[Jingxue Wang]

Dear all

I am using standard Smagorinsky in my LES simulation. And I also adopted the vanDriest function. The codes are following:

 LESModel        Smagorinsky;
    SmagorinskyCoeffs
    {
        Ce              1.05;
        Ck              0.07; // 0.094;
    }

    printCoeffs     on;

    delta           vanDriest;
    vanDriestCoeffs
    {
        delta           cubeRootVol;
        cubeRootVolCoeffs
        {
            deltaCoeff      1;
        }
        Aplus           26;
        Cdelta          0.158;
    }

May I understand that the subgrid model I used above is stand Smagorinsky model? How I should revise my code if I want to use Smagorinsky-Lilly model or dynamic model？I checked the codes in OpenFoam. I only found the Smagorinsky.C and Smagorinsky.H.

Many thanks for this!

[Santiago]

Judging by the input I say that you are using the static smagorinsky model.

What do you mean by "Smagorinsky-lilly or dynamic (...)"?

[Jingxue Wang]

Hi, Santiago

Thanks for your reply.

I read some papers and also find some paper used Smagrinsky-Lilly model, as shown in the attached file. And in the fluent, the choice of sub-grid models also include Smagrinsky-Lilly. I am wondering that standard Smagrinsky model and Smagrinsky-Lilly are the same? If not, what's are their diffence?

[Jingxue Wang]

Hi, Santiago

Thanks for your reply.

Because I can not manage to insert my attached file when replying to you directly, so I choose to just post my reply. Please see them in the previous reply.

[Santiago]

Originally Posted by Jingxue Wang

Hi, Santiago

Thanks for your reply.

I read some papers and also find some paper used Smagrinsky-Lilly model, as shown in the attached file. And in the fluent, the choice of sub-grid models also include Smagrinsky-Lilly. I am wondering that standard Smagrinsky model and Smagrinsky-Lilly are the same? If not, what's are their diffence?

The static and "Dynamic" (whichever flavor you choose) Smagorinsky models are NOT the same. The former assumes $C_s$ to be a constant, whereas the other(s) take the parameter $C_s=C_s(x,y,z,t)$. The so-called Smagorinsky-Lilly model is just one of the many approaches that you can use in order to minimize the error induced by the use of the Smagorinsky model in the Germano Identity. In the paper of Lilly (1992), no stabilization via averaging along homogeneous directions is made. I think that's what the models in FLUENT mean when they say "Samgorinsky-Lilly".

I don't know what is actually being used for the determination of the parameter in FLUENT, but generally it should not be of much importance as long as the model does not unstabilize your whole simulation.

[Jingxue Wang]

I understand.

Many thanks for your detailed reply.

By now, my current calculation is ok and seems stable when using my Smagrinsky model.

[HakikiCanakkaleli]

Hi,

'Dynamic' is a procedure rather than a closure model to the filtered N-S equations. OpenFOAM v1712 has two options for the dynamic procedure: dynamicKEqn and dynamicLagrangian, the latter of which uses Smagorinsky closure model with a Lagrangian-concept averaging procedure. For the 'Lilly', it has been a long time that I had completed my studies on it, so I might be misleading, Lilly attempted to designate a value of Cs universal to all types of flows through a number of assumptions and mean dissipation. Therefore, Smagorinsky model which uses Lilly's universal constant is tagged as 'Smagorinsky-Lilly' model, if my memory serves me correctly. Moreover, Smagorinsky model with the default coefficients in OpenFOAM should actually be 'Smagorinsky-Lilly' as well. You may need to check the default coefs for this deduction.

Plus, please note that there should not be static Smagorinsky model as quoted below, because the physical modelling approaches (such as LES, DES etc.) wherein Smagorinsky model is used are inherently transient. I've just wanted correct the phrase:

Originally Posted by Santiago

Judging by the input I say that you are using the static smagorinsky model.

What do you mean by "Smagorinsky-lilly or dynamic (...)"?

[Santiago]

Originally Posted by HakikiCanakkaleli

Hi,

'Dynamic' is a procedure rather than a closure model to the filtered N-S equations. OpenFOAM v1712 has two options for the dynamic procedure: dynamicKEqn and dynamicLagrangian, the latter of which uses Smagorinsky closure model with a Lagrangian-concept averaging procedure. For the 'Lilly', it has been a long time that I had completed my studies on it, so I might be misleading, Lilly attempted to designate a value of Cs universal to all types of flows through a number of assumptions and mean dissipation. Therefore, Smagorinsky model which uses Lilly's universal constant is tagged as 'Smagorinsky-Lilly' model, if my memory serves me correctly. Moreover, Smagorinsky model with the default coefficients in OpenFOAM should actually be 'Smagorinsky-Lilly' as well. You may need to check the default coefs for this deduction.

Plus, please note that there should not be static Smagorinsky model as quoted below, because the physical modelling approaches (such as LES, DES etc.) wherein Smagorinsky model is used are inherently transient. I've just wanted correct the phrase:

Some comments:

1. The term 'static', as juxtaposed to the term 'dynamic', does not refer to whether the simulation, solver, or solution fields are transient. Instead it refers to whether Cs is calculated inside the solution cycle or just set as a constant during the simulation.

2. The smagorinsky-lilly model IS DYNAMIC, as the method proposed by Lilly (1992) gives a "procedure" to calculate Cs each iteration, resulting in a field. In a following paper, using his method, he gives an average value for Cs for homogeneous isotropic flows, but that doesnt mean that he proposed a STATIC model. In fact, once upon a time there used to be a smagorinsky lilly model in openfoam, but was removed because there were inconsistencies, there is a bug report by Alberto Passalaqua on this.

So I kindly, but firmly, disagree with you

[HakikiCanakkaleli]

Hi,

Originally Posted by Santiago

Some comments:

1. The term 'static', as juxtaposed to the term 'dynamic', does not refer to whether the simulation, solver, or solution fields are transient. Instead it refers to whether Cs is calculated inside the solution cycle or just set as a constant during the simulation.

This is much clear, thank you. I thought the questioner may understand this wrongly.

Originally Posted by Santiago

Some comments:

2. The smagorinsky-lilly model IS DYNAMIC, as the method proposed by Lilly (1992) gives a "procedure" to calculate Cs each iteration, resulting in a field. In a following paper, using his method, he gives an average value for Cs for homogeneous isotropic flows, but that doesnt mean that he proposed a STATIC model. In fact, once upon a time there used to be a smagorinsky lilly model in openfoam, but was removed because there were inconsistencies, there is a bug report by Alberto Passalaqua on this.

IMHO, it seems to me that you confuse yourself with Germano-Lilly dynamic procedure as Smagorinsky-Lilly is a closure model rather than a dynamic procedure, wherein, as I mentioned above, Lilly's set of coefficients were in use in Smagorinsky model; hence, the name convention Smagorinky-Lilly.

PS: BTW, how did you make that emoji?

[Santiago]

Originally Posted by HakikiCanakkaleli

Hi,



This is much clear, thank you. I thought the questioner may understand this wrongly.



IMHO, it seems to me that you confuse yourself with Germano-Lilly dynamic procedure as Smagorinsky-Lilly is a closure model rather than a dynamic procedure, wherein, as I mentioned above, Lilly's set of coefficients were in use in Smagorinsky model; hence, the name convention Smagorinky-Lilly.

PS: BTW, how did you make that emoji?

About the emoji, i think comes from the cellphone app.

About the terminology: I have not found the method, or procedure, being called Germano-Lilly but called Smagorinsky-Lilly, in my field of work at least. We can agree on the Lilly, and the dynamic, part i guess.

[HakikiCanakkaleli]

Hi,

I think. it seems we talk about the same thing, yet with different terminologies.

Thanks

[sr71blackbird]

Originally Posted by Santiago

Some comments:

1. The term 'static', as juxtaposed to the term 'dynamic', does not refer to whether the simulation, solver, or solution fields are transient. Instead it refers to whether Cs is calculated inside the solution cycle or just set as a constant during the simulation.

2. The smagorinsky-lilly model IS DYNAMIC, as the method proposed by Lilly (1992) gives a "procedure" to calculate Cs each iteration, resulting in a field. In a following paper, using his method, he gives an average value for Cs for homogeneous isotropic flows, but that doesnt mean that he proposed a STATIC model. In fact, once upon a time there used to be a smagorinsky lilly model in openfoam, but was removed because there were inconsistencies, there is a bug report by Alberto Passalaqua on this.

So I kindly, but firmly, disagree with you

Thanks, everyone for a detailed discussion on the dynamic Smagorinsky model. I wish to run a non-reacting, turbulent particle-laden jet flow using dynamic smagorinsky SGS model and I am using OpenFOAM-5.0.x. Can anyone suggest if there are any tutorial available to guide in how to use dynamic smagorinsky ?

[cryabroad]

There are no built-in dynamic Smagorinsky in FOAM, but people have done the implementation, just search for it. I think for your case maybe the existing one in FOAM (the one with a static constant) is enough though, why bother using the dynamic one?