[roadRunner]

Dear all,

I'm somewhat experienced in using openFoam, but mostly I've been using already existing stuff so far. Now I'm looking into implementing other things, in this case turbulence model.
But before I do that, I have a question about the existing implementation of kEpsilon model.

Why are source terms treated differently? Equation:

    // Dissipation equation                                                     
    tmp<fvScalarMatrix> epsEqn
    (
         fvm::ddt(epsilon_)
      + fvm::div(phi_, epsilon_)
      - fvm::laplacian(DepsilonEff(), epsilon_)
     ==
        C1_*G*epsilon_/k_
      - fvm::Sp(C2_*epsilon_/k_, epsilon_)
    );

So, is first term treated explicitly and the second one implicitly (if I understand correctly)? Is it because of velocity gradient? If so, why? Pressure-velocity coupling?

This is maybe a silly question, but I would like to understand it.

Thanks for reading!

[roadRunner]

OK, what I know so far is that G is evaluated explicitly due to the fvc::grad:
nut_*2*magSqr(symm(fvc::grad(U_))

But the epsilon_ could still be treated implicitly? Or?

Thank you (and please help)

[ssss]

All the therms that have fvm:: are treated as implicit. What I'm not sure about is the term:

C1_*G*epsilon_/k_

My intuition says that this term is treated as explicit. You could try to change the code to:

fvm::Sp(C1_*G/k_,epsilon_)

And see if it makes any difference

[kebsiali]

Total agree with your reasoning here. Why explicit?
Maybe the epsilon generation term is taken as a function of the last epsilon
Like saying the amount of epsilon that we had is the cause of the generation.
but then why is the destruction term implicit?
maybe both were chosen like this for stability reasons?

[roadRunner]

Originally Posted by ssss

All the therms that have fvm:: are treated as implicit. What I'm not sure about is the term:

C1_*G*epsilon_/k_

My intuition says that this term is treated as explicit. You could try to change the code to:

fvm::Sp(C1_*G/k_,epsilon_)

And see if it makes any difference

Ok, but for what do I check? And how? On which case?

Originally Posted by kebsiali

Total agree with your reasoning here. Why explicit?
Maybe the epsilon generation term is taken as a function of the last epsilon
Like saying the amount of epsilon that we had is the cause of the generation.
but then why is the destruction term implicit?
maybe both were chosen like this for stability reasons?

I think stability reasons are most likely in this case, I would like to know what they are
Thanks for your answers!

[GerhardHolzinger]

Originally Posted by roadRunner

Dear all,

I'm somewhat experienced in using openFoam, but mostly I've been using already existing stuff so far. Now I'm looking into implementing other things, in this case turbulence model.
But before I do that, I have a question about the existing implementation of kEpsilon model.

Why are source terms treated differently? Equation:

    // Dissipation equation                                                     
    tmp<fvScalarMatrix> epsEqn
    (
         fvm::ddt(epsilon_)
      + fvm::div(phi_, epsilon_)
      - fvm::laplacian(DepsilonEff(), epsilon_)
     ==
        C1_*G*epsilon_/k_
      - fvm::Sp(C2_*epsilon_/k_, epsilon_)
    );

So, is first term treated explicitly and the second one implicitly (if I understand correctly)? Is it because of velocity gradient? If so, why? Pressure-velocity coupling?

This is maybe a silly question, but I would like to understand it.

Thanks for reading!

The difference between the two source terms is the sign. Both depend on epsilon, so you could in theory handle them both implicitely. However, the sign determines whether the implicit handling improves numerical behaviour or degrades it.
A negative term on the RHS is a positive term if brought to the LHS. Further, implicit terms go into the main diagonal of the system matrix of the discretized equation system. So, by treating negative source terms (on the RHS) implicitely, we add to the main diagonal of the system matrix. Thus we increase the so called diagonal dominance of the matrix which in turn helps numerical iterative solution procedures.

[roadRunner]

Thank you.