[Munki]

Hello everyone,
I'm trying to set up a transient simulation of a mixing vessel with pimpleDyMFoam (Openfoam 4.1). A sketch of the setup is shown in figure 1. The vessel is filled with a liquid the rotating speed is roughly 30 rpm. Laminar flow is to be expected.
Currently all my attempts lead to unphysical velocity/pressure distributions in certain areas close to the impeller (see figure 4). Based on the impeller dimension and rotational speed the expected velocity should lie in the order of <1 m/s. The simulation results in peak velocities up to ~200 m/s.
The case-file can be downloaded here:
https://www.dropbox.com/s/psjr8p9a0v...el.tar.gz?dl=0

My setup is as follows:
- the mesh was built with Star-CCM's trimmer model (comparable to snappyHexMesh, see figure 2)
- mesh conversation was done with wyldckat's ccm26ToFoam (see https://github.com/wyldckat/localCCM26ToFOAM)
- since the geometry is pretty simple and doesn't contain any baffles, etc. no interfaces are needed and the whole mesh is put into one cellZone (topoSet) --> this means the whole mesh rotates during the simulation
- laminar flow --> 0 directory only contains p and u
- boundary conditions:
u: top --> slip wall --> zeroGradient
stirrer --> no-slip wall --> movingWallVelocity (0 0 0)
tank_wall --> no-slip wall --> fixedValue (0 0 0)
p: top --> slip wall --> zeroGradient
stirrer --> no-slip wall --> zeroGradient
tank_wall --> no-slip wall --> zeroGradient
- fvSchemes & fvSolution based on tutorial test cases



Things I've noticed:
- despite the high velocities the simulation doesn't actually crash
- after some iterations with reasonable velocity/pressure distributions, said fields blow up
- so far I ran the simulation for half an hour and the velocities didn't really change for the better

Things I've tried to fix the problem:
- mesh optimisation --> checkMesh still complaints about concave cells, but - as far as I understand - this can be explained with the splitHex meshing approach
- Originally high velocities occurred in the transition region between shaft and anchor impeller (see figure 3). Since the mesh quality in this region was at least questionable, I decided to remove the shaft from the model. But still the unphysical velocity distribution didn't vanish. It just moved from said position to the outer region of the anchor (see figure 4).
- played with fvSchemes & fvSolutions settings



So guys, what do you think?
Any help is greatly appreciated.


PS: I'm aware of the fact that you can probably simulate given problem with a reference frame approach (e.g. SRFSimpleFoam), but still I want to perform a transient simulation with pimpleDyMFoam

[Munki]

*additional screenshots

[BlnPhoenix]

Hello,

from the looks, i think you have a mesh issue. First i would try to get a lower growth ratio for the cells in the mixer (You have very large cells in the middle, and very fine cells in the boundary layer, to capture the gradients) try to get the cells more evenly sized. Solvers don't like sudden increases in cell size.

Secondly i don't understand your problem setup. Is your impeller filled with water?

[Munki]

Hi BlnPhoenix,
thank you for your reply. Based on your advice I spend the last days "optimizing" my mesh (see attached screenshots). Although I'm not happy with the mesh*, at least the simulation does create meaningful results now
Concerning your question about the impeller: No, the impeller is not filled with fluid. It's just a solid rotating in the vessel.

The case-file can be downloaded here:
https://www.dropbox.com/s/ijgweshnd8...v2.tar.gz?dl=0



*Basically, I switched to polyhedron cells, removed the prism layer at the boundaries and tried to reduce the expansion ratio between cells as good as possible.

[Munki]

Hi, it's me again
I used the last few days to come back to the problem mentioned above. Although I could find a "solution" (thanks to the help of BlnPhoenix) it never felt right to blame the mesh for my results.
To rule out the possibility that the mesh might be the reason for the strange flow field, I decided to create a simplified 2d version of my case while maintaining the same boundary conditions and solver settings (see figure 1).
... as expected, the flow field is at least questionable. With a rotational speed of 3.14 rad/sec and a impeller diameter of 0.085 mm one would assume velocities in the order of <1. The solver calculates velocities up to 40 m/s.

So what's the problem with my simulation? My current thoughts are:

• I've done similar simulations with commercial solvers and everything went fine.
• my case (solver settings, boundary conditions) is based on "../openfoam4/tutorials/incompressible/pimpleDyMFoam/mixerVesselAMI2D/"
• the only difference between my case and the tutorial case is the lack of baffles at the tank wall --> that's why i can (at least i think so) put everything in one rotating domain and do not have to use interfaces between the stationary and the rotating parts
• ... and this might be the problem: does anyone know if pimpleDyMFoam allows simulations in a single rotating domain with counter-rotating walls? And if so what are the correct boundary conditions for said problem?

The case-file is attached below. Any help is greatly appreciated.

case-file: https://www.dropbox.com/s/zap0811tfd...2D.tar.gz?dl=0