[PerryLJohnson]

Alberto,

Sorry to bug you some more on this code. I was revisiting your code, and it seems to me you have left out two coefficients in your MM equation (Germano et al. Eq 15, right hand side): the 2 out in front and the sqrt(2) in the | Sij | = sqrt (2 * Sij * Sij).

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

But I also see that the sqrt(2) is cancelled out in the definition of nuSgs because it is not included there either.

nuSgs_ = max(cD(D)*sqr(delta())*sqrt(magSqr(D)), -nu());

The leading coefficient of 2 (coming from tau_ij = 2 * nu_eff * S_ij) does not appear to cancel out since the definition of the effective stress is higher up in the turbulence model code.

00096 tmp<fvVectorMatrix> GenEddyVisc::divDevBeff(volVectorField& U) const
00097 {
00098 return
00099 (
00100 - fvm::laplacian(nuEff(), U) - fvc::div(nuEff()*dev(T(fvc::grad(U))))
00101 );
00102 }

So it seems to me there should be something to the effect of
MM *= 2;
directly after your definition of the MM tensor.

Please correct me if I am missing something here.

Regards,
Perry

[alberto]

Hi, my code is entirely based on what OF calls "homogeneousDynSmagorinsky" model. I think the reference is not Germano et al. but Lilly (1992) "A proposed modification of the Germano subgrid-scale closure method".

If there is a bug in the original code, then there is a bug in mine too :-D. I will check into that as soon as I can.

[Bernhard]

I also have the impression there is something wrong with the definition of MM, see this post: http://www.cfd-online.com/Forums/ope...details-2.html But if you can point me at my mistake there you're very welcome

@alberto: In the Lilly formulation c(D) is devided by two (Cs=LM/(2*M*M)), which is not incorporated here, so if you apply his suggestion, MM *= 2, then you would end up with exactly that right?

[alberto]

I am sorry, I pointed to the wrong reference in my last post. The implementation of LES models in OpenFOAM is based on:

C. Fureby, G. Tabor, H. G. Weller, and A. D. Gosman, A comparative study of subgrid scale models in homogeneous isotropic turbulence, Phys. Fluids 9, 1416 (1997); doi:10.1063/1.869254

According to this work:

k_sgs = c_I*(delta*|D|)^2

nu_sgs = c_D*(delta^2)*|D|

In the dynamic approach, they define:

c_I = <0.5 tr(L) m>/<mm>

c_D = <L_D M>/<M M>

with L_D, m and M defined as a function of D, in a way that seems consistent to me (please, check, I went through it quickly) with the OF implementation.

[PerryLJohnson]

I am in agreement with Bernhard. In Lilly's paper, the coefficient '2' in equation 8 (Fureby equation 6) becomes manifest in the coefficient '1/2' in equation 11. It seems to mysteriously vanish in Fureby's derivation, however, in his equation 7.

[alberto]

Hi, I agree. However, since the problem also affects homogeneousDynSmagorinsky, I think you should report it on OpenCFD bug tracker. This might provide an explanation too.

Best,

[PerryLJohnson]

I put in a bug report, we'll see if an explanation is forthcoming.

[alberto]

Bug corrected upstream, and fix merged in my modified version. You can download it from the usual git repository ( https://github.com/AlbertoPa/dynamicSmagorinsky ).

Thanks for pointing this out.

Best,

[deniggo]

Hi,
Alberto, thanks for providing the code. I compiled it in OF 2.0 and changed input-files as described in readme.
My case works but it stays laminar, while in the standard Smagorinsky with constant coefficients turbulence is developed. I use pisoFoam and pimpleFoam. Did I forget something to change or is the model not applicable to piso/pimpleFoam?

Thanks for your help!

Nico

[alberto]

Hi, depending on your case you need to initialise the simulation with a perturbed field to favour the development of the turbulent structures. There are is a tool in the forum for channel flows (perturbU) and probably some user adapted it to pipe flows too.

Best,

[engma3]

Hi alberto,
Thanks for providing the code. I use your code for simulation of a swirl jet. The problem is in calculation of kinetic energy , It gives me negative values.

Thanks,
engma3

[alberto]

The value should be bounded to kMin_.

Best,

[feldy77]

Dear Alberto,
Thank you very much for your contribution. I tried to use your model with buoyantBoussinesqPimleFoam. The compilation process was completed successfully. Then I modified LESProperties and controlDictionary files according to your instructions. Unfortunately something screws up when I try to run the simulation. I am getting the following warning:
--> FOAM Warning :
From function dlLibraryTable:pen(const fileName& functionLibName)
in file db/dlLibraryTable/dlLibraryTable.C at line 78
could not load /home/yurifeld/OpenFOAM/OpenFOAM-1.7.1/lib/linux64GccDPOpt/libdynLocalAverageSmagorinskyModel.so: undefined symbol: _ZNK4Foam14incompressible9LESModels11GenEddyVisc1B Ev


Do you have any idea what can be the problem?
Thank you very much, Yuri

[PerryLJohnson]

Originally Posted by feldy77

Dear Alberto,
Thank you very much for your contribution. I tried to use your model with buoyantBoussinesqPimleFoam. The compilation process was completed successfully. Then I modified LESProperties and controlDictionary files according to your instructions. Unfortunately something screws up when I try to run the simulation. I am getting the following warning:
--> FOAM Warning :
From function dlLibraryTable:pen(const fileName& functionLibName)
in file db/dlLibraryTable/dlLibraryTable.C at line 78
could not load /home/yurifeld/OpenFOAM/OpenFOAM-1.7.1/lib/linux64GccDPOpt/libdynLocalAverageSmagorinskyModel.so: undefined symbol: _ZNK4Foam14incompressible9LESModels11GenEddyVisc1B Ev


Do you have any idea what can be the problem?
Thank you very much, Yuri

Yuri,

You should be able to safely ignore this warning, see post #21.

Perry

[alberto]

Yes, Perry is right :-)

Btw, you could also consider the langrangianDynSmagorinsky model in OpenFOAM 2.x, which is based on a more rigorous treatment of the dynamic coefficient.

Best,

[alberto]

For your information, I moved my repository to bitbucket, so future versions will be here:

https://bitbucket.org/albertop/dynamicsmagorinsky

[ehsankf]

Dear Alberto,

I would like to know if this new dynamic smagorinsky model could be changed to do averaging in homogenous direction ( not whole domain) in case of channel flow .
Thanks in advance

[babakflame]

Dear Alberto

I am trying to include dynamicSmagorinsky SGS model for reactingFoam solver, I did exactly as u have said in
https://bitbucket.org/albertop/dynamicsmagorinsky
However, When I executed the file; I got this error:

Selecting LES turbulence model dynamicSmagorinsky

--> FOAM FATAL ERROR: 
Unknown LESModel type dynamicSmagorinsky

Valid LESModel types:

6
(
DeardorffDiffStress
Smagorinsky
SpalartAllmaras
homogeneousDynOneEqEddy
lowReOneEqEddy
oneEqEddy
)


    From function LESModel::New(const volScalarField&, const volVectorField&, const surfaceScalarField&, const basicThermo&, const word&)
    in file LESModel/LESModel.C at line 140.

FOAM exiting

It seems that the code has not recognized dynamicSmagorinsky.H & .C files.
I should note that I have put the dynamicSmagorinsky folder in My home directory and not in OpenFOAM sources, My openFOAM version is 2.1.x

Would You PLZ help me?

Any hint is appreciated

Best Regards
Bobi

[ehsankf]

you should include the new model in controlDict.

[babakflame]

Dear Ehsan
Hi
I have included this line at the end of controldict file:

libs ( "libOpenFOAM.so" "libdynamicSmagorinskyModel.so" ) ;

However, The problem is still apparent

Bobi