way and I could create blockMeshDict from a script
• could use a simple cyclic condition on appropriate faces
• I can't create the yellow block because it would contain concave cells
• Can't divide the difficult block in any reasonable way, especially from a script.

snappyHexMesh / cfMesh

• It takes ages for both meshers to make a good mesh
• cannot control cyclic faces and have to resort to cyclicAMI
• cyclicAMI fails when creating thin boundary layers because meshes are usually not conforming to exact geometry on both sides so not all cells have their neighbour cell

SALOME

• a ton of wrestling with scripting, geometry creation, patch assignment etc.
• netGen and similar conform to geometry perfectly (although slowly)
• only tetrahedral meshes
• no boundary layers (or very unreliable)

It is important to me that mesh creation is as fast as possible but even more that it is reliable. I'm trying to run a large number of similar cases with geometry generated by optimization algorithm from an external script.

If anyone has any experience or ideas with this kind of problems I'll be very happy to listen

Thank you!