Hello,
OK, I obtain results similar to yours using the constant time step option. To avoid divergence, I need either to reduce the number of MPI ranks (from 48 to 4 works,), or keep the number of MPI ranks but change linear solver options to ones that are less dependent on the partitioning (ie. Jacobi instead of Gauss-Seidel, ....).
I'll try to see with colleagues if they have any idea about this, and I may run a few tests with sub-iterations on my side.
Best regards,
Yvan
Output forces on surface
Forum rules
Please read the forum usage recommendations before posting.
Please read the forum usage recommendations before posting.