[mwaqas]

Hello everyone,


I am simulating a PCB steady state multi region heat transfer case (unfortunately, I cannot share my geometry). It involves only conduction as I have used convective BC. So, no fluid region is involved in the simulation. The simulation is working fine giving physical results but the convergence is too slow (getting 1e-5 residual in 100000 iterations). I have seen couple of posts on this problem but couldn't find any solution to speed up my simulation.

I have a good quality mesh with pure hexa mesh having small number for skewness and non-orthogonality. This checkMesh report is for one region but all other regions have more or less same quality with pure hexa cells.

Overall number of cells of each type:
hexahedra: 34272
prisms: 0
wedges: 0
pyramids: 0
tet wedges: 0
tetrahedra: 0
polyhedra: 0

Checking topology...
Boundary definition OK.
Cell to face addressing OK.
Point usage OK.
Upper triangular ordering OK.
Face vertices OK.
Number of regions: 1 (OK).

Checking patch topology for multiply connected surfaces...
Patch Faces Points Surface topology
top_1_top_1_to_solid_1460 504 ok (non-closed singly connected)
walls_top_1 5908 5952 ok (non-closed singly connected)

Checking geometry...
Overall domain bounding box (0.0003568386 0.0045414847 0.032217899) (0.00108 0.0094434212 0.038171206)
Mesh has 3 geometric (non-empty/wedge) directions (1 1 1)
Mesh has 3 solution (non-empty) directions (1 1 1)
Boundary openness (-6.8998151e-16 -1.1445138e-17 -5.2797844e-18) OK.
Max cell openness = 3.3008723e-16 OK.
Max aspect ratio = 9.1835275 OK.
Minimum face area = 3.3217452e-09. Maximum face area = 3.2063938e-08. Face area magnitudes OK.
Min volume = 5.8680568e-13. Max volume = 6.4026871e-13. Total volume = 2.0891145e-08. Cell volumes OK.
Mesh non-orthogonality Max: 16.708037 average: 1.5778153
Non-orthogonality check OK.
Face pyramids OK.
Max skewness = 0.61245144 OK.
Coupled point location match (average 0) OK.

Mesh OK.

End

When I use nNonOrthogonalCorrectors = 0 (case I), I get very slow convergence (log.without.txt). So, I used nNonOrthogonalCorrectors = 1 (case II) to speed up convergence. it has improved the convergence rate in 2nd loop (log.with.txt). But the initial residual for 1st loop in both cases always remain same for every iteration and both cases produce same temperature field temperature field at the end of same number of iteration.
It means with nNonOrthogonalCorrectors = 1 (case II), I get much lower residual at the end of 2nd loop and converged solution in much lower number of iterations. But the temperature field corresponds to the residual of 1st loop as of case I. Hence, it is not the actual temperature field.

Here is my fvSolution, I also tried PCG solver but it didn't help. Moreover, I am already using maximum relaxationFactor = 1, because it is a stable simulation.

solvers
{
h
{
solver GAMG;//PCG;
smoother GaussSeidel;
preconditioner DIC;
tolerance 1e-10;
maxIter 500;
// relTol 0.01;
}
}

SIMPLE
{
nNonOrthogonalCorrectors 0;
residualControl
{
h
{
tolerance 1e-6;
relTol 0;
}
}

relaxationFactors
{
equations
{
h 1;
}
}

Can you guys please tell me that what else case I do to speed-up the simulation.
Thank you

Regards