[alberto]

Originally Posted by jiejie

Hi alberto

Thanks for your reply.

Again, does this forbid the back scattering and does not allow the energy transfer from the small scale to the large scale?

or is it possible just to remove the averaging part of the code in the dynSmagorinsky model and how bad is it in terms of stability?

I guess this returning question is a problem with the nomenclature.

The effective viscosity is nuEff = nu + nuSGS, being nu the molecular viscosity. As a consequence, clipping the effective viscosity nuEff to zero does not prevent backscattering, or all the effort of the dynamic procedure would lose at least part of its importance.

It is not possible to "directly remove the averaging", you need to implement some procedure to ensure that you do not divide by zero when computing the dynamic coefficients. There are different ways of achieving this reported in the literature. We used a very simple one, which consists in performing a local average over the neighbours. In terms of stability, it seems to work fine, and results are in agreement with what reported by other Authors.

Best,

[alberto]

Originally Posted by FMDenaro

Hi to all,
this topic is full of interesting issues for LES practioneers and researchers.

I am working on the quality and realibility of LES implemented in OF along with some collegues.
We decided to test the compressible version of the code in low-Mach conditions, performing the classic plane channel turbulent flow at Re_tau=590.
The dynamic procedure for the SGS modelling is supplied.

Testing the code in such condition was not so easy because of the requirement of adding the proper forcing term in momentum and energy equations.

At present, results obtained on a 64x100x64 cartesian grid are not satisfactory both in the zero-order statistics (the streamwise averaged profile) and in the energy spectra. This latter shows for the three velocity components along the streamwise wavenumber a very strange peak in the inertial range, transfer of energy is therefore not monotone. This is not a numerical pile-up since the peak is about 10 wavenumbers before the Nyquist wavenumber, the energy is well dissipated at the cut-off wavenumber.
The SGS model does not change the presence of the peak (no-model solution showed the same behaviour)
No other peaks appear in the energy spectra along the spanwise wavenumber.

I am quite sure that the reason is due to some acoustic interference caused by the low-Mach condition. The PISO solver should be able to manage such critical case...

Does anyone had similar experience?

Thanks
Filippo

Hi Filippo,

I think I saw part of those results, since they were in the paper Andrea sent me.
Did you try to run one case with the incompressible version of the model to see if the same behaviour is present?

Best,

[waynezw0618]

Originally Posted by jiejie

Hi waynezw0618

I have read Lilly's paper, I probably found the reference of OpenFOAM dynSamgorinsky from "A comparative study of subgrid scale models in homogeneous isotropic turbulence" by Fureby C. 1996.

It can only used for
homogeneous isotropic turbulence

[FMDenaro]

Originally Posted by alberto

Hi Filippo,

I think I saw part of those results, since they were in the paper Andrea sent me.
Did you try to run one case with the incompressible version of the model to see if the same behaviour is present?

Best,

Hi Alberto,
yes it is possible you have seen them since Andrea was asking for getting opinions from people involved in OF. At present I am coordinator of the group LESinItaly, collecting activities from the Politecnico of Milano, Torino, Unvisersity of Pisa/Udine and Napoli. We have a project for testing several CFD/LES codes on the turbulent channel flow.

With Andrea just decided a couple of days ago to repeat the same simulation with the incompressible version of the code and see what happens in the spectra. We have now to run and wait for the solution to get an energy equilibrium state and sample the fields for the statistics.
Some time is therefore required....


However, the compressible code gave no peaks in the energy spectra at Re_tau=180. But the lenght of the longitudinal domain is wider, maybe permits to dissipate acoustic waves, also because the dissipation is greater....

Do you have some experience of spurious solutions with the compressible version of PISO in low-Mach flows??

Thanks
Filippo

[FMDenaro]

Originally Posted by waynezw0618

It can only used for
homogeneous isotropic turbulence

Hi, perhaps if the filter lenght is ensured to lie in the inertial region for all scales, it is common to suppose that SGS scales have a quite universal (and homogeneous) behaviour ....
Often the HIT is at basis of several SGS models used for non-homogenous flows...
But this is one of the controversial issues in LES ...

[mmahdinia]

Hi,

Does anyone has a journal paper for "clipping" of effective viscosity such that (nuSGS+nu)>0 in dynamic LES mehod?

Sincerely,
Maani

[FMDenaro]

Originally Posted by mmahdinia

Hi,

Does anyone has a journal paper for "clipping" of effective viscosity such that (nuSGS+nu)>0 in dynamic LES mehod?

Sincerely,
Maani

Hi, yes, you can see my paper on TCFD where we discussed about the clipping criteria.

F.M. Denaro, G. De Stefano, A new development of the dynamic procedure in large-eddy simulation based on a Finite Volume integral approach. Application to stratified turbulence. Theor. Comp. Fluid Dyn., online DOI10.1007/s00162-010-0202-x, 2010

[mmahdinia]

Thanks Filippo

Please excuse me. But also I wanted to know if anyone has used QUICK scheme in LES despite it is too dissipative.

According to Mittal and Moin (1997) the QUICK is too dissipative for reactive flows and generated noise flows and there central schemes should be used. But for ordinary flows it is nearly sufficient.

I would appreciate very much it if anyone could tell me about any resent journal articles in which QUICK (or B-QUICK) is used.

Sincerely,
Maani

[FMDenaro]

Originally Posted by mmahdinia

Thanks

Please excuse me. But also I wanted to know if anyone has used QUICK scheme in LES despite it is too dissipative.

According to Mittal and Moin (1997) the QUICK is too dissipative for reactive flows and generated noise flows and there central schemes should be used. But for ordinary flows it is nearly sufficient.

I would appreciate very much it if anyone could tell me about any resent journal articles in which QUICK (or B-QUICK) is used.

Sincerely,
Maani

There are many argumentations about using upwind schemes, I also published a paper on JCP regarding their using ...
However:
1) QUICK is not suitable at all for unsteady flows, it was developed for steady flows, therefore is not only a problem of using it in LES. The unsteady version was developed by Leonard, called QUICKEST, and it is quite different having a time-intagration built-in inside.
2) More in general, upwind schemes are well-suited for DNS since their artificial dissipation lie in the part of the spectrum where the molecular dissipation is acted on. But in LES, the action of the numerical dissipation acts on the resolved scales, prevalently in the highest part of the resolved spectrum, therefore even if you provide an "ideal" perfect SGS model it would not work properly since of the artificial smoothing of the resolved components.
3) However, some high order (more than third order) upwind schemes implemented in the FV framework, along with some suitable recovering of the smoothed components (deconvolution) can be used. Mittal and Moin showed unsatisfactory results by using FD upwind. You can find on my paper on JCP the discussion about the differences

Regards
Filippo

[alberto]

Originally Posted by FMDenaro

Do you have some experience of spurious solutions with the compressible version of PISO in low-Mach flows??

Not really. I used it for some application in a complex geometry and did not notice anything strange.

Best,

[mdjames]

With all this talk of openFOAM's unfaithful implementation of dynSmag, it seems a good place to ask what the code is using the supplied coefficient for

i.e. in CASE/constant/LESProperties :

dynSmagorinskyCoeffs
{
filter simple;
ce 1.048;
}

is ce used to define the second filtering size (~ in Germano 1991) or what? I've looked through dynSmagorinsky.C for the answer but can't seem to find "ce" even used anywhere.

Also, I suspect it's not a trivial endeavor to change the code to average over horizontal planes rather than the entire domain?

[alberto]

Hello,

Originally Posted by mdjames

is ce used to define the second filtering size (~ in Germano 1991) or what? I've looked through dynSmagorinsky.C for the answer but can't seem to find "ce" even used anywhere.

Ce is not Cs. Ce is used to compute the dissipation rate.

Also, I suspect it's not a trivial endeavor to change the code to average over horizontal planes rather than the entire domain?

In a general geometry it is not appropriate to average along planes either. You can average locally however.

Best,

[mdjames]

Originally Posted by alberto

Ce is not Cs. Ce is used to compute the dissipation rate.

RIght, Cs is determined by the algorithm. I was unsure of what you meant by dissipation rate, but it appears as though we're talking about the same thing in different ways. I wondered if Ce is just the ratio of the coarse filter to the grid-sized filter. (Ce=delta_tilde/delta_bar). After some more reading it appears that this is indeed what Ce adjusts.

In a general geometry it is not appropriate to average along planes either. You can average locally however.

I agree and should have been more specific. It is appropriate, however, to average along planes parallel to a wall; no?

Thanks very much for your speedy and helpful response, as usual.

[FMDenaro]

Hi,
just coming back to this topic in order to acquire further insight on the dynamic procedure implemented in the compressible version of OF.
In this case, the computed eddy viscosity is local, not averaged on the whole volume, that's right?
It is allowed to have some negative value (corresponding to an energy back-scatter) or the value has a clipping?
Many thanks

Filippo

[alberto]

Hi Filippo (and happy holiday!),

there is no dynamic Smagorinsky model for compressible flow implemented in OpenFOAM currently, just the standard Smagorinsky for compressible flows. There is the dynOneEqEddy model, however, which averages the coefficients on the whole domain.

Notice that nuSgs is local also in the dynSmagorinsky implementation in OpenFOAM, since it depends on the local value of the magnitude of the filtered deformation rate tensor. What is not local, but averaged, is the coefficient Cs, which should be local instead.

Best,

[FMDenaro]

Originally Posted by alberto

Hi Filippo (and happy holiday!),

there is no dynamic Smagorinsky model for compressible flow implemented in OpenFOAM currently, just the standard Smagorinsky for compressible flows. There is the dynOneEqEddy model, however, which averages the coefficients on the whole domain.

Notice that nuSgs is local also in the dynSmagorinsky implementation in OpenFOAM, since it depends on the local value of the magnitude of the filtered deformation rate tensor. What is not local, but averaged, is the coefficient Cs, which should be local instead.

Best,

Hi Alberto,

thanks for your reply,

I wish you a Merry Christmas

Filippo

[tehache]

talking about dynamic smagorinsky reminded me of this this thread:


http://www.cfd-online.com/Forums/ope...agorinsky.html

[grandgo]

hello!

im working on the tutorial pitzDaily (incompressible/pisoFoam) in LES mode with the dynSmagorinsky model and i have a question:

if i'm right, i only need the p, U and k files for this case. according to dynSmagorinsky.C . but with deleting the nusgs file in folder 0 i get the error message which tells me, that nusgs file is missing.

why??

best regards
grandgo

[fs82]

This is easy to answer. Please have a look to dynSmagorinsky.C:

void dynSmagorinsky::updateSubGridScaleFields(const volSymmTensorField& D)
{
    nuSgs_ = cD(D)*sqr(delta())*sqrt(magSqr(D));
    nuSgs_.correctBoundaryConditions();
}

The information about the boundary condition is needed in nuSgs_.correctBoundaryConditions(); and thus the field nuSgs is needed at start. Also you could use the boundary conditions for nuSgs to specify a wall function for a specific boundary. Thats he reason why he is complaining about the missing nuSgs file.

[grandgo]

Originally Posted by fs82

This is easy to answer. Please have a look to dynSmagorinsky.C:

void dynSmagorinsky::updateSubGridScaleFields(const volSymmTensorField& D)
{
    nuSgs_ = cD(D)*sqr(delta())*sqrt(magSqr(D));
    nuSgs_.correctBoundaryConditions();
}

The information about the boundary condition is needed in nuSgs_.correctBoundaryConditions(); and thus the field nuSgs is needed at start. Also you could use the boundary conditions for nuSgs to specify a wall function for a specific boundary. Thats he reason why he is complaining about the missing nuSgs file.

thanks fabian!