Elmer Discussion Forum

The whole case is here
https://github.com/mrkearden/case

Thank you very much for doing that. It makes sense. I was using gmsh fragment option that does the same thing. However, there were some issues with that.

I noticed that you have Mortar BC = 3 in the boundary condition. but I did not see any boundaries named 3. did you use mortar boundary in this case?


! pos
Boundary Condition 1
Target Boundaries(1) = 1
Name = "VoltageP"
Potential = 1
Mortar BC = 3
End

No I did not the mesh was conforming

Sometimes, I have a mesh that is non-conformal. If you don't mind can you try it with mortar bc and see if that works?

I really appreciate it

Hi

Just some thoughts about meshing vs. mortar methods:

Meshing may be tedious and then there is a hope that one could use existing pieces of meshes and glue them together. In finite element methods you can do the glueing using so called mortar finite elements. These basically integrate over the shared boundary with the Galerkin method creating a weak condition between the two boundaries. Another alternative is to find the nodes (or even edges) directly in the mesh creating a strong projector between the values. This introduces significant errors.

Both of these methods introduce a constraint matrix and even if you are succesfull in creating the conditions you end up with a saddle-point problem which is considerably more challenging solve. You may in some cases perform elimination but this is still partly work in progress.

Hence, if possible, I would try to use conformal meshes. Those provide optimal accuracy. Only if you truly have rotating/sliding BC's are these BC's impossible to avoid. Time spent in meshing will be gained in more robust numerics.

-Peter

Thank you Peter,

I spent some time and made the mesh conformal. I am still figuring out how to make mortar bc works in this case, just for future reference.

Here is an example of a mortar connection.

Thank you