[lakeat]

Anyone knows how to get the production and dissipation of TKE (turbulence kinetic energy)? Thanks

Is it in this way:

        volSymmTensorField D = symm(fvc::grad(U));
        //volTensorField Dprim = symm(fvc::grad(U - UMean));

        volScalarField prod = -((U - UMean)*(U - UMean)) && D;
        volScalarField epsilon = sgsModel->epsilon();

----------------------------------------------

• Rate of production of turbulence kinetic energy from the mean flow(gradient):

• Rate of dissipation of turbulence kinetic energy per unit mass due to viscous stresses:

[Heesei]

Dear Daniel

I am also interested in this topic. Therefore, I would like to know if you found an answer to your question in the meantime.


Cheers,

Irina

[lakeat]

Sorry, it'e been a long time, but doesn't the codes in my last post work for you?
you can try with a channel case and see if its distribution is correct.

Bye

[johndeas]

Hi,

why did you take the symmetrical part of grad(U) ? Why didn't you compute

R && fvc::grad(Umean)

?

Regards,

JD

[alberto]

Notice that the method epsilon() returns the SGS dissipation rate. Check doxygen to see how it is computed, depending on the specific model.

[johndeas]

Hi Alberto,

thank you for your answer. When I take a look at src\turbulenceModels\LES\incompressible\oneEqEddy\ oneEqEddy.C, I read:

void oneEqEddy::correct(const tmp<volTensorField>& gradU)
{
    GenEddyVisc::correct(gradU);

    volScalarField G = 2.0*nuSgs_*magSqr(symm(gradU));

    solve
    (
       fvm::ddt(k_)
     + fvm::div(phi(), k_)
     - fvm::laplacian(DkEff(), k_)
    ==
       G
     - fvm::Sp(ce_*sqrt(k_)/delta(), k_)
    );

    bound(k_, k0());

    nuSgs_ = ck_*sqrt(k_)*delta();
    nuSgs_.correctBoundaryConditions();
}

it appears that, in src\turbulenceModels\LES\incompressible\GenEddyVis c\GenEddyVisc.H (for example), epsilon() represents

ce_*sqrt(k_)/delta()

. But, what I am interrested in is G, the production. Since I am doing DNS computations, I have access to both the Reynolds Stress and the gradient of the mean component of velocity, and I do not want to use any eddy viscosity to compute it, like in

2.0*nuSgs_*magSqr(symm(gradU))

. I just need the formula to combine R and fvc::grad(Umean) properly.

I also do not understand why the symmetric part of the gradient of U is mentionned in the previous post.

Regards,

JD

[boger]

The double-inner product of a symmetric and asymmetric second-rank tensor is zero, so in that sense, only the symmetric part of gradU "survives" anyway. So using only the symmetric part is harmless, but perhaps unnecessary.

[alberto]

Originally Posted by johndeas

Hi Alberto,

thank you for your answer. When I take a look at src\turbulenceModels\LES\incompressible\oneEqEddy\ oneEqEddy.C, I read:

void oneEqEddy::correct(const tmp<volTensorField>& gradU)
{
    GenEddyVisc::correct(gradU);

    volScalarField G = 2.0*nuSgs_*magSqr(symm(gradU));

    solve
    (
       fvm::ddt(k_)
     + fvm::div(phi(), k_)
     - fvm::laplacian(DkEff(), k_)
    ==
       G
     - fvm::Sp(ce_*sqrt(k_)/delta(), k_)
    );

    bound(k_, k0());

    nuSgs_ = ck_*sqrt(k_)*delta();
    nuSgs_.correctBoundaryConditions();
}

it appears that, in src\turbulenceModels\LES\incompressible\GenEddyVis c\GenEddyVisc.H (for example), epsilon() represents

ce_*sqrt(k_)/delta()

. But, what I am interrested in is G, the production.

My comment was to point out that

volScalarField epsilon = sgsModel->epsilon();

does not return what you want. That epsilon is only the SGS epsilon.

Since you do DNS, just apply the definition and you will be fine :-)

Best,