[kandelabr]

I have this case of a centrifugal pump.

The mesh is divided into three parts, suction pipe, impeller and volute casing. I spent a good month meshing the impeller and volute with all possible refinements, boundary layers, hexahedral and tetrahedral cells etc. What I have now, I believe, is a pretty decent mesh.

The three parts of the mesh are coupled with two AMIs with weights ranging from 0.7 to 1.1, which I think is OK. With bad meshes I either get problems with AMI or points with extreme pressure and velocity values. Impeller is a cellZone within MRF or dynamicMesh (two different cases).

My boundary conditions are as follows:
Walls: noSlip for velocity, zeroGradient for pressure
Inlet: flowRateInletVelocity for velocity, zeroGradient for pressure
Outlet: zeroGradient for velocity, fixedValue for pressure
At the moment I don't bother with turbulence modelling at all.

I tried three different approaches, first with no MRF/moving mesh. That works but obviously doesn't give any useful results except that it shows possible mesh problems. There aren't any and the solution converges.

MRF and dynamicMesh both diverge. The pressure keeps getting higher and higher on the inlet, velocities as well, but smoothly (not like with a bad mesh). The values are way too high and I know what the results should be because I have results from a real-life test of the exact pump.

I checked everything a hundred times over and at the moment I have no idea what else I could check.
I'd appreciate if someone took a look and maybe give me an idea what to work on next... I attached all case files but the mesh (there are roughly 500 million reasons for that )

Thanks!

[huangxianbei]

Apart from the settings, a important fact is that the mesh you use will significantly affect the numerical stability. My advice is to generate a corser mesh or a mesh with higher quality. To my experience, it would work fine.

[kandelabr]

so you are saying that my settings are fine and i should rather work on my mesh?

[Alisa_W]

Have you solved your problem? I am simulating a pump as well, but I can't even make it solve smoothly.

[kandelabr]

I made the nicest possible mesh using Salome with Netgen and viscous layers. There are no more high- or low-pressure points in results but pressure is much higher than physically possible although smooth.

i'm still juggling with boundary conditions. i'm just trying different combinations because that's all that i can do at the moment. i can't find anything useful in the internetz.

[Alisa_W]

I have set a fixed inlet velocity and a fixed outlet pressure with other boundaries set as zeroGradient for both p U.However, the solving process always diverged.

[kandelabr]

I'm noticing a strange thing.
Obviously pressure at the inlet pipe of the pump should be lower than that on the outlet but every other MRF iteration it 'changes sides' - I guess the system with my boundary conditions has two solutions.

But none of the solutions are correct because pressure is too low and too high by roughly three orders of magnitude. The pump head should be around 4.5m (that's 0.45 bar or 45 p/rho) but I consistently get numbers above 14e+3...

[kandelabr]

It does seem to be a mesh problem after all.
If I turn MRF off and do a very irrelevant case of flow through a fixed geometry with boundary conditions from the pitz-daily case, the results are still wrong.

I know k-epsilon and other turbulence models pose special requirements on mesh near walls but I'm currently trying to do a simulation without turbulence modelling. Are there any other guidelines for meshing in general?