[tareqkh]

Hello All,

I am doing a validation study on a backward-facing step and trying to implement k-epsilon with proper boundary conditions. The dynamic viscosity for the problem is 1.96*10^-6 N.s/m2, Re 5100, density 1kg/m3, Ux=1m/s, Uy=0. I have made the proper mesh according to y+ NASA calculator. I have 29000 number of nodes for the mesh. Everything is being step properly. Now, I want to implement k-epsilon with a give value of k=0.015m2/s2 and turbulence intensity of 10%. The aim of the study is to compare my data with the reattachment point. I have tried to set up k and epsilon properly. However, the results do not seem reasonable. I am attaching both k and epsilon boundary conditions and would be more than happy if someone of you could help in this regards.

Boundary condtions.zip

[icaroaccordi]

First of all, it seems that you are using the standard k-e. Am I correct? However you are using bc's for lowRe k-e. If you are using y+ <= 1 you need to choose some low-Re model (e.g. LaunderSharma_ke). If your y+ is greater than 20 you can use standard k-e with the wall functions that you have commented.

[tareqkh]

Thank you for your reply. My question is about whether or not boundary conditions at the top and bottom wall is being set up properly.

[icaroaccordi]

I only can reply if I know what kind of k-e you are using. If you are using standard k-e they are completely wrong because you are using low-Re bc's.

[tareqkh]

I am using the files that are provided in the pitzily example in the simplefoam. Since I am using y+, I need to change the top and bottom boundary to a fixed wall. If you could guid me in this regard, that would be nice.

[icaroaccordi]

Do you have a yPlus target?

[tareqkh]

Yes, it is 0.0002529.

[icaroaccordi]

First of all you need to change to LaunderSharma k-e because the standard cannot work as a Low-Re model.

[tareqkh]

Do you know any examples in the simplefoam uses this turbulence model. Or can you provide me the boundary for this type?

[icaroaccordi]

The only example that I know in OpenFoam is compressible and the Lunder Sharma is working as a high-Re model (you can use a low-Re model as a high-Re but not the oposite).
You can use at wall:
nut = 0;
k = 0;
e = zeroGradient;

But the best would be use the new bc's of 2.3.versions:

for k:
myWall
{
type kLowReWallFunction;
}

for e:
myWall
{
type epsilonLowReWallFunction;
}

for nut:
myWall
{
type nutLowReWallFunction;
}

These are good because they are automatic and can work as wall-functions or resolving the boundary layer in agreement with your mesh.

Why so litle y+. The target usualy is less than 1 but not so less.

[icaroaccordi]

If you want to take a look at the example is compressible/sonicFoam/RAS/nacaAirfoil.
Pleas do not use the bc's of this example.

If you want quick information about low-Re models pleas take a look at:
http://www.cfd-online.com/Wiki/Low-Re_k-epsilon_models

[tareqkh]

Can you send me your 0 file for this turbulence if you could?

[JasonWang3]

Originally Posted by icaroaccordi

If you want to take a look at the example is compressible/sonicFoam/RAS/nacaAirfoil.
Pleas do not use the bc's of this example.

If you want quick information about low-Re models pleas take a look at:
http://www.cfd-online.com/Wiki/Low-Re_k-epsilon_models

Hi

I generate two mesh, y+ near 1 and y+ near 20. When I use the y+=1 grid, I choose Lauder-sharma with kLowReWallFunction, while the coarse grid, use k-epsilon with kqRWallFunction.

My question is whether kLowReWallFunction is needing when applying Low-Re turbulent model? In most case, wall functions are designed for High-Re turbulence model. And what is the difference of this two wall functions?

Regards,
Jason

[icaroaccordi]

Actually you do not need the kLowReWallFunction but it is a good idea. kLowReWallFunction is a automatic function that uses different approaches to calculate the turbulent quantities according with the local y+. It can be used by low and high-Re models. Using it with a low-Re model you can have more flexibility when dealing with different y+.
How I have said before you can just use:
nut = 0;
k = 0;
e = zeroGradient;

nutLowReWallFunction is simpler than epsilonLowReWallFunction and kLowReWallFunction and only works in low-Re. It has only the function to gives you access to the y+ of your case. The yPlusRAS application do not work when you uses fixedValue nut=0. To have a completely automatic setting of wall functions it is better to use nutUSpaldingWallFunction instead of nutLowReWallFunction.

[icaroaccordi]

omegaWallFunction is already a automatic wall function that is why there is no omegaLowReWallFunction.

[RLFerreira]

Hello Foamers, I am using 2.2.2 openFoam version to simulate flow through radial diffuser with LaunderSharma k-epsilon model, because Reynolds number is close to 3000. The case files are attached below. All my attempts diverged before 5 iteractions. I used for domain walls (patch names: walls and reed-wall) nutLowReWallFunction, epsilonLowReFunction and kLowReWallFunctions for nut, epsilon and k, respectively. Besides of that, I did a lot of others attempts without success, including the model with wall functions disable.

I really appreciate your help!

Best Regards,
Ricardo.

[icaroaccordi]

Hi Ricardo,

There is a miss concept in your question. The use of low-Re model is not exactly about the Reynolds number of the flow but about the y+ of the first element near wall. Of course that for high-Re flows you are going to need a finer mesh owing to the narrow boundary layer. Please check if there is recirculation at the exit and make it longer if necessary.
checkMesh to see if everything is OK.
If the angle is grater than 80 it is a waste of time run it.
Change all divSchemes:
- bounded Gauss limitedLinearV 1;
- bounded Gauss limitedLinear 1;
to bounded Gauss upwind;
You can come back to the old settings after some iterations. I would use linearUpwind but limitedLinear is OK as well.
Change U outlet from zeroGradient to inletOutlet.

It should work.

[RLFerreira]

Hi Icaro, thank you for your quick answer! I am grateful about your highlight in my error in concept in using low-Re model, as it made me study carefully about this subject.
Is well-known in the literature that RNG k-epsilon without wall functions is the best RANS model to deal with radial diffuser problem. So, I have some questions, if you can help me a little bit more.
1) Using RNG k-epsilon, I noted that convergence is easier using wall functions than without one. As it (wall functions) is implemented in OF, I would be missing to use it in flow regions where yPlus<1?
2) As wall boundary condition for nut, the "calculated" is the same of "nutLowReWallFunction"?
3) As wall boundary condition for k and epsilon, kLowReWallFunction and epsilonLowReWallFunction works for both low and high Reynolds setup? If yes, I prefer always use it.

Thank you very much!

[fabian_roesler]

Originally Posted by icaroaccordi

The only example that I know in OpenFoam is compressible and the Lunder Sharma is working as a high-Re model (you can use a low-Re model as a high-Re but not the oposite).
You can use at wall:
nut = 0;
k = 0;
e = zeroGradient;

But the best would be use the new bc's of 2.3.versions:

for k:
myWall
{
type kLowReWallFunction;
}

for e:
myWall
{
type epsilonLowReWallFunction;
}

for nut:
myWall
{
type nutLowReWallFunction;
}

These are good because they are automatic and can work as wall-functions or resolving the boundary layer in agreement with your mesh.

Why so litle y+. The target usualy is less than 1 but not so less.

I'd like to rise a small discussion on the kLowReWallFunction:

As far as I know this wall function was designed for v2f model. It gives a fixedValue k=0 for y+ well below yPlusLam. However, when you are in the buffer or log region, the values will be wrong for all models except v2f.
I'd appreciate a small confirmation to this statement

Cheers

Fabian

[davibarreira]

Hey Fabian, this thread http://www.cfd-online.com/Forums/ope...ion-2-2-a.html seems to confirm what you are saying. Right now, I'm reading this paper http://web.stanford.edu/group/uq/pdf...als/jcp_wf.pdf together with OpenFOAM code. If I get to any conclusions, I will report.

[edited]

I didn't get to a final conclusion, but I think that it is indeed only for v2f. The v2f.h file says:

Description
    Lien and Kalitzin's v2-f turbulence model for incompressible flows, with
    a limit imposed on the turbulent viscosity given by Davidson et al.
    The model solves for turbulence k and epsilon, with additional equations
    for the turbulence stress normal to streamlines, v2, and elliptic damping
    function, f.  The variant implemented employs N=6, such that f=0 on walls.
    Wall boundary conditions are:
        k       = kLowReWallFunction
        epsilon = epsilonLowReWallFunction
        v2      = v2WallFunction
        f       = fWallFunction
    These are applicable to both low- and high-Reynolds number flows.