CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   about MRFSimpleFoam and ggi (http://www.cfd-online.com/Forums/openfoam-solving/84662-about-mrfsimplefoam-ggi.html)

wllmk1 February 5, 2011 04:52

about MRFSimpleFoam and ggi
 
Hello.

I have previously posted pretty much these same questions to another thread but I have been unable to receive any answers for a while so I'll try my luck here. To get familiar with my case you can check:
http://www.cfd-online.com/Forums/openfoam/83498-rotating-rotor-inside-frame-2.html

Shortly: Im trying to simulate an airflow inside an electric motor affected by rotor with 1000-3000 rpm (1.5-dev). My case is basicly two concentric cylinders with some holes and blades. I have been trying to solve this with turbDyFoam and MRFSimpleFoam, both cases with ggi-interface. But for now I'm just desperately trying to get a steady state solution with MRFSimpleFoam.

I have builded two element models with Femap v10.2. First is a full 360 degree model with pretty coarse mesh ( about 1 000 000 cells ). Mesh is considered coarse because airgap between rotor and stator is only 2 mm when diameter of rotor is 400 mm. This makes elements in the air gap pretty bad quality. Another model is a 22,5 degree model with a symmetric BC (cyclicGgi). This model has about 1,2 M cells because I have thickened the mesh towards the air gap. So mesh is coarse elsewhere but steadily thickenes when heading towards the air gap.

Mesh is created likes this: ideasUnvToFoam --> autoPatch --> createPatch.
So I have defined boundaries manually because mesh generation crashes if I try to include any boundaries in ideasunvToFoam conversion. Values and settings for BC, k, U, epsilon, controlDict, FvSolution etc.. have been made similar with many examples available of MRFSimpleFoam and ggi.

I'm having problems with both (cyclic and full) cases and these problems are actually pretty similar. Basicly after 10 ... 60 time steps the values of epsilon and k start to rise unsteadily till I get a floating point error and U reaches infinite. I have been able to avoid this problem by reducing relaxation factors to 0.01 or 0.001 but then it seems that solution never converges. I have also tried to start with smaller rotational speed which seems to help a little but still crashes after some time. When I view solved steps (with small relax. factors) in paraview the values and direction of U vectors seem to be in right scale however.. When I view values of epsilon and k there seems to be only a one vector with very high value..

Could this be a problem related to ggi interfaces? faces in insideSlider and outsideSlider are both planar and cylindrical if you know what I mean. So ggi interface is like a barrel with covers. Should I define ggi separately for planar and cylindrical faces?

Could this be a problem related to MRFZones? I should have a cellZone with all the cells in rotating zone right? And MRFSimpleFoam can define faceZones itself right? What patches should I have for patches in MRFZone file? I assume that wall of rotor? Howabout cyclicGgi's ? Howabout nonrotatingPatches? normal ggi interfaces and wall of stator?

Could this be a problem related to something else? I have been searching this info myself but it seems that most of available info is out of date or very contradictory.

Im sorry that I cannot share you any kind of pics or files of my mesh because of copyrights. I might be able to share all the other files however if that is needed.

I really appreciate your help for any of these questions

-wllmk1


All times are GMT -4. The time now is 15:36.