Lex,
In my FSI problem, there are 4 vector-valued variables + 3 scalar variables.
Most of them are strongly coupled with each other. It's hard to decouple the
system and design a good preconditoner. Previously, I solved the whole system
with the direct solver (UMFPACK) + Newton's iteration method.
Several days ago, I added deal.ii's new nonlinear solver (SNES) into my code,
and now its default method is jacobian-free newton-krylov. However, it behaves
bad and never get converged in my case. Maybe I am using it in a wrong way, as
I heard that SNES solver is powerful.
My first question is *what* doesn't converge. Is it the linear solver, or the
nonlinear solver?
I am not too familiar with PETSc's SNES, but it should be relatively easy to
use the KINSOL solvers (see step-77) instead where the choice of linear solver
is independent of the nonlinear solver. So you can continue to use the direct
solver for the linear systems (solving with the Jacobian) and still get the
benefit from using an excellent nonlinear solver package. If that works, you
can then independently think about using a better linear solver.
Best
W.
--
------------------------------------------------------------------------
Wolfgang Bangerth email: bange...@colostate.edu
www: http://www.math.colostate.edu/~bangerth/
--
The deal.II project is located at http://www.dealii.org/
For mailing list/forum options, see
https://groups.google.com/d/forum/dealii?hl=en
---
You received this message because you are subscribed to the Google Groups "deal.II User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to dealii+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/dealii/1975dbf1-1ef4-8122-fe7f-e37ca520d05a%40colostate.edu.
