improving boundary condition mapping

CHLNDDEV / OceanMesh2D

A two-dimensional triangular mesh generator with pre- and post-processing utilities written in pure MATLAB (no toolboxes required) designed specifically to build models that solve shallow-water equations or wave equations in a coastal environment (ADCIRC, FVCOM, WaveWatch3, SWAN, SCHISM, Telemac, etc.).

https://github.com/sponsors/krober10nd

GNU General Public License v3.0

179 stars 65 forks source link

improving boundary condition mapping #190

Closed WPringle closed 3 years ago

WPringle commented 3 years ago

changing intersect to ismember for the b.c. mapping as this allows for duplicate indices which can often occur for weirs as well as for islands (first==last).
Some other small miscellaneous changes also included

krober10nd commented 3 years ago

Perhaps this should land prior to #194 or be combined?

krober10nd commented 3 years ago

Lots of things have changed here. Could you summarize so I can take a better look at what to test?

WPringle commented 3 years ago

Yep, please see the recent commit of the changelog for the summary.

krober10nd commented 3 years ago

Thanks.

Please fix the typo "thr" in the docstring for msh.remove_attribute().
Please add a description for the new kwarg slope_calc to the docstring in msh.interp.
Could you elaborate more on what you mean by "- fix for msh.make_bc() with empty gdat"? Why would you have an empty gdat? Shouldn't this be an error?
It seems that several things have changed w.r.t. to the handling/mapping of complex boundary conditions when creating subdomains. Do you have an example or test to demonstrate this capability?

krober10nd commented 3 years ago

In the case of mapping boundaries, depending on how the user extracts the subdomain, it can create what I would term "dangling" arcs that connect two disjoint pieces of the mesh. Should these segments exist or be removed?

See the lhs in the image below. dangling_boundaries

krober10nd commented 3 years ago

A closer up on the lhs. The island and mainland boundary end up crisscrossing. dangling_boundaries2

WPringle commented 3 years ago

Thanks.

Please fix the typo "thr" in the docstring for msh.remove_attribute().

Please add a description for the new kwarg slope_calc to the docstring in msh.interp.

Could you elaborate more on what you mean by "- fix for msh.make_bc() with empty gdat"? Why would you have an empty gdat? Shouldn't this be an error?

It seems that several things have changed w.r.t. to the handling/mapping of complex boundary conditions when creating subdomains. Do you have an example or test to demonstrate this capability?

Made the fixes.

WPringle commented 3 years ago

As regard to the boundary condition handling and your example above. Yes, the boundary condition mapping will be unable to produce a "correct" b.c in many cases as you show.

One of things I concentrated on was to have this work for the case of weirs even when your subset intersected the weir. But this was really for a mesh where I only had boundary conditions on the weirs. I think once you intersect the larger boundary like that it is impossible to map the bcs over correctly and really you just need to recompute them.

WPringle commented 3 years ago

In summary, it will work as expected for boundaries fully enclosed within the subdomain and for simple intersections as long as that boundary doesn't wrap around back into the domain.

krober10nd commented 3 years ago

Great, thanks.

So in regard to to your comment here:

`msh()` more efficient storing of boundary conditions read in from fort.xx files, and `msh.write()` can write out arbitrary vertex indices (instead of just 1 to NP).

How does one go about writing a subset of vertex indices instead of 1 to NP?

WPringle commented 3 years ago

If you use the extract_subdomain with the new keep_numbering option it will keep the full mesh triangulation numbering. Then in write it just uses unique(m.t) to get the vertex numbers.

krober10nd commented 3 years ago

Alright thanks