ERROR:: ElementMetric: Degenerate 3D element

[mzenker]

Hi,

I would like to reactivate this thread since I have got the error again. I have built a testcase which is attached, and I would appreciate someone with more knowledge than I have to have a look at it. Since the mesh is 25 MB large, I do not post it here, but I post the geo file (gmsh format) which permits to reproduce it (I have used gmsh 2.6.2 svn). Just load the geo, mesh in 3D and save. I have tried to import the mesh into Elmer with and without -autoclean, with ElmerGUI and with ElmerGrid. The questions are still the same as before: what is going wrong here, and what can I do to avoid it?

Thanks for any help,

Matthias

EDIT: I forgot to upload the files, done now.

[raback]

Hi Matthias

If it works now then no need to sent. Usually the degenerate elements are due to problems in initial mesh generation or in deforming mesh cases some excessive deformation. The first might not have anything to do with Elmer and the second one is more a problem in setup than a bug. However, there could sometimes be a problem of mesh import as well, and then it is clearly something that should be resolved. Or some mesh deformation routines could be broken. But if you don't think this is the case then maybe we can forget it.

-Peter

[mzenker]

Hi Peter,

thank you for your reply.
Maybe I wasn't clear enough: This is a new case of a 3D mesh giving the error message. So as of now, it doesn't work unfortunately...
I don't have a deforming mesh so this can be excluded for this case.
Questions: What does Elmer consider a degenerate 3D element? Are there limits/tolerances to observe? What should I try to check in order to detect problems before ElmerSolver throws the error message? Can I inspect the mesh from within ElmerSolver/Grid?

Thanks,

Matthias

[mzenker]

Addendum:
I can save the mesh as vtk from gmsh and load it in ParaView, which has a "mesh quality" filter. But I can choose between many quality measures for tetrahedra:
* edge ratio
* aspect beta
* aspect gamma
* aspect frobenius
* aspect ratio
* collapse ratio
* condition
* distortion
* jacobian
* minimum dihedral angle
* radius ratio
* relative size squared
* scaled jacobian
* shape
* shape and size
* volume

I have to consult the documentation for the exact definition of those.
What are Elmer's criteria for mesh quality?

Matthias

[raback]

Hi Matthias

In Elmer the determinant of the Jacobian must be larger than epsilon. So I would look at the "jacobian" or "volume". In practice this means that the volume (area in 2D) goes to zero. This can happen when two nodes are two close to each other. So it is therefore also related to aspect ratio.

-Peter

[mzenker]

Thank you Peter, I will see what I find out and report back...

Matthias

[mzenker]

Hi,

the problem is (hopefully) solved!
In the 3D mesh generated by gmsh, there were tetrahedra with all 4 points almost aligned. In the case I have examined more closely (thanks to the ability of Elmergrid to write out gmsh format), the deviation from exact alignment was smaller than 5e-13. Apparently Elmersolver considers such elements degenerate. It seems that this kind of elements is called "sliver".
The gmsh developpers pointed me to the "optimize" function for 3D meshes which (hopefully) eliminates degenerate elements. It made the error messages disappear in my case.
So for all gmsh users: always do "Optimize 3D" and/or "Optimize 3D (Netgen)" on the 3D mesh before saving it!

Matthias