[psceco]

Hello everyone,


I'm new to OpenFoam but there are still things I don't understand.


I'm working on a 2D simplefoam case with Re = 10^5 and k-epsilon model. The boundaries are in freestream and the obstacle is a wall.


With gmsh, I made a size field to create a boundary layer around the obstacle, and checkMesh says the mesh is OK.


I don't understand why, but the simulation diverges from a random spot near the wall in the boundary layer set in the mesh.


Apparently, the problem doesn't come from the conversion of the mesh file (with gmshToFoam) so it must be the simulation.


That's where I need your help : does it come from the boundary conditions, from the simulation settings (I use the basic simplefoam k-epsilon model) or something I don't suspect ?


Please find attached the 0 archive and the .geo file (create mesh with "gmsh retro4P1.geo -3 -o retro4P1.msh" then "gmshToFoam retro4P1.msh" commands). Thanks for your help !

[HPE]

Out of my head: kEpsilon might be unstable due to epsilon. Have a go for kOmegaSST.

[psceco]

Hi HPE,


Thanks for your reply. I apologize for answering a bit late, I was testing the solution you gave me.


It appears the k omega SST is a bit more stable than k epsilon, because I managed to solve some of my cases.


Still, there are a few cases that keep diverging after a few iterations, still for the same reason : a spot near the boundary layer. I don't understand why, because the inputs are the same that the cases that went well. Maybe it comes from the mesh. Could it be that it is too complex and the solver can't follow ?


For instance, the mesh I use is too heavy to be uploaded to this message, so I must give you the gmsh .geo file to be converted to OF.


To create the case, you must create the mesh with gmsh by bashing the "commands" file I put at the root of the case and then replace the boundaries file in polymesh by the one at the root of the case.


Any opinion or help would be greatly appreciated !
psceco