Continuing the discussion from A developer/preview PrePoMax version v1.5.0 has been released:
Hi Matej,
great new update! Will it be possible in the future or is it planned to add a local refinement/number of elements to edges? The current “Mesh Refinement” feature seems not to work properly for those meshes:
Currently, the mesh refinement is implemented but might not work as expected. Can you try the tetrahedral Gmsh option and try if the mesh refinement works?
I tried but as shown the mesh refinement is not taken into account (same result):
Also using “Tetrahedral_Gmsh” with “Transfinte…” set to “No” leads to strange meshing results:
Hmm, I see you were using .stl files for the last example. I think the mesh refinement will not work for stl geometries in Gmsh.
Using the step file, I get the following result using a single line mesh refinement:
Okay i didn’t realized that i imported a .stl file in the last example but i didn’t use mesh refinement for this. So using “Tetrahedral_Gmsh” works not so good for .stl files (i tried other .stl files with similar results).
Mesh refinement on compound parts seems to be the problem. No refinement is created in this example either:
Okay, currently the mesh size is applied in Gmsh to the points with the same IDs. So maybe/probably the IDs get mixed up. A better way would be to identify the points by their coordinates not IDs.
Sometimes the best option to capture stress concentration is to use submodel. In such a case you can use fine mesh in small part of the model without problem. Works in the PrePoMax perfectly well.
I agree, but I think I must improve how the mesh refinement works using the Gmsh.
Nice, just realized that there is also an advantage in time for meshing compound parts with “Meshing_Parameters” + “Transfinite_Mesh” (B) vs just using “Meshing_Parameters” (A).
Is it mainly because (A) uses netgen while (B) uses gmsh?
Even with setting “Recombine” to “No”, option (B) is nearly two times faster than (A) even though a double number of nodes and elements are created.
Model - Type - elements/nodes - Time(s)
1A - C3D10 - 845/1442 - 4,77
1B - C3D20 - 116/703 - 2,88
2A - C3D10 - 6722/10959 - 7,8
2B - C3D20 - 2592/12301 - 4,55
3A - C3D10 - 34055/52305 - 17,89
3B - C3D20 - 11940/53583 - 8,19
You can also use “Meshing_Parameters” + “Tetrahedral_Gmsh” to create tetrahedral meshes using Gmsh mesher.
Now in v2.0.1 mesh refinement using gmsh on compound parts is principle working. But unfortunately the refinement is often applied to other geometries than selected - see the right side:
Of course for a free tet-mesh netgen can be used, but i am more thinking of refining mapped hexa-meshes.
I hope it can be fixed 
Test.pmx (74.6 KB)
You are right. I focused only on a single part meshes. I will also try to find the solution for the compound parts.
