[mechkween]

Hi everyone,


I'm looking to solve a problem using rhoPimpleFoam (OF 2.4) but it looks like the solution can be obtained a little more efficiently. The problem at hand is a 2D rectangular section with a fixed mass flow rate at the inflow and a fixed static pressure at the outlet. I put in a forcing term in the UEqn.H and off my solver goes. It takes 40000 (2000 time loops and 20 pimple corrector loops) iterations to get to a good enough solution. I would like to reduce this because the solution is steady-state and that many iterations doesn't sound right (This is only a test case and the actual problem will be 3D and unsteady). I'm using the laminar model with sensible enthalpy. Here's a snapshot of the thermo properties:

thermoType
{
    type            hePsiThermo;
    mixture         pureMixture;
    transport       sutherland;
    thermo          hConst;
    equationOfState perfectGas;
    specie          specie;
    energy          sensibleEnthalpy;
}

First question, sensible internal energy seems to blow up after a couple iterations while sensible enthalpy is stable. Any suggestions why and how to get the former working?

Here's the 0 folder with the boundary conditions (the variable values are non-dimensionalised, Cp, As, Ts, molWeight were changed to match the non-dimensionalisation):
p:

internalField   uniform 10.26; //p = rho*T/gamma*M^2

boundaryField
{
    inlet
    {
        type            waveTransmissive;
        gamma           1.4;
    }
    outlet
    {
        type            fixedValue;
        value           uniform 10.26;
    }
    top
    {
        type            cyclic;
    }
    bottom
    {
        type            cyclic;
    }
    front
    {
        type            empty;
    }
    back
    {
        type            empty;
    }
}

U:

internalField   uniform (1 0 0);

boundaryField
{
    inlet
    {
        type            flowRateInletVelocity;
        rho             rho;
        rhoInlet        1;
        massFlowRate    constant 0.2;
    }
    outlet
    {
        type            waveTransmissive;
        gamma           1.4;
    }
    top
    {
        type            cyclic;
    }
    bottom
    {
        type            cyclic;
    }
    front
    {
        type            empty;
    }
    back
    {
        type            empty;
    }
}

T:

internalField   uniform 1;

boundaryField
{
    inlet
    {
        type            fixedValue;
       value           $internalField;

    }
    outlet
    {
        type            waveTransmissive;
        gamma           1.4;
    }
    top
    {
        type            cyclic;
    }
    bottom
    {
        type            cyclic;
    }
    front
    {
        type            empty;
    }
    back
    {
        type            empty;
    }
}

I replaced the zeroGradients with waveTransmissives and found the solution converging a little faster. Here is the pimple dictionary for this case:

PIMPLE
{
    nOuterCorrectors 20;
    nCorrectors     2;
    nNonOrthogonalCorrectors 0;
    momentumPredictor yes;
    rhoMin          rhoMin [ 1 -3 0 0 0 ] 0;
    rhoMax          rhoMax [ 1 -3 0 0 0 ] 2;
    pRefCell        0;
    pRefValue       10.26;
}

Using a low value of nOuter didn't help with oscillations being set up in the flow. This number seemed to do the job of driving to a steady state solution. I'm attaching the residuals for the cumulative continuity and initial Ux. My concern is with the oscillatory cumulative residual behaviour. What can I do to reduce this with a lesser number of iterations?


Thanks!