CDT for 3D surface meshes intersection

[VinceChiarut]

A little message to thank the CDT developers. I am currently developing a mesh backend for a FE solver using intel TBB and I needed an opensource constrained Delaunay kernel for surface meshing. I was very pleased with the easy interface offered by CDT and its very good performance for my usage. I have developed a triangle-surface mesh intersection process for now. This approach is performed in two stages:

1. A geometric triangle/triangle intersection stage with multi-threading and BVH, delivering a thread-safe conformal segments/vertices data structure.
2. The embarrassingly parallel conforming remeshing of each of the intersected triangles using CDT.

Here is an example of an academic application: 6859 intersecting spheres with 8.7 Mtriangles. The conform intersection (17.2 Mtriangles, 110 for the worst triangle quality) was obtained in 11.7s with 6 threads and 36.6s with 1 thread on a 6 core laptop (21GB of memory required during the operations).

Once again, I extend my deepest appreciation to the creators of CDT for their excellent open source contribution.

[artem-ogre]

Hi @VinceChiarut,

Thank you very much for the kind words! This made my day :) It was also very interesting to know about your CDT usage.

Do you also use robust predicates for detecting or resolving triangle-triangle intersections?

[VinceChiarut]

Hi @artem-ogre,
For my application I use a global tolerance to identify singular case (edge/edge, edge/vertex or vertex/plane) intersections in the geometric part. It's not perfect, but I works pretty well.