CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   slow convergence in chtMultiRegionSimpleFoam (http://www.cfd-online.com/Forums/openfoam-solving/85691-slow-convergence-chtmultiregionsimplefoam.html)

sven82 March 4, 2011 09:24

slow convergence in chtMultiRegionSimpleFoam
 
Hi Foamers,

I simulate a cht-case.
the case setup for the fluid
p{
solver GAMG;
tolerance 1e-6;
relTol 0.01;
}

".*" //other fields
{
solver PBiCG; //also GAMG
tolerance 1e-7;
relTol 0.1; //also 0.01 and 0
}

The relaxation are
T 1;
k,omega 0.7;
U 0.4;
p 0.7;

the same for the solid temperaturefield.
The problem is the very slow convergence,
the delta temprature between the iteration steps (deltat)
are something like 0.003 Kelvin !!!????

Im not sure how I come two a faster convergence?
I try also relTol from 0.01 and 0.
and lower relaxations.
But for my eyes the relaxation
got high values

M3hdi January 9, 2012 04:40

Hi sven82,

I am facing the same problem of low convrgence for T field in the solid region in a CHT case.

Have you solved the issu since then ?

Thanks in advance

M.

ulli January 9, 2012 05:53

Dear all,

we got a much fast convergence for chtMultiRegionSimpleFoam in many cases when we increase the number of iterations for T in the solid region (e.g. more nonOrthogonalCorrectors in the solid regions).
Hope this helps.

Ulli

Toorop January 24, 2012 08:40

Hi Ull,

great tip, the increased iteration number really helps the cht solver in the solid domain.

Anyone with experience on the topic of chtMultiRegionFoam solvers? When to use the steady-state / transient version?

anothr_acc October 2, 2014 10:15

I was in the same place, using layers in OF 2.3.0 with compressible::turbulentTemperatureCoupledBaffleMix ed. While I don't understand why convergence is so different with solids with and without layers, my incredibly slow convergence (days with 48 cores, steady state and a few hundred thousand cells) was accelerated by adjusting the relaxation factor for h up from 0.7 to 0.9. I saw no difference with nNonOrthogonalCorrectors as my convergence criteria were being met on the second sub-iteration. Successive iterations caused by setting nNonOrthogonalCorrectors gave no change to h residuals.

Good luck all.

Mark.


All times are GMT -4. The time now is 19:43.