OpenFOAM-CalculiX FSI Diverging Sim Never Ends

[Makis]

@Makis

I have the feeling I have seen this sitatuion before, but I cannot find the discussion right now.

I remembered where I have seen this before: Simulation gets stuck at same time window - #14 by rachaelesmith

@Mike_Tree

I’m explicitly instructing OpenFOAM to not adapt its time step. I’m attempting to set a solver coupling time step that is equivalent to each solver’s time steps such that sub-cycling won’t be required and I can keep total wall time to a minimum. This may not prove possible.

Avoiding subcycling is always a good first debugging step!

@Mike_Tree

Yes, I believe this was because I did not set a lower limit on the Calculix time step and it was driving itself to zero? I’m not entirely sure, but something was obviously causing the Calculix solver to diverge and leaving it to crunch hard on its solution.

Where do you see the divergence? You mean that it comes suddenly in the last time step and it never gets reported? It should still exit at the maximum number of iterations.

@Mike_Tree

I haven’t ventured to explicit coupling because I understand that the nature of my application has the best chance at convergence with implicit coupling. The low Re, almost equivalent density nature of the problem means a tight coupling will be required to keep things in check, right? I’ve since switched from Aitken to IQN-ILS. I also switched to a nearest-projection mapping, hoping that may increase my displacement accuracy and improve convergence.

Definitely these are steps that can help the convergence, but I am just trying to clear any software or configuration related issues first. While of course we all want perfect mapping and acceleration in a production case, I would usually start from explicit coupling, nearest neighbor mapping, no subcycling, sockets communication, etc, for debugging purposes.