Extract high-quality 3D mesh from level set

[jmark]

I have the following use case:

I do process simulations of MEMS devices. From that I get level sets representing the etched structure of the MEMS. The next step is to transfer the level set to a 3D volume mesh after clipping the mask and the underlying substrate. The 3D mesh is supposed to be an input for FEM simulations (mechanical stress, temperature gradients, conductivity, etc.) done by, for example, ANSYS. Thus, the 3D mesh must be of decent quality: water-tight, no degenerate or tiny elements.

The 3D mesh does not have to be perfectly aligned with the level set. Minor deviations are acceptable for now. Mesh quality is more important.

I achieved the workflow with Paraview reading the VTU files written by ViennaPS. But the resulting mesh quality is very bad.

My question: Is there a robust way inside of the ViennaTools framework (or outside) to do this? Do you have similiar use cases? If so, what is your approach?

[tobre1]

At the moment, ViennaTools/ViennaPS does not provide a robust “FEM-grade” 3D meshing pipeline from the level-set geometry.

What you see in ParaView is expected: the current VTU output is mainly meant for visualization and post-processing, not for producing high-quality, watertight, well-shaped tetra meshes. In particular, we do not currently run a dedicated volumetric mesher (e.g., constrained Delaunay / quality tetrahedralization + cleanup), and generating a reliable Delaunay-quality mesh from complex clipped geometries is a non-trivial problem.

We are actively looking into better mesh generation strategies for FEM follow-up simulations, but this is not available yet. If you find a reliable external pipeline (especially one that works well for clipped mask + substrate cases), please share it. This is a use case we also care about and would like to support better.