evelknevel |
July 27, 2016 07:19 |
Convergence problem laplacian
Hey guys,
I programmed a little solver for solving the compressible Reynoldsequation.
Code:
surfaceScalarField phis = fvc::interpolate(U*h/(R*T)) & mesh.Sf();
{
tmp<fvScalarMatrix> PEqn
(
fvm::laplacian(p*(pow(h,3))/ (12 * (nu)),p)
- fvc::div(phis,p)
);
PEqn().solve();
For a high velocity of the shaft i get highly oscillating results. With a really really low relaxation (0.0001) its converge very slow to the expected Values for p.
My schemes i use are:
Code:
divSchemes
{
default none;
div((interpolate(((Uwelle*h)|(R*T)))&S),p) Gauss upwind;
}
laplacianSchemes
{
default Gauss linear corrected;
laplacian(((rho*pow(h,3))|(12*nu)),p) Gauss linear corrected;
}
I tried to use an upwind scheme for the laplacian term as well but i got some Errors.
How can i discretize the laplacian term with an upwind scheme, and do you have some other ideas to get it stable ?
Thanks,
Daniel :)
|