LES Dynamic models details
 

Dear all,

If I am not mistaken, the current dynamic models calculate the coefficients as a single scalar, not scalarfield, I mean C is not C(y), or C(x,y,z,t). The averaging process is not done in homogeneous directions.

I am a little confused, the original basic idea of Dynamic procedure would make sense better if averaging in homogeneous directions, wouldn't it?

Another question I have is, is the test filter used in OpenFOAM double size of the original filter? If I am not mistaken, simpleFilter is just another explicit filter used in dynamic procedure that differs from the implicit filter, but isn't the explicit simpleFilter (top-hat) the same as the implicit filter (top-hat too!)?

Please correct me if I was wrong.

Thank you very much.

dynOneEqEddy vs locDynOneEqEddy
 

Take a look at the source of "locDynOneEqEddy" which is a localized, dynamic LES model i.e C(x,y,z,t).

Contrast this with "dynOneEqEddy" which is dynamic but with a global coefficient i.e C(t).

Michael

I see, but what about the second question,
"is the test filter used in OpenFOAM double size of the original filter?"

Thank you, Michael!

the ratio of the two filter width is 2,this number was choosed by some researchers,they demonstrated it was optimum.

Which line in the source code, please?

the test filter size is (at least in dynSmag) controlled by the variable ce in the turbulenceProperties dict, I think.

The test filter width is set to twice the grid filter width by the following line in dynSmagorinsky:

volSymmTensorField MM =
sqr(delta())*(filter_(mag(D)*(D)) - 4*mag(filter_(D))*filter_(D));

The factor 4 results from substituting 2*delta for the test filter width. where delta is the grid filter width.

If I'm not mistaken, I think even the locDynOneEqEddy model is not localized..

EDIT: I stand corrected. The coefficients are volScalarFields, so that would imply local coefficients..

Well, you have to give it some sort of name :)

And I just don;t believe there is an universal shortcut to model all types or turbulence in a non-DNS mesh.

Personally, I think the development of LES small scale turbulence models has almost reached its peak.

But it seems hybrid method still has a long way to go.

Test filter
 

moved to new thread