Parameterization of a spherical surface or solid

in sections, like a soccer ball?

"Bill Chernoff" wrote in news:yYAVe.478787$5V4.19309 @pd7tw3no:

Or maybe a band around the middle and caps at the poles?

That sort of works with a sphere, but if I go to an ellipsoid it is harder to do.

What about using exactly half of an ellipse? If you cut the ellipse in half along either the major or minor axis, you should get the desired result.

I think you would still get the degenerate parameterization around the poles. The problem is that most surface creation algorithms are based on a rectilinear 2D approach that doesn't map well to some surfaces.

I spent several hours on a problem with splitting a model and found that SW can't get away from the underlying surface parameterization, even with the fill command.

In the cases I am working on a sphere is subdivided into 7 "chunks", 6 of which are identical, and one is a cube. In this way the sphere can be hex meshed with a very nice, well behaved brick element.

Can you give any more details about the analysis you are trying to perform (loading, material, program)?

I wouldn't fret too much about getting perfectly rectangular bricks. There are other ways to get decent results.

If you're doing solids curved surfaces, a P-tet mesh should work fine. Keep locally refining mesh until you get convergence in regions with high stress gradients.

Another strategy is to cut your sphere by layers, with an outer layer that is at least 4 elements deep in thickness. Auto quad-mesh the inner surface of the shell and offset the quad elements outward. Mesh the sphere interior with a tet mesh.

This is a contact problem. Bricks allow me to cut down on the dof while still getting good accuracy. I'm breaking up a sphere to get seven six sided chunks. It's the only way to get a decent mesh providing the underlying surfaces don't create problems. In this way there is very little distortion in the bricks so I can use the 8 noded variety which helps convergence.

The problem with trying to refine a P-Tet mesh in a non-linear problem with contact should be obvious. The high stress gradients move around from iteration to iteration.

In this problem there is a ball bearing between two other surfaces.