[krzychu111]

Hi,
is it possible to set residuals per direction for velocity in fvSolution? I've tried changing "(U)Final" to "(Ux)Final" or "(U1)Final" but it didn't work. Does anybody have any idea?



Background:
I'm working on axisimmetric no swirl case. My simulation doesn't stop only because of Uz residual which is not reducing below 1e-4. On my simulation Uz speed is in range from -1e-14 to 1e-14 m/s (which is still big enough to spoil the residual) and I found out (searching through the Internet) that setting Uz speed to zero require adjusting openfoam code and recompiling of which I don't feel catable

[pete20r2]

Is it axisymmetric as in a wedge mesh? 1e-14 is a good as 0 considering the machine tolerance. Because the numbers are so small, and near machine tolerance, rounding errors and oscillations cause large RELATIVE differences in the filed between steps, hence the fairly large residuals.
I think you need to find another way to decide if the case is converged if you want it to stop automatically. Evidently, you know that by watching the min/max plot for Uz and deciding that the Uz field is effectively 0.
I also found this:
Residual controls per direction
Courtesy of Bruno.

    residualControl
    {
        p 1e-6;
        Ux 1e-6;
        Uy 1e-6;
        Uz 1e-1;
        "(k|epsilon|omega|f|v2)" 1e-6;
    }

Chuck this in your SIMPLE/PISO/PIMPLE section of fvSolution.

[yambanshee]

I don't see the sense in wanting the residual of a variable you do not care about to want to settle. As pete said, the residual is a result of the actual number being so small that rounding errors become significant between iterations. If the rest of your residuals are converged, and any physical properties (force, pressure difference, etc...) are converged, then the simulation itself is likely converged.

[krzychu111]

@pete20r2 Thanks a lot


Now my simulation stops when it should. The reason why I wanted the simulation to stop itself is that I'm going to run ~4000 cases so it's crucial to limit unnecessary iterations.

[Diro7]

Sorry for reopening.

Is this "residual-control-per-component" feature still available?
I'm working with OF6 and specifying residual control settings like

residualControl
{
    p     1e-5;
    Ux    1e-5;
}

seems not to be working.
The solver just skips the convergence check on U.

I'm asking since I'm doing some simple 2D computations and one of the 2 components is zero. Checking on the whole vector prevents the residual control check to work properly.

Andrea

[hconel]

    residualControl
    {
        p 1e-5;
        Ux 1e-5;
        Uy 1e-5;
        "(k|epsilon|omega|f|v2)" 1e-5;
    }

Hi,
I am setting the residual control above for SIMPLE solver in OpenFOAM-v1906 and it does not work. Convergence on U is completely ignored. The only thing that works is the following:

    residualControl
    {
        p 1e-5;
        U 1e-5;
        "(k|epsilon|omega|f|v2)" 1e-5;
    }

Did anyone actually make Ux Uy Uz convergence control work?

[sourav90]

Originally Posted by hconel

    residualControl
    {
        p 1e-5;
        Ux 1e-5;
        Uy 1e-5;
        "(k|epsilon|omega|f|v2)" 1e-5;
    }

Hi,
I am setting the residual control above for SIMPLE solver in OpenFOAM-v1906 and it does not work. Convergence on U is completely ignored. The only thing that works is the following:

    residualControl
    {
        p 1e-5;
        U 1e-5;
        "(k|epsilon|omega|f|v2)" 1e-5;
    }

Did anyone actually make Ux Uy Uz convergence control work?

Hello,

Have you tried setting very low values? For me, in some axisymmetric wedge geometry simulation, residualControl of

PHP Code:

1E -3 


didn't work but when I changed to

PHP Code:

1E -8 (Ux & Uy) 


and

PHP Code:

1E-6 (p) 


respectively, it converged towards the right values in ~100 iterations.