[mike.franky]

Hi,

I am running a case with dynamic mesh (interDyMFoam) which also includes topological changes. Specifically, layers next to a boundary are added or removed. When I run it in parallel, everything is working fine up to the point where the layer to be deleted is located next to the processor's boundary layer (not the actual boundary of the geometry).

One thing I tried is to run the mesh in parallel until a problem occurs. Then, I reconstruct (both mesh and field), decompose again and re-run. However, for some reason, when the simulation restarts the pressure is sometimes inconsistent with what it was before the simulation stopped, and often leads to instabilities which cause the simulation to terminate.

The other thing I am considering is using manual decomposition and add all the layers that might get deleted into one processor. However, I can't find an example of what the datafile for the decomposition should like like. I understand it needs to map each cell to a processor. Does anyone have such an example?

Finally, the region where the layers will be deleted is a significant part of the mesh. Therefore, manual decomposition will be extremely inefficient. If anyone has a better idea please let me know.

Thanks in advance for any help

Cheers
Mike

[guptaayush164]

Hi
Surprisingly I am working on a similar problem and facing the same issue where manual decomposition provides no advantage when the case is run in parallel. While simple, hierarchical, and scotch methods didn't work. Did you find a solution to it?