[chriss85]

I'm currently running a simulation using the P1 radiation model and I notice extremely high residuals which don't appear to converge:

DICPCG:  Solving for G, Initial residual = 1, Final residual = 3.98945e+15, No Iterations 1001

In the next time step the simulation will crash due to the completely wrong radiation field.

Any idea what could be causing this?

checkMesh seems to be ok, only thing is that one patch is "multiply connected (shared edge)". Still, it fails no quality check.

fvSolution:

"G.*"
    {
        solver          PCG;
        preconditioner  DIC;
        tolerance       1e-08;
        relTol          0;
    }

The boundary condition for the radiation field G uses the Marshak BC with emissivity=1.

[chriss85]

I switched to the GAMG multigrid algorithm and it no longer crashed in the first time step and looks reasonable.
However, now it crashes in the second time step

Can anyone explain why some algorithms fail sometimes while others work?

[chriss85]

It seems to be working now, I think there was an error in my input data (material libraries), however, I still don't understand why the matrix solvers behave differently regarding convergence behavior (not convergence speed, but rather convergence itself). Do they have different stability ranges?