Unite 2 meshes after import from gmsh

[raback]

Hi Matthias

1) I found the following definition of conformal meshes on the web: "The intersection between any two elements is a sub-element of both: a face, an edge, a node or nothing". So in this sense it does not make the mesh conforming. It only add a linear interpolation of the boundary elements in the matrix that ensures continuity.

2) Sorry, both are bad in separate. You can see the spoiled mesh structure in looking at the bandwidth Elmer gives after bandwidth optimization. Maybe I was too pessimistic, but the added interpolation matrix does have an effect on the matrix structure i.e. what i&j combinations exist in A_ij.

-Peter

[mzenker]

Hi Peter,

1) As long as the physical result is OK, the mesh needs not to be conformal...
2) Can I conclude that it should be safe to use iterative solvers without parallelization?

Thank you,

Matthias

[raback]

You can use any linear solver that works, even the direct ones. For real periodic BCs there are some glitches in ElmerGrid that ensure that they come to the same partition on both sides. For this kind of internal BCs they do not however work.

-Peter

[mzenker]

OK, I will try it when merging nodes with Elmergrid fails.

Matthias

[mzenker]

Hi again,

I am wondering if the -merge option in Elmergrid could be extended/changed to act on selected boundaries only. This way, one could manage the situation where, for example, the adjacent boundaries 1 and 2 are meshed with a width of 0.01, and the adjacent boundaries 3 and 4 with a width of 0.5, and in both cases the nodes have to be merged.

Would this be feasible?

Matthias