mesh_tri.isValid()

[JarosT55]

Hi,
I accoutered the problem in the following sequence of the code:

$meshT = MeshTri().refined(2)$
$isValid = meshT.is⎽valid()$

produces: $\quad\quad\quad isValid = False$

The problem stems from the comparison inside $is⎽valid()$ .
$mesh.doflocs.shape[0] != mesh.elem.refdom.dim()\quad\quad\quad$ # 2 != 3

The similar code for $MeshQuad()$ does not exhibit any issues.

I looked inside .skfem\refdom.py, but the finesse of what was going on there is beyond my rusty comprehension.

Any ideas ...?

Regards
MichaelT

[kinnala]

I am unable to reproduce the issue:

In [1]: from skfem import *

In [2]: MeshTri().refined(2).is_valid()
Out[2]: True

I am using the latest main branch.

[JarosT55]

Hi Mr Kinnala,

Thank you for checking out.
I updated to the latest version of skfem
Successfully uninstalled scikit-fem-8.1.0
Successfully installed scikit-fem-9.0.1
and I also re-checked in the minimalist code in the separate Jupyter notebook,
The result is positive—no validity problem … for vanilla-type mesh.
The problem existed in the original code and only when it was run for the second and successive times. Import/dealings with 3D surface triangular meshes triggered it.

As I understand, skfem does not support such 3D surface meshes and respective capable elements.
Can they be ‘simulated’ by ‘flattened’ ElementTet()s? (rather crazy idea!)

Regards
MichaelT

[kinnala]

Yes they can, it is not a crazy idea, I have used it a few times. It is currently the only way to use such surface meshes in scikit-fem. I have tried to support them natively once or twice but they seem to require so many changes here and there that I have given up (due to lack of motivation - I don't study such problems in my own research).

[kinnala]

If you read a surface mesh, you should add to each element a fourth node to make it a tetrahedron. Then in the final matrices you should slice only the nodes belonging to the surface. That way you can use scikit-fem for surface meshes.

[JarosT55]

Hi Mr Kinnala,

... I was thinking along this line but in the opposite direction, ... by using a sledgehammer to flatten tetrahedra :).
Although possible, full implementation of a 3D surface in skfem would be pretty challenging. Nevertheless, the package's fabric and its good structure seems suitable for such a task.
I think I will try finding other 'countermeasures' to deal with 3D surface meshes, … anyway, only for visualization purposes.
Thank you for the input.

Regards
MichaelT

[JarosT55]

Hi Mr Kinnala,

I have dealt … a little with '3D surface' issue. Despite some progress, I encountered a 'topology' problem.
Perhaps, it deserves, in its own right, the separate discussion thread.
I will start the new one under the title 'Mesh and Normals' ⏳.

Regards
MichaelT