[dachs]

Hello everyone,

I am simulating oil/air in a rolling bearing and have the problem that numerical instabilities form around my interfaces. Does anyone have an idea why this could be? I use an MRF zone to move the oil through the bearing.

[Alczem]

Hey,


Do you have a turbulence model? Is your mesh fine enough in gaps between the rollers and the walls?


I am not sure interFoam is suited for your case depending on the mesh resolution. It might have trouble capturing the interface. Maybe a Euler-Euler solver is more appropriate.


Share your case if you can, it will be helpful for others.

[dachs]

Originally Posted by Alczem

Hey,


Do you have a turbulence model? Is your mesh fine enough in gaps between the rollers and the walls?


I am not sure interFoam is suited for your case depending on the mesh resolution. It might have trouble capturing the interface. Maybe a Euler-Euler solver is more appropriate.


Share your case if you can, it will be helpful for others.

Hi,

thanks for your help. Unfortunately my file is too big to upload directly here. I was able to upload the file to Gigamove via my university, if you want you can download the file here and have a look:

https://gigamove.rwth-aachen.de/de/d...fefcfadd3bf57a


I also tried to start the simulation with multiphaseEuler, but I always get an error message because of my force functions in my controlDict. Apparently I have to specify a coordinateRotaion, but it is always defined incorrectly.


I would be very grateful for any further feedback.

[Alczem]

Hey,


As usual, I would suggest to create a new, smaller case based on the real one, and turn on the needed features one by one (laminar, then turbulent, no MRF then MRF etc). For instance, start with just one cylinder, make sure the mesh is OK and fine enough, do not add inflation layers right off the bat, and use 1st order for all variables. If that works, progressively add more complexity to the case.


It might be a lot of effort, but it is the most reliable way to pinpoint the issue.




EDIT:


I checked your case, and I am not sure about the boundary conditions and MRF properties. You specify a lot of MRF zones, and from what I understand, there might be overlaps. Not sure how OF v11 handles this but try to run your case without any MRFs, and turn them on one by one. Why are you using MRF by the way? I would expect a boundary condition on the rotating wall to take care of the rotation, not a MRF.

[dachs]

Hi,

I use MRF zones to give my liquid a rotational movement. In a rolling bearing, the rolling elements rotate once around themselves and translate through the bearing. Since I don't know how to represent this translational motion, I have chosen a relative approach in which I have to give the fluid a velocity around the shaft's axis of rotation. I do this via the MRF zone ‘FreeFlow’. Currently, only this MRF zone is active, but it is already causing the problems. As I initially had continuity problems, the MRF zones have proved useful for each rotation (I now believe that this was due to my poor mesh quality at the time). To avoid overlapping, I define FreeFlow last in topoSet and then delete all cells from the other zones here. Is this probably not the most elegant way or what do you think?

I have tried multiphase Euler (without force calculation) and also get instability here. So I think there is still a problem somewhere with my BC or my mesh. This is also my first attempt with a tetra mesh instead of a hexa mesh.

A question of understanding: Why should I not start with inflation layers? Isn't the mesh always better with these layers than without?

I have also made smaller cases where, for example, only one rolling element rotates in an oil bath or the fluid rotates in a ring segment around the shaft rotation axis. These simplified cases all work very well, but the combination turns out to be complicated for me

Thanks for your input and your time!

[Alczem]

Hey,


Concerning MRF, I would create a single cellzone containing the outer ring and the cylinders. Check out the attached picture: in red, you have the MRF and in blue the fixed cells. I don't think you need more to represent the moving cylinders, except a rotating boundary condition. I hope it is clear enough I haven't had the time to run your case sorry


EDIT:


Hexa mesh is usually the better choice over tetra for OpenFOAM, try to stick with this if your geometry allows it. Despite its flaws, snappyHexMesh gives a decent mesh if you allow it to refine at sufficient levels.


Boundary layers are nice to look at, but they often worsen the mesh quality. Starting with a coarser mesh stabilizes the simulation by introducing more diffusion. It might be less accurate, but it is more robust. Make sure to have enough cells per gap though (bare minimum is 3 IMO, and ideally at least 5). Once your case is running, you can try to refine the mesh.

[dachs]

Originally Posted by Alczem

Hey,


Concerning MRF, I would create a single cellzone containing the outer ring and the cylinders. Check out the attached picture: in red, you have the MRF and in blue the fixed cells. I don't think you need more to represent the moving cylinders, except a rotating boundary condition. I hope it is clear enough I haven't had the time to run your case sorry


EDIT:


Hexa mesh is usually the better choice over tetra for OpenFOAM, try to stick with this if your geometry allows it. Despite its flaws, snappyHexMesh gives a decent mesh if you allow it to refine at sufficient levels.


Boundary layers are nice to look at, but they often worsen the mesh quality. Starting with a coarser mesh stabilizes the simulation by introducing more diffusion. It might be less accurate, but it is more robust. Make sure to have enough cells per gap though (bare minimum is 3 IMO, and ideally at least 5). Once your case is running, you can try to refine the mesh.

Hi,

thanks for your advice, I think it was spot on and I'm now on the right track (although I don't want to celebrate too soon).

I have replaced my mesh with a hex mesh without BL, and only one MRF zone. Although I find snappyHexMesh extremely frustrating to use and it takes an extremely long time, my simulation has now been running for a day without showing any instability. So thank you very much for your help!