[alberto]

Dear all,

I am checking the implementation of dynamic LES models in OpenFOAM, and I have a doubt on the calculation of the dynamic coefficients in locDynOneEqEddy.

The ck coefficient is computed dynamically as

volSymmTensorField LL =
        simpleFilter_(dev(filter_(sqr(U())) - (sqr(filter_(U())))));

    volSymmTensorField MM = simpleFilter_(-2.0*delta()*pow(KK, 0.5)*filter_(D));

    volScalarField ck =
        simpleFilter_(0.5*(LL && MM))
       /(
            simpleFilter_(magSqr(MM))
          + dimensionedScalar("small", sqr(MM.dimensions()), VSMALL)
        );

    return 0.5*(mag(ck) + ck);

In calculating LL and other quantities, the filter operation is applied twice, first with filter_, then with simpleFilter.

However, U (the one solved for in LES momentum equation) is already the filtered velocity, based on the filtering applied by the mesh. Why is the filter operation applied twice? Shouldn't it be applied once only, since the solved velocity is already filtered once by the mesh?

Is there any reference for this implementation?

Best,

[Tarak]

Hii Alberto,

I am also looking at the LDKM implementation in OF. Can you please tell me whether the in following part, in line 148, should it be sqrt(k_)?? Or is it supposed to be k^(3/2) ?

fvScalarMatrix kEqn
00142 (
00143 fvm::ddt(k_)
00144 + fvm::div(phi(), k_)
00145 - fvm::laplacian(DkEff(), k_)
00146 ==
00147 P
00148 - fvm::Sp(ce(D, KK)*sqrt(k_)/delta(), k_)
00149 );

Also can you provide me any proper reference for the model?

And if you have managed to have a look at it completely, can you
please let me know whether the implementation is perfect or not?

Thanks,
Tarak

[Tarak]

Hii Alberto,

I was running the backward facing step case (Le & Moin for Re_h=5100 using LDKM in Openfoam. But I got a shorter recirculation length (~5h whereas published results show ~6h). Can you please let me know if you have run this case, and/or if you have any other comments.

Thanks,
Tarak

[alberto]

Hi, I did not run that case. You might want to give a try to the new Lagrangian dynamic Smagorinsky model in OpenFOAM 2.0.x

Best,

[lakeat]

Originally Posted by alberto

Dear all,

I am checking the implementation of dynamic LES models in OpenFOAM, and I have a doubt on the calculation of the dynamic coefficients in locDynOneEqEddy.

The ck coefficient is computed dynamically as

volSymmTensorField LL =
        simpleFilter_(dev(filter_(sqr(U())) - (sqr(filter_(U())))));

    volSymmTensorField MM = simpleFilter_(-2.0*delta()*pow(KK, 0.5)*filter_(D));

    volScalarField ck =
        simpleFilter_(0.5*(LL && MM))
       /(
            simpleFilter_(magSqr(MM))
          + dimensionedScalar("small", sqr(MM.dimensions()), VSMALL)
        );

    return 0.5*(mag(ck) + ck);

In calculating LL and other quantities, the filter operation is applied twice, first with filter_, then with simpleFilter.

However, U (the one solved for in LES momentum equation) is already the filtered velocity, based on the filtering applied by the mesh. Why is the filter operation applied twice? Shouldn't it be applied once only, since the solved velocity is already filtered once by the mesh?

Is there any reference for this implementation?

Best,

Interesting, I am reading this dynamic localization model.
And it seems to me that this is the best model in OpenFOAM LES modeling so far, (I have no idea of what going on with 2.0), but I guess this model is based on 1995 Ghosal's paper. The simplefilter.() function, in my opinion, is not the first filtering, neither is it a test filter, it simply means to do an "integral over the volume", and I think this was supposed to do it in homogeneous direction as initially suggested by Germano (see dyn S model in OF, it is just an average operation), and improved by Lili and many others like Ghosal, Lund, Moin, etc., now this is done in OF really localized (within a cell??), it's all about the averging process so to calculate C.

But I failed to see the current implementation has fully realized what suggested in that paper.

I will study it further

[Tarak]

I would really like to know upon which formulation is the LDKM model in OpenFOAM based on, is it Ghosal's, Menon's or a modifed Germano approach? Also I would appreciate if anyone shares his opinion on using this model for cases like backstep, circular cylinder etc.

Thanks,
Tarak

[lakeat]

Originally Posted by Tarak

I would really like to know upon which formulation is the LDKM model in OpenFOAM based on, is it Ghosal's, Menon's or a modifed Germano approach? Also I would appreciate if anyone shares his opinion on using this model for cases like backstep, circular cylinder etc.

Thanks,
Tarak

Sorry I modified my original post a little bit. Just Google this dynamic model, you will find it.

[lakeat]

And I have no further experience, which means I really don't know how could this localization (I mean the way OF is using now) overcome the long standing notorious stability problem. Hope someone would enlighten me by some theoretical words.

Thanks

[Tarak]

ya, the LDKM formulation is indeed Sandeep Ghosal's (JFM 1995). But I am asking about the formulation in Openfoam, although it is intended to be LDKM, but I highly doubt on what formulation is it really based on.

[lakeat]

Originally Posted by Tarak

ya, the LDKM formulation is indeed Sandeep Ghosal's (JFM 1995). But I am asking about the formulation in Openfoam, although it is intended to be LDKM, but I highly doubt on what formulation is it really based on.

highly doubt? Why? I think it is pretty close.