Remeshing support

[MakisH]

As of v3.2.0, preCICE supports dynamic meshes with remeshing. This also needs the adapter to support it.

The support should be relatively easy, and we could add it soon.

We would need a simple test case, ideally based on one of our tutorials. OpenFOAM seems to offer the dynamicRefineFvMesh class, I am not sure how well the adapter would already play with that. We could then use the perpendicular flap, for example.

Another interesting case could be the lid-driven cavity with a flexible bottom, which would be a nice validation example as well.

@fsimonis have you already tried anything with OpenFOAM and remeshing?

Triggered by https://precice.discourse.group/t/dynamic-coupling-interface-for-openfoam/2662

[precice-bot]

This issue has been mentioned on preCICE Forum on Discourse. There might be relevant details there:

https://precice.discourse.group/t/dynamic-coupling-interface-for-openfoam/2662/2

[fsimonis]

This feature has also been requested by @Crazy-Rich-Meghan.

[fsimonis]

Some notes we learned:

Timing

Currently, the meshing is separately timed/controlled as the remeshing using mesh.update() is out of the control of preCICE. We need to align this with the requirement of preCICE to only remesh in the first iteration of a time window.

Using the new mesh.controlledUpdate() could help. This is used in some solvers for beta testing like pimpleFoam.
More on gitlab

Checkpointing

The mesh update change the mesh as well as the dofs of fields defined on it. This directly influences the checkpointing, meaning the mesh update and the writing of the checkpoint needs to be synchronized.

If we assume that the remeshing is only allowed in the first iteration of a time window, then the procedure should looks as follows:

1. time window converges
2. read data from precice
3. update mesh if required which also takes care of remapping the read data to the new discretization. (needs confirmation)
4. create the checkpoint with the updated dofs.
5. perform the time loop

[MakisH]

We should also check whether some parts of the code assume a fixed interface mesh size throughout the simulation. See, for example, #398 (comment)