[babakflame]

Dear Fellows

I have written a non-dimensional custom solver based on interFoam in OpenFoam 2.1.x.
In this solver there is a Poisson equation which is solved for a custom variable. (here ).
However, my solver crashes by insanely reducing time step if I increase the Dirichlet values at the boundaries.

The snippet related to Poisson equation is as follows:

fvScalarMatrix UeEqn
            (
                fvm::laplacian(eps,Ue) == -rhoE * gamma
            );
 UeEqn.solve();

Here, Ue is the only unknown variable. rhoE is solved in the previous equation and gamma is a scale appears by non-dimensionalization.


These are my relevant settings in fvSolutions and fvSchemes files:

"(Ue|UeExt)"
    {
        solver         PCG;
        preconditioner   DIC;
        tolerance        1e-14;
        relTol           0.0;
    };

laplacianSchemes
{
    default         Gauss linear corrected;
}

I have tried the following approaches to improve solver performance but failed

• employing smoothsolver and/or increasing relTol
• employing under-relaxation
• increasing the Dirichlet boundary condition from lower values to desired value gradually.

Is there any way to solve that Poisson equation with high boundary values? I am thinking of moving my solver to OpenFoam-dev or OpenFoam 4. Is there any chance that fvSolvers in these recent versions act more stably?

Has any body experienced this issue or any suggesstion??

Thanks

[Tobi]

Hi Bobi,

I have no idea about your solver and equations and I also never had »large boundary value problems (whatever large means)« ... one thing that might help. Under relaxation and making outer loops till the equation converged. Or field relaxation. For the pressure (poission eqn) we also use field relaxation. Out of the box I have no other idea -- sorry.

[babakflame]

Thanks Tobi