CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM (http://www.cfd-online.com/Forums/openfoam/)
-   -   Initial residuals do not decrease... (http://www.cfd-online.com/Forums/openfoam/90018-initial-residuals-do-not-decrease.html)

Igor_2011 June 28, 2011 15:51

Initial residuals do not decrease...
 
Hi everyone!

I'm rookie on the forum and I'd like to start by greeting all the members for the very interesting topics that are being discussed here.

I recently tried to simulate the unsteady around a basic 2-D cylinder using the solvers available in OpenFoam, in particular pimpleDyMFoam. Basically the code I use is the looks like:

Code:



        // Do any mesh changes
        mesh.update();

        // Make the fluxes relative to the mesh motion
        fvc::makeRelative(phi, U);

        if (mesh.changing() )
        {
            #include "meshCourantNo.H"
        }

      for (int ocorr=0; ocorr<nOuterCorr; ocorr++)
        {
        // Pressure-velocity SIMPLE corrector
        {
            #include "UEqn.H"
       
            #include "pEqn.H"

        }
       
        turbulence->correct();
        }

where the velocity and pressure are solved in the following parts:

Code:


    // Solve the Momentum equation

    fvVectorMatrix UEqn
    (
    fvm::ddt(U)
      +fvm::div(phi, U)
      + turbulence->divDevReff(U)
    );

    UEqn.relax();

    eqnResidual = solve
    (
        UEqn == -fvc::grad(p)
    ).initialResidual();

    maxResidual = max(eqnResidual, maxResidual);

Code:


    p.boundaryField().updateCoeffs();

    volScalarField AU = UEqn.A();
    U = UEqn.H()/AU;
    //UEqn.clear();
    phi = fvc::interpolate(U) & mesh.Sf();
    adjustPhi(phi, U, p);

    // Non-orthogonal pressure corrector loop
    for (int nonOrth=0; nonOrth<=nNonOrthCorr; nonOrth++)
    {
        fvScalarMatrix pEqn
        (
            fvm::laplacian(1.0/AU, p) == fvc::div(phi)
        );

        pEqn.setReference(pRefCell, pRefValue);

        if (nonOrth == 0)
        {
            eqnResidual = pEqn.solve().initialResidual();
            maxResidual = max(eqnResidual, maxResidual);
        }
        else if( ocorr == nOuterCorr-1 && nonOrth == nNonOrthCorr)
        {
                        pEqn.solve(mesh.solver(p.name() + "Final"));
        }
        else
        {
            pEqn.solve();
        }

        if (nonOrth == nNonOrthCorr)
        {
            phi -= pEqn.flux();
        }
    }

#  include "continuityErrs.H"

    // Explicitly relax pressure for momentum corrector

    p.relax();

    // Momentum corrector
    U -= fvc::grad(p)/AU;
    U.correctBoundaryConditions();

As you can see, two inner loops are performed during each iterations. From top to bottom:
  1. loop for non-orthogonality correction on the pressure equation (nCorr
  2. velocity- pressure correcion
My question here concerns the initial residual at the beginning of each inner iterations. Typically for nNonOrthCorr=3 and nOuterCorr=2 the following residual history is obtained:

Code:


Mesh Courant Number mean: 1.03403 max: 1645.15
smoothSolver:  Solving for Ux, Initial residual = 0.000663609, Final residual = 2.88668e-06, No Iterations 8
smoothSolver:  Solving for Uz, Initial residual = 0.00548648, Final residual = 2.28391e-05, No Iterations 8
DICPCG:  Solving for p, Initial residual = 0.828448, Final residual = 9.83331e-07, No Iterations 506
DICPCG:  Solving for p, Initial residual = 0.211827, Final residual = 9.78121e-07, No Iterations 478
DICPCG:  Solving for p, Initial residual = 0.0673258, Final residual = 9.56774e-07, No Iterations 468
time step continuity errors : sum local = 5.6259e-12, global = 4.41739e-20, cumulative = 1.4129e-19
smoothSolver:  Solving for omega, Initial residual = 7.3622e-06, Final residual = 7.40485e-09, No Iterations 4
smoothSolver:  Solving for k, Initial residual = 0.0114753, Final residual = 3.82431e-06, No Iterations 3

smoothSolver:  Solving for Ux, Initial residual = 0.000132713, Final residual = 1.01532e-06, No Iterations 8
smoothSolver:  Solving for Uz, Initial residual = 0.000974607, Final residual = 6.52005e-06, No Iterations 8
DICPCG:  Solving for p, Initial residual = 0.804658, Final residual = 9.75985e-07, No Iterations 513
DICPCG:  Solving for p, Initial residual = 0.241636, Final residual = 9.50152e-07, No Iterations 481
DICPCG:  Solving for p, Initial residual = 0.0714927, Final residual = 9.76869e-07, No Iterations 458
time step continuity errors : sum local = 5.33943e-12, global = 1.72288e-20, cumulative = 1.58519e-19
smoothSolver:  Solving for omega, Initial residual = 1.39838e-06, Final residual = 8.80479e-09, No Iterations 3
smoothSolver:  Solving for k, Initial residual = 0.00313893, Final residual = 1.69267e-06, No Iterations 3

and I don't understand why my initial residual for the pressure is not equal to the final residual at the previous sub-iteration. In particular, should I consider the case to be converged when the first initial residual is below a pre-defined threshold?

I'm a beginner in CFD so do not hesitate to correct or highlight any misinterpretations or vocabulary misuses.

Great thanks!

Igor

L1011 July 11, 2011 14:05

I'm also interested by this post, nobody has a clue?

Yahoo March 6, 2013 14:07

I am trying to solve equation

fvScalarMatrix CEqn
(
fvm::ddt(C)
+ fvm::div(phi, C)

);
CeqnResidual = CEqn.solve().initialResidual();

In case where phi is zero (i.e. no flow), my initial residuals are always high (~ 0.3). When I am forcing the solver to exit without satisfying the convergence criterion (by setting a maximum for the iterations), I see that phi is zero everywhere and C is constant, as I expect. But, I don't know why CeqnResidual is high!


All times are GMT -4. The time now is 23:56.