This removes some experimental code that was accidentally uncommented some time ago and also brings ALE support to the split solvers (tested with IPCS-A).
Convergence with moving mesh tested on the Taylor-Green test case and only with pure translation for now. The convergence is not perfect, but now gives much better results than before this pull request. Convergence results can be seen in the individual commit messages
There are still some remaining ALE issues:
Convergence is not perfect even with a purely translating mesh velocity field
Mesh movement with velocity fields that have divergence is probably more problematic (not tested at this time, used to be a problem so I assume it still is).
The XDMF output assumes that the mesh does not move (to create smaller files by only outputting the mesh once per file). This configuration should be turned off when the mesh morpher is active (xdmf_file.parameters['rewrite_function_mesh'] should depend on the mesh_morpher status)
Some post processors (the PlaneProbe I believe, probably the iso surface finder in the level set code, maybe more) assume that the mesh does not change (for efficiency). This is probably OK, they are not vital parts of Ocellaris and not on by default, but maybe there should be an error if the mesh_morpher is active or something
Included commits:
Start work on resurrecting the ALE functionality
Ensure that tests still work (disable ALE for tests)
Unbreak the Coupled ALE CG solver
The coupled DG solver is also close to converging with ALE
Convergence with IPCS-A and ALE now close to the Coupled solver
This removes some experimental code that was accidentally uncommented some time ago and also brings ALE support to the split solvers (tested with IPCS-A).
Convergence with moving mesh tested on the Taylor-Green test case and only with pure translation for now. The convergence is not perfect, but now gives much better results than before this pull request. Convergence results can be seen in the individual commit messages
There are still some remaining ALE issues:
Convergence is not perfect even with a purely translating mesh velocity field
Mesh movement with velocity fields that have divergence is probably more problematic (not tested at this time, used to be a problem so I assume it still is).
The XDMF output assumes that the mesh does not move (to create smaller files by only outputting the mesh once per file). This configuration should be turned off when the mesh morpher is active (
xdmf_file.parameters['rewrite_function_mesh']should depend on the mesh_morpher status)Some post processors (the PlaneProbe I believe, probably the iso surface finder in the level set code, maybe more) assume that the mesh does not change (for efficiency). This is probably OK, they are not vital parts of Ocellaris and not on by default, but maybe there should be an error if the mesh_morpher is active or something
Included commits:
Start work on resurrecting the ALE functionality
Ensure that tests still work (disable ALE for tests)
Unbreak the Coupled ALE CG solver
The coupled DG solver is also close to converging with ALE
Convergence with IPCS-A and ALE now close to the Coupled solver
Small fixes