I have a mesh file (describing a dense volume in space with tag specifying the martial of each tetrahedron for a FEM simulation. However, this mesh is too detailed for my simulation. I would want to reduce the total elements in it (thus the total DoF to solve the simulation), with the best effort to respect the material boundaries in space and varying mesh size based on knowledge. I don't have the cad file for it.
I encoded the geometry into Gmsh using python interface and using compound to combine smaller triangles I wanted to merge, it worked fine for a small simple. but when I tried to define a compound made up to ~200k triangles, the RAM usage goes up to 32 GB (all my PC have) and crashed.
so here are my questions.
- is this an expected behavior as the compound processing somehow cost a lot with large group?
- this there some walk around inside Gmsh, I am thinking about divide the grouping surface into smaller groups, but the next question goes to me is that how to divide them. Intuitively, for a fixed group size, a long and narrow group is not supposed to get optimized as good as a round one.
- if Gmsh is not good at dealing with this, is there some way using other tools, searching CGAL, I found this , but my problem goes differently at there are different martials and surfaces, if I process each volume separately, I'm afraid I can't get a merged dense msh where all tetrahedron meets and share the exact same faces, I think it's possible to use the algo indicated and handle all the merging /boundary check little by little, but I'm afraid it's far too complicated for me.