[sylvester]

Hi,

As a test setup we have placed an empty MRF zone in a flow with with uniform freestream conditions, calculated using MRFSimpleFoam.
The MRF zone is a cylinder with the axis pointing in Z-direction. The freestream velocity is in X-direction. The cylinder has a radius of 1 m, see attached 'geometry-s.jpg'.

This is our MRF definition:

    cylinder
    {
        nonRotatingPatches ();
        origin  origin  [0 1 0 0 0 0 0] ( 0. 0. 0. );
        axis    axis    [0 0 0 0 0 0 0] ( 0. 0. 1. );
        omega   omega   [0 0 -1 0 0 0 0] 120.;
    }

The velocity is specified as:

internalField   uniform (50 0 0);

boundaryField
{
    sphere
    {
        type            freestream;
        freestreamValue $internalField;
    }
}

And the pressure as:

internalField   uniform 0;

boundaryField
{
    sphere
    {
        type            freestreamPressure;
    }
}

Our expectation was that, as the MRF zone is empty, the flowfield would be unaffected by it. Clearly this is not the case, as the screenshots of the resulting flowfield show.

We contacted OpenCFD about this (http://www.openfoam.com/mantisbt/view.php?id=78), but their reply was that an MRF region will always affect a cross-flow even if it is empty. But surely an empty MRF zone cannot have such a big effect?

Is this really what should happen, or is something going wrong here?

Sylvester

ps. OF 1.7.x is used, updated today. The mesh consists of tetrahedra only, created using ANSA.

[Ohbuchi]

Hi,

I'm really surprised with your result.
Then, I've tried another case with sphere boundary as wall, and initial U=0.
In this case, fluctuation caused by MRF was considerably small.
So I suppose this phenomena was caused by interaction free stream boundary with MRF.

[eugene]

Unfortunately, empty MRFs do have a huge influence. I bashed my head against the problem for more than a week and dissected the governing equations down to the bones. What you see is "correct". In general, I would say an MRF with spin axis perpendicular to the flow does not behave very nicely unless it is encased (i.e. has some kind of solid rim).

A free MRF with flow perpendicular spin axis and no rim will produce non-physical (but correct according to the model) results.

[bastil]

Interesting quote Eugene. I was not aware of this. C0uld you please go little more into detail?
This means there should be not MRF-Zone exceeding solid rims, right?

Regards Bastian

Originally Posted by eugene

Unfortunately, empty MRFs do have a huge influence. I bashed my head against the problem for more than a week and dissected the governing equations down to the bones. What you see is "correct". In general, I would say an MRF with spin axis perpendicular to the flow does not behave very nicely unless it is encased (i.e. has some kind of solid rim).

A free MRF with flow perpendicular spin axis and no rim will produce non-physical (but correct according to the model) results.

[eugene]

Hi Bastil,

I can speculate on what the all this means, but I am not really sure, so don't take anything I say here as the gospel truth.

If you take the MRF equations and write them all in terms of the absolute velocity, you end up with at least one term that does not contain a gradient and will be non-zero in a uniform cross-flow: (omega x Ui). If the flow is purely swirling around the axis, this term exactly balances another term that does have a gradient of U in it.

If the directions of omega and Ui are the same, then the cross product is zero. If not, then it results in a net force. So a perfectly uniform flow field passing through an empty MRF can be deflected, just as if there was some kind of rotor occupying the volume. In general, I think an MRF should be constructed such that the cell zone matches the swept volume as closely as possible. Extending the MRF to any volume outside this, could potentially lead to additional non-physical forces.

In an ideal world, we would have automatic detection of the swept volume based on solid body rotation and some kind of flux jump boundaries to take care of the non-perpendicular MRF internal boundary faces. Maybe, we could just turn off the additional MRF momentum term in non-swept parts of the MRF volume. The main thing to note, is that there is a potential source of error inherent in using MRF and caution should be exercised when it is being applied.

[sylvester]

Eugene, thanks for your excellent explanation. We tested it on a spinning wheel and indeed when we use a non-protruding MRF zone the cross-flow no longer appears.