I’m reiterating here what we’ve previously discussed at the workshop to catalog the issue. I’m trying to set up a problem imposing Dirichlet boundary conditions on a 2D fault stepover. This model has dynamic rupturing faults with no friction and no prescribed slip such that slip will resolve on them as a function of the boundary conditions. Additionally, the material is divided into two regions: a purely elastic deforming one in most of the model and a smaller elastoplastic region near the tips of the faults.
After a few timesteps, the model fails to converge. After discussions and troubleshooting with Matt Knepley at the workshop, it appears that this issue is not related to convergence parameters I’ve chosen, but is rather associated with the underlying solvers used. This can be run either just using run.cfg or using run.cfg and custom_pc.cfg which Matt suggested using (although it doesn’t solve the issue).
I’m aware that currently, an update to version 2.2.1 is being prepared for July that will allow Paraview to plot solver residuals in order to help diagnose this issue. I’m looking forward to the results.