CFD Online Discussion Forums

CFD Online Discussion Forums (
-   OpenFOAM Mesh Utilities (
-   -   failed CheckMesh in spite of good quality (

-mAx- May 23, 2013 04:18

failed CheckMesh in spite of good quality
I am used to work and convert mesh from gambit.
Nevertheless, with the use of prismatic layers (BL), and a good mesh quality, I observed that the checkMesh in OF always fails: non -orthogonal faces and over all skew faces.
Regarding skewness I have a max skewness around 0.9 in Gambit, which is ok, but in OF the skewness is around 4.8 (which is not acceptable).
I displayed all the skewed faces from skewFaces Set in Paraview, and they are all quads (which confirms that the problem lies on prismatic layers)
I am aware that skewness calculation's method are different, and I shouldn't compare both quality.
But on Fluent should this mesh pass the check mesh successfully, which is not the case in OF.
Any idea?
Calculation runs and converges well despite checkMesh failed, but... :rolleyes:

Radically method: is there a way to delete all cells, which contains face(s) beloging to nonOrthoFaces and/or skewFaces?

mturcios777 May 23, 2013 13:05

Just because your mesh fails any of the checks in checkMesh, does not mean that it will fail to give you a good solution. A better quality mesh will be more accurate and converge faster but there are algorithms in place to help mitigate the problems (sometime you just can't get rid of mesh issues in certain regions). Also remember that the mesh quality metrics can be changed at any time (in general, the default ones are sufficient for most work).

OpenFOAM is interested in giving you the best possible mesh, especially since we may be dealing with work in academia which requires a bit more rigor. In engineering, sometime we just need it to work well enough.

To actually answer your question instead of waxing poetic, you can remove faces with the removeFaces tool. You specify a faceSet you want to remove and *BAM*, its gone. That said, just removing them is no guarantee you have fixed the problem.

-mAx- May 24, 2013 01:48

Thank you for your reply.
I converged very nicely below 1e-5, but I got bounding epsilon.
And off course the localisation of unrealistic epsilon values are exactly where skew faces are. From my side it means that something is really wrong with those Faces, and I will have to pay more attention to prismatic layers. (even if it doesn't sound to affect my solution)

To the second point, I quickly tried yesterday to digg the question, and I found removeFaces utility (as you mentionned), but after running it a checkMesh failed (something wrong with number of faces within cells). I don't know exactly what does removeFaces, but if it just removes faces, then adjacent cells are open, aren't they?

Then I found on wiki another example with setSet, but after deleting all the cells adjacent to nonOrthoFaces and skewFaces, a checkMesh also failed (something wrong with points)
Anyway, my basic idea was to try ignoring those cells, and as you said *BAM*, if the problem disappears, then I fixed the problem (I am talking about 1000 faces on a 4millions cells nesh)
Thanks again for your interest

All times are GMT -4. The time now is 03:35.