[obscureed]

Following on from a previous post (Removing Zero-sized Patches ), but with a new title etc.

We are still seeing zero-sized faces generated by snappyHexMesh when using version 2306 of ESI's OpenFOAM suite.
- These are detected in meshCheck.
- The collapseEdges utility sometimes removes them (and then we can proceed), but not always.
- They seem to cause fatal trouble in using the mesh (either for solving or createBaffles etc).

I cannot currently share any cases -- they are large and confidential. From a small sample, it seems that that the trouble occurs where there are internal surfaces (intended to be internal baffle walls, or surfaces for flow monitoring) sharing edges with exterior surfaces.

Can anyone give me suggestions for how to tackle this, please?

In snappyHexMesh:
- I've compared the output from snappyHexMesh in version 10 of OpenFOAM Foundation's suite. The sequence of steps appears to be very similar, with slightly different cell counts, etc, as you might expect. The mesh from the v10 snappyHexMesh does not have zero-sized faces (including if I apply the v2306 checkMesh to it). This workflow has been consistent and reliable in v10.
- The v10 output mentions a warning "Did not find surface near face centre" many times. The v2306 output does not contain these warnings. (I would not object to suppressing these warnings, because they occupy nearly 100,000 lines of log file, even when the result is a good mesh.)
- The v10 output mentions a step at the end, "Removing zero-sized patches", not reported in v2306. I tried the advice of running createPatch with a blank createPatchDict to remove zero-sized patches -- it did not work, but I have not pursued it very far. I do not believe that zero-sized patches can be the source of the trouble, because I think they will be ignored by subsequent utilities.
- The v2306 output mentions an extra step, "Refinement transition refinement", not reported in the v10 output. (As far as I can, I'm using equivalent versions of snappyHexMeshDict, but this is not straightforward because the OpenFOAM Foundation's syntax keeps changing.) This is somewhere in the middle of the sequence, after "Dangling coarse cells refinement".

In collapseEdges:
- It is strange that there are zero-sized faces detected in checkMesh, but collapseEdges sometimes does not remove them.
- I have been using the collapseDict from /usr/lib/openfoam/openfoam2306/etc/caseDicts/annotated/collapseDict.
- I have tried changing to "controlMeshQuality off;" in ./system/collapseDict.
- Are there other settings I should try?

I have gone looking in ./constant/polyMesh/points.gz, and there are vertices with identical coordinates. (This is not necessarily wrong -- they could be duplicated to avoid non-manifold edges -- but it does suggest that the zero-sized faces could be real.) There are no duplicate vertices in the v10 points.gz.


Any help much appreciated.
Ed

[obscureed]

Actually, I've just looked at the collapseDict settings again, and changed the minimumEdgeLength setting from 1e-6 to 1e-3 (which is still small in my geometry).
- Now, even with controlMeshQuality back on, it has zapped the zero-sized faces in a case where it previously didn't.
- The mesh now passes meshCheck apart from a few cells with skewness up to 8, and these were present anyway. (The OF10 mesh had skewness up to 5.6, max non-orthogonality 65.)
- The max non-orthogonality has increased from 65 to 68 -- a fairly minor change, though in the wrong direction.

So this now looks like a fix for that case, and I'll try running it. In general, though, I would like to know why this is happening in v2306 but not in v10. I would like to avoid zero-sized faces rather than using collapseEdges to fix them at a cost to mesh quality. Any ideas?