[aep14]

All:

I am attempting to create an unstructured surface domain for an iced airfoil. I imported a *.stp file that has ~500,000 triangular elements. When I build the domain on the database, most (~80%?) of the resulting triangles are directly on the domain, as you would expect. The remaining triangles go "through" the database, and link to other vertices on the domain.

I've attached a picture. The triangles in question tend to form along the boundaries of the domain.

Here are some steps I've taken, and what I've observed:

1) Enforcing a max angle (5-10) doesn't help, but does increase the number of triangles.
2) Enforcing a max deviation (0.001-0.04) doesn't help, but does increase the number of triangles.
3) Pointwise appears to ignore the max/min edge length. When I put them as ~1.0 the maximum edge length in the domain ends up to be ~33 (!!) and the minimum edge length is ~1e-3.
4) Delaunay triangulation is much more robust in keeping triangles on the database than advancing fronts.

Any advice/tips would be greatly appreciated.

Thanks!

[cnsidero]

Originally Posted by aep14

All:

I am attempting to create an unstructured surface domain for an iced airfoil. I imported a *.stp file that has ~500,000 triangular elements. When I build the domain on the database, most (~80%?) of the resulting triangles are directly on the domain, as you would expect. The remaining triangles go "through" the database, and link to other vertices on the domain.

Is there an upper and lower surface to the airfoil, i.e. a seam at the leading edge? Can you post a screen or two of the just the geometry?

3) Pointwise appears to ignore the max/min edge length. When I put them as ~1.0 the maximum edge length in the domain ends up to be ~33 (!!) and the minimum edge length is ~1e-3.

Note, if you have either max angle or max deviation on, they will take precedence. However, the large edges are likely due to its inability to project properly to the surface and creating bogus triangles (with long edges).

4) Delaunay triangulation is much more robust in keeping triangles on the database than advancing fronts.

That has been my observation as well.