CFD/Thermal Solver Mapping Configuration

Using preCICE
1

Hi,

I’m seeking advice for a custom interface between a CFD solver and a thermal solver. I am sending face-centered heat fluxes to the thermal solver, and receiving nodal surface temperatures and displacements
(the thermal solver has a moving mesh along with the heat transfer solution). The moving mesh velocity is quite small, but not negligible. The thermal solver typically consists of all tetrahedron, while the CFD mesh is all hexes.

For a nearest-neighbor mapping scheme, what is the recommended mapping constraint for either participant in this situation? I have experimented with “read-consistent” for both sides, but I keep noticing that some nodes on the thermal side get “hot spots” with respect to surrounding nodes at time points later during the coupled simulation. Should I go for a nearest-projection mapping in this case, or consider a consistent-conservative scheme?

2

Hi,
nearest-neighbour is best suited for matching meshes.

Have a look at our perpendicular flap tutorial. Here we use rbf mappings.

3

Both temperature and heat flux should indeed use a consistent mapping, as they don’t depend on the area. See, for example, tutorials/flow-over-heated-plate/precice-config.xml at master · precice/tutorials · GitHub

A nearest-neighbor mapping can definitely lead to such issues, especially if the mesh sizes are very different. The direction then is also important. I would try to sample from the fine mesh into the coarse, not from the coarse into the fine.

A conservative mapping would be wrong here. As @fsimonis already wrote, using an RBF mapping would be best.