On Thu, Aug 27, 2015 at 6:38 AM, <martin.lue...@geo.uzh.ch> wrote: > > When using the FemSystem class, what is the best way to implement > Neumann and Robin boundary conditions?
For "standard" boundary Neumann/Robin terms, override side_time_derivative in your FEMSystem subclass. That function is only called for elements on the boundary of the domain by default, so just check if the boundary_id of the element matches the boundary of the boundary where these conditions apply and add the terms to the residual. > Is this still by adding penalty > terms to matrix and RHS? > Those were for Dirichlet boundary conditions, so I'm not quite sure what you mean here in the context of Neumann/Robin boundary conditions. > If yes, does this collide with using the DirichletBoundary class (I > think it shouldn't)? No, it won't collide with DirichletBoundary constraints. > Or should all bcs be enforced with penalties? > You're free to use penalty boundary conditions, but I advise using DirichletBoundary for Dirichlet boundary conditions unless you're playing with Nitsche-type method. Please let us know if you need more clarification. Best, Paul ------------------------------------------------------------------------------ _______________________________________________ Libmesh-users mailing list Libmesh-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-users
