code_saturne User's Forum

Hello Matt,

I checked your mesh using Code_Saturne's mesh quality criteria, and although most of it looks good at first, there are some faces with almost 90% non-orthogonality. Filtering the associated cells under ParaView, they seem linked to degenerate (almost flat) cells.

Removing those degenerate cells from the mesh should make a big difference.

Regards,

Yvan

Thank you Yvan. I reviewed the RESU/check_mesh/check_mesh.log file more carefully. Presumably it has similar information in it as you refer to. Indeed, it does complain a lot, though I must say, the messages are rather cryptic from where I stand. Which Filter in particular are you using in ParaVIS to find these cells? And which files besides the .HDF file do you have to load into it to see them? Can you load the .MED file into ParaVIS directly? So far, I have only ever worked with the .HDF file and then the RESU/yyyymmdd-hhmm/postprocessing/RESULTS.case file.

Also, over the weekend, I made some progress towards implementing in FDS/SMV a much simplified model of what I wanted to do in Salome/Code_Saturne. This program uses a rectilinear mesh to start with. It is fairly easy to set up a case, and it seems to converge reliably and in a snap. I am estimating no more than an hour or two of run time for the full (i.e. not even "symmetric") problem. It has less physical accuracy than Salome/Code_Saturne (simply because I am not modelling curvatures, and much has to fall into a rectilinear "straight jacket"), but it seems to have a lot more numerical accuracy - and more importantly stability - than the latter. Based on what you said, I suspect that is because it starts out with known good (perfectly square) cells in the mesh.

Finally, if it wasn't so difficult to set up a case in OpenFOAM, I would want to try that, but I suspect it too would fail, given that the mesh is bad to start with. So, I guess I will spend a little more time trying to fix the mesh (if I can figure out how to identify the bad cells), but I really cannot afford to put too much more time into it, especially since FDS/SMV seems to offer a workable alternative now.

Thanks,

Matt

Hello,

For postprocessing, you can use the "check mesh" option in Code_Saturne, or use the "mesh quality" run type, then visualize cell quality using the standard postprocessing output.

Under PARAVIS (SALOME's ParaView + plugins), you can visualize the MED file directly, and use ParaView's mesh quality criteria also (different, but should help you find degenerate cells).

Regards,

Yvan