A new release version of the PrePoMax contains the following FEM features:
- added support for sweep meshing
- bug fixes
The sweep meshing was tested on a couple of examples. All of them could be sweep meshed with appropriate algorithm settings.
The user is supposed to select the source surfaces (one or multiple), then the algorithm determines the side surfaces and the target surface (one).
Limitations of the algorithm:
- all side surfaces must be 4-sided surfaces
- if a side surface is split into multiple 4-sided surfaces that form layers away from the source surfaces, all other side surfaces must be split at the same positions, creating the same number of layers on all side surfaces
- a periodic side surface with a single seam (a cylinder with a single edge in the axial direction) must be split at least one more time to form two 4-sided surfaces
- the target surface must be only one
Currently, the algorithm is based on topology constraints only, and no geometrical constraints are included. That may result in strange/invalid results in special cases. Some geometrical constraints will be added in the future if necessary.
In most cases, the sweep mesh algorithm can replace both the extrude and the revolve procedures.
3 Likes
Can you please upload examples for these special cases? I did some tests and also often got extremely distorted and collapsed elements, but was able to prevent this by changing some options. So I would really like to test this with the parts you mentioned.
Did you experience any problems where the meshing failed but should succeed? Or succeeds in other software?
I did not test all the possibilities; I only imagined them 
For example, consider a solid extruded profile with a rectangular cross-section and one additional division at the half extrusion length. Then, increase the size of the middle cross-section only. At some point, the start angle of 180° of the surfaces at the middle cross-section becomes so small the mesh becomes unusable. I know an experienced user will not attempt something like this, but the goal is to apply sweep messing on compound parts automatically, and then a robust algorithm will be needed that could predict such unusable cases.
Yes, with the default settings even for simple bended tube parts and coarse mesh in sweep direction like this:
It looks good from the outside but generates colabsed elemens:
I have always had good results with these settings so far (most important here’s to deactivate “Global smooth steps”):
So I was wondering if this setting might be more useful as the default setting… But that would require testing more different parts.
Ok, the global smooth steps are mainly required if the source and target surfaces are not simple filled surfaces of their edges but have a concave or convex shape.
In such cases, I could limit the smoothing to only a couple of element layers at the source and target surfaces while the global smoothing is applied to all elements in the mesh. In the case of bent meshes, the smoothing tends to move the nodes towards the bent center, as is probably the case in your example.
Changing the default setting could also be a solution.
** Edit **
The mesh optimization is part of the Gmsh 2D meshing procedures. In your case, it only affects the mesh at the source surface, so I think it should not greatly improve the mesh quality.
