[Getfem-users] [getfem-users] Build a mesh_fem over an array of meshes

[Domenico Notaro]

Dear Users,

I have a generic question about the feasibility of the following problem.

As in previous topics I am dealing with a 1D network immersed in a 3D domain. We solve the coupled 1D-3D model in both pressure and velocity.

Concerning only the 1D network problem, we would like to implement a fragmented mesh for velocity in order to deal with compatibility conditions at junction points (*). More specifically, we want a unique mesh for pressure in the whole network and an array of meshes for velocity in each branch of the network. Then we need to build the local matrices (including mixed pressure-velocity ones) and to copy them to the monolithic iteration matrix.

Up to now, I am able to assemble the std::vector<getfem::mesh>. My concerns are about defining the finite element methods on multiple meshes. I was thinking to implement a std::vector<getfem::mesh_fem> in which each element has been linked to one mesh. As an example, I defined a simple constant function over a Y-shaped domain:
f=alpha  on Branch 0
f=beta   on Branch 1
f=gamma  on Branch 2
but when I try to export the values over the three meshes to the same vtk file I got only values over the first mesh points (Branch 0).

Do you think there is a way to handle with this problem? Maybe within the context of interpolated fems?

(*) Since velocity is discontinuous at junctions, we need different dofs for velocities over each branch affecting the node in order to impose mass conservation


Best Regards,
domenico_notaro