potentially better force function

meshpro / dmsh

:spider_web: Simple mesh generator inspired by distmesh.

GNU General Public License v3.0

210 stars 26 forks source link

potentially better force function #85

Closed krober10nd closed 3 years ago

krober10nd commented 3 years ago

I've been experimenting with this force function from the paper attached Eq. 3.9
It indeed converges faster and produces higher mesh qualities than the default DistMesh one. .
Perhaps it would be useful to implement here?

imp_distmesh2d.pdf

nschloe commented 3 years ago

Interesting! I'll check it out later.

nschloe commented 3 years ago

I played around with the other forcing function a little bit and found virtually no improvement. Not sure what the difference to your observations is.

krober10nd commented 3 years ago

When building so-called geophysical multiscale meshes with orders of magnitude (like 100-m to 10 km in one domain) variations in mesh size, it converged faster than the Persson-Strang function.

What were your test cases?

krober10nd commented 3 years ago

Maybe try the rectangle with refinement case

nschloe commented 3 years ago

I tried with a disk mesh. If it's not better for that, I have little hope for other domains tbh.

krober10nd commented 3 years ago

I observed it influenced the rate of convergence of mesh quality. I'd like to run some comparisons at some point and will post them here or on Sm. Feel free to close it of course. Just found the paper interesting.

nschloe commented 3 years ago

I observed it influenced the rate of convergence of mesh quality.

The mesh quality is almost always 98% and more because the starting grid is so good. I'm not sure how you measure a faster rate of convergence.