CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   using petsc for large coupled linear systems (

sd_vakip January 31, 2011 13:22

using petsc for large coupled linear systems
Hello to all,

I am in trouble using petsc.
I have solved Free-Convection flow in annulus on a 90x100 grid using PGMRES-ILUT of Saad's SPARSKIT. I am using a coupled algorithm with PWIM for face velocities to solve continuty equation with momentum and energy equations simultaneously. to It worked well up to 90x100 cells (90x100x4 equations) with non-dimensionalized time step t=5. To study the grid independency, I used 180x200 grid with the same solver settings and B.C.'snd t=5. However, it does not converge on the fine grid. I reduced the time step size up to t=0.001, but it did not work. Up to now, I have no idea about this issue: it is caused by the solver or the problem is too taugh to solve.
Therefore, I used PETSc in order to utilize more options on both preconditioners as well as solver and also, to use its parallel processing benefit.
I tests almost all preconditioners and ksp solver of petsc and they can not find a solution (do not converege) even though, for 90x100 cells!
1- How could I use the same preconditioner and solver (i.e. ILUT and GMRES with fill-in=100 and droptol=0.01) by petsc to get the same result for 90x100 cells?
2- Do you have any idea about what makes the solver to diverge when I use 180x200?

Many Thanks,

All times are GMT -4. The time now is 18:40.