[chriss85]

I've recently run into problems where snappyHexMesh will generate polyhedra cells with a nonzero volume, but two faces directly on top of each other.

This results in problems because the Courant number will be extremely high in these cells, and adaptive time step methods will fail.

This problem is not detected by checkMesh, since common quality measures like non-orthogonality or cell volume are unaffected. I think this issue needs to be considered in the detection of invalid cells during undo stage in snappyHexMesh.

For reference, here's the checkMesh output:

    Overall domain bounding box (-2.10476e-08 1.47089e-05 1.07802e-05) (0.017315 0.0186159 0.0183828)
    Mesh (non-empty, non-wedge) directions (1 1 1)
    Mesh (non-empty) directions (1 1 1)
    Boundary openness (-8.91807e-17 -2.58947e-17 -6.77004e-18) OK.
    Max cell openness = 3.89026e-16 OK.
    Max aspect ratio = 5.73767 OK.
    Minimum face area = 4.2135e-10. Maximum face area = 4.23884e-08.  Face area magnitudes OK.
    Min volume = 4.94644e-14. Max volume = 7.47925e-12.  Total volume = 3.39481e-07.  Cell volumes OK.
    Mesh non-orthogonality Max: 55.4823 average: 10.0927
    Non-orthogonality check OK.
    Face pyramids OK.
    Max skewness = 1.68602 OK.
    Coupled point location match (average 0) OK.

If any further information is required, feel free to ask me.