The Quadwild mesher is also able to mesh .stl files using quads only. The mesher works based on a triangulation of the surface, so first, a good triangulation is needed. The resulting quad elements get the size of the underlying triangles, so direct element sizing is not possible. The size of the quads can be further scaled using the Quad edge multiplier setting.
example figure shown in LinkedIn have a lot of partition for both clamp and piston models, tet mesh for input preparation of clamp models shown continuous mesh but hex result is mismatch. This can be different with currently available (experimental) of hexahedral dominant in Gmsh, previous tet mesh being hex as many as possible in volume parts. Unfortunately, percentage of tet element still exist, a linear one known not reliable and quadratic generates hanging nodes. May this limitation can be improved in PrePoMax preprocessor to connect with neighbor nodes by equation, CGX author state is possible doing this task also.
old version of 3.x took long in computation and generate hex-dominant mesh, even memory is large still made it temporarily frozen in second to minutes. In version 4.x being reactivated and run faster drastically in the same computer, overall result look similar with previous except in density of mesh required finer. Hex-dominant mesh algorithm being prepared for final release of version 5.x currently being developed.
QuadWild looks generate more uniform mesh when Quasi Structured Quads (Gmsh) needed some help by extra edge partitions. Currently disadvantages are in working with Stl files and sweep method did not work flawlessly and capable.
it seems a curve cannot be segmental by approximating number of edge line in QuadWild, but Gmsh still capable at this condition. Below example using Step file with true curve edge instead of segmental straight line.
It doesn’t respect the mesh size directly, hence the difference in mesh densities. The mesh handles transitions better and is more regular in general, but can sometimes fail locally:
Can you share the model where Quadwild fails? First, try the Preview option to see that the triangulation works. If I understood correctly, you divided the CAD edge into multiple segments to achieve a mesh with a precise element size?
The Quadwild can be improved by adding edge seed constraints (changing the source code, compiling, and testing), but I did not test it to see how that reflects the final mesh.
To diagnose such issues, it would be good to access the temporary files that QuadWild generates and then they are deleted immediately after meshing completes or fails:
