SimpleFoam convergen problem
 

Hi, Now I am using simpleFoam to calculate a pipe flow.
Now I face a converge problem.I am using low-Re k-epsilon model. After about 10 000 times irritation, k ,epsilon nearly converged, but ux and uy an not converge. Because the mesh number is very huge, and I have tried many ways, but still can not solve this problem. Mesh check is OK and the max skewness is about 0.5
Hope someone can give a help.
Below is the output:
================================================== ===============
smoothSolver: Solving for Ux, Initial residual = 0.00828136, Final residual = 7.04211e-05, No Iterations 9
smoothSolver: Solving for Uy, Initial residual = 0.0071747, Final residual = 4.91001e-05, No Iterations 10
smoothSolver: Solving for Uz, Initial residual = 7.61274e-06, Final residual = 8.03272e-07, No Iterations 4
GAMG: Solving for p_rgh, Initial residual = 0.000732179, Final residual = 7.03776e-05, No Iterations 13
GAMG: Solving for p_rgh, Initial residual = 0.000109843, Final residual = 1.07213e-05, No Iterations 27
time step continuity errors : sum local = 1.73118e-09, global = 8.26877e-13, cumulative = -5.21255e-10
smoothSolver: Solving for epsilon, Initial residual = 1.02623e-05, Final residual = 4.30998e-07, No Iterations 3
smoothSolver: Solving for k, Initial residual = 4.14421e-09, Final residual = 4.14421e-09, No Iterations 0
================================================== ===============
U:
Boundary condition:
inlet
{
type fixedValue;
value uniform (0 0 0.501);
}
outlet
{
type zeroGradient;
}
wall
{
type fixedValue;
value uniform (0 0 0);
}
}
p:
outlet
{
type fixedValue;
value uniform 0;
}
inlet
{
type zeroGradient;
}
wall
{
type zeroGradient;
}

fvSchemes:
divSchemes
{
default no;
div(rho*phi,U) Gauss linearUpwindV cellLimited Gauss linear 1;
div(phi,alpha1) Gauss linearUpwind Gauss linear;
div(phirb,alphai1) Gauss linearUpwind Gauss linear;
div(phi,k) Gauss linearUpwind Gauss linear;
div(phi,epsilon) Gauss linearUpwind Gauss linear;
div(phi,R) Gauss linear;
}

laplacianSchemes
{
default Gauss linear corrected;

}

interpolationSchemes
{
default linear;
}

snGradSchemes
{
default corrected;
}

fvSolution:
p_rgh
{
solver GAMG;
preconditioner DIC;
tolerance 1e-06;
relTol 0.1;
minIter 5;
maxIter 50;
smoother GaussSeidel;
nPreSweeps 1;
nPostSweeps 3;
nFinestSweeps 3;
scaleCorrection true;
directSolveCoarsest false;
cacheAgglomeration on;
nCellsInCoarsestLevel 50;
agglomerator faceAreaPair;
mergeLevels 1;
}
k
{
solver smoothSolver;
smoother GaussSeidel;
tolerance 1e-06;
relTol 0.01;
nSweeps 1;
maxIter 20;
}

epsilon
{
solver smoothSolver;
smoother GaussSeidel;
tolerance 1e-06;
relTol 0.01;
nSweeps 1;
maxIter 20;
}
U
{
solver smoothSolver;
smoother GaussSeidel;
tolerance 1e-06;
relTol 0.01;
nSweeps 1;
maxIter 20;
}

Have you tried with a more fine mesh?
Try refineMesh command and repeat simulation again.

Thanks, I'll try that later.

You can try to decrease your relTol parameter, as it is probably preventing the linear solver to reach the specified tolerance.
The values you are using seems to be a bit high.

I tried to decrease reTol parameter of p to 1e-3. the others to 0, but still, the residual of ux, uy and p decrease at first to about 1e-4, and then start to increase.

Hi Maolong LIU,

If you are having convergence problems with Above solver settings then try with PCG solver for p and PBiCG solver for other variables. It may take more time but it seems that these solvers are more stable as compared to GAMG solver.

You can also try Gauss linearUpwind cellLimited Gauss linear 1 scheme for divScheme for div (phi, U). for other variables limitedLinear 1.

Regards

Hrushikesh

Thanks, Hrushikesh
Now I am trying to start with a higher viscosity, and then gradually decrease it to the real value. This seems work, the calculation is still going on. If this fail, I will try your advice.
Thanks again.

I think I find the reason why my calculation seen hard to cenverge.
Now I am doing a 3D simulation, and through the calculation results I found tat ux and uy is very small, about 1e-5.
I heard some said that the convergence criteria of OpenFOAM for this kond very small results may have some problem.
Is this true?
Thanks.