Couping with explicit Structure solver

Hi everyone,

Recently I’m doing the FSI tutorial to implicitly couple my explicit FEM solver with openfoam, and I found that the explicit FEM solver shows much more instability when compare with implicit solver.

We already know that explicit FEM solver will produce noise in its results, so this kind of contaminated displacement results will cause crash in openfoam.

For example, in dam breaking case, the 2nd order CrankNicolson scheme in time discretization will lead to crash in simulation, but only the 1st order Euler can be used for convergence. And in turek-hron-fsi3 case, the fluid simulation crash in 0.025s when the displacement are in merely 10^-8 level.

I tried to subcycle openfoam in turek-hron-fsi3 (Although it’s not recommended), the simulation seems work well in the beginning, but as the openfoam-adapter says, subcycling will lead to instability, the pressure in subcycling time step will oscillate, and eventually leading to crash in about 3s. But results seems fine in first 2s.


So my questions are:
1: Is there any suggestion of dealing with the unstable coupling problem with explicit FEM solver?
2: Is openfoam’s subcycling feature developed ready now? If it is not, is there any instruction to develop it?

Thank you!

This is a benchmark case designed to reveal such issues. You essentially have to use an implicit solver and an implicit coupling scheme.

This is mostly a matter of checkpointing. We are still not keeping full checkpoints in the OpenFOAM adapter, mainly due to limitations (of our understanding of) the function objects interface. See some related issues: Issues · precice/openfoam-adapter · GitHub

This is expected.

Well, not so much is happening in these first 2s, right? Note also that the magnitude and frequency seem to be very different than the ones deal.II leads to.

1 Like