Hi Hugo,
On 23/06/2021 16:27, hugo.levy@gmail.com wrote:
Hi Robert,
See sfepy/discrete/fem/geometry_element.py, the geometry_data dict. This information is accessible through IDdomain.gel. The ifa above relates to the row index into the faces array.
Thanks! Indeed I was looking for info in the discrete/iga directory exclusively. Now the surface integral in 3D also works flawlessly .
pb.set_bcs() only deals with the BCs definitions. The actual calculation of BCs' values occurs in Equations.time_update() -> Variables.equation_mapping() -> EquationMap.map_equations() (sfepy/discrete/common/dof_info.py), where Field.set_dofs() (sfepy/discrete/common/fields.py) is called on the user-provided data. There: from sfepy.discrete.projections import project_to_facets, and finally project_to_facets() calls field.get_surface_basis(region). It's a bit convoluted :).
BTW. the test could be just a problem description file and a non-box-like domain (e.g. your circle), where you set ebcs by a function, and in a postprocess() function call variable.evaluate_at() on the boundary to verify the values provided.
Okay, that's quite intricate as you say. I did the test you suggested by setting the dirichlet condition on the inner circle to the x-coordinate. I then solved a laplace pde and called field.evaluate_at() at various locations on the inner circle. No bugs observed here, the values I got were the expected ones. Again, I can send you the problem description file I used in case you want to verify by yourself.
Yes, that would be great. It would be helpful to have for the eventual proper implementation of surface integrals.
Anyway, thank you for your help and patience. I hope this discussion thread will be helpful to others eventually. I will definitely continue to work with sfepy for my IGA-related applications!
You are welcome! If you have a nice IGA-related example, that could replace or complement the current ones, you may consider making a pull request - contributions are appreciated!
Cheers, r.