CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   Wall treatment with geometrical restriction (http://www.cfd-online.com/Forums/openfoam-solving/112509-wall-treatment-geometrical-restriction.html)

-mAx- January 30, 2013 07:48

[Solved]-Wall treatment with geometrical restriction
 
1 Attachment(s)
Hello,
I am turrning around since a week for wall treatment.
My geometries don't allow me keeping first cell's y+ >30.
If I generate BL with first cell's height computed for y+ =30, then I don't have enough cells for resolving flowfield in smallest section
I tried to switch on Low Re turb. model (Lauder-Sharma) for resolving the BL, but I get awful convergence. And I am doubtfully using Low-Re model for my applications (incompressible, simpleFoam with Reynolds-range between 2000-6000
Despite the y+ theory, I get acceptable results with realizable-ke and default wall function (with yPlusRAS giving me min & max y+ respectively 0.5 and 30)
Results (drop pressure) are confirmed with test
Any advice from anyone?

PS: I post picture of my mesh in minimal section (y+~1)
Attachment 18629

-mAx- February 4, 2013 03:56

Problem solved by using k-Omega SST model with nutUSpaldingWallFunction on fine grid.
Then k and omega were set with uniform value 1e-10 instead of zeroGradient
Once model converged yPlusRAS utility gave me 0.005 and 2.19 for min and max on walls.

immortality February 4, 2013 05:49

hi.congratiolation for solving the problem;)
How much should yplus be at walls for kOmegaSST?
Could you introduce me an article about this model?

-mAx- February 4, 2013 06:45

in my case y+ goes from 0.005 till 2.2 (results from yPlusRAS)
I don't have article about this, but I searched a lot in the forum.
For example: http://www.cfd-online.com/Forums/ope...megasst-3.html

immortality February 4, 2013 07:58

thanks.then how you are sure about domain of y+ that is suitable or not?is there any appropriate values near wall?

-mAx- February 4, 2013 09:29

what I understood: this wall treatment switchs automatically between lowRe and standard wall function dependantly on your local y+
In my case, I build my mesh with y+~1

immortality February 4, 2013 10:41

thanks.could you send me your turbulency folders?

-mAx- February 5, 2013 06:51

/constant/RASProperties
RASModel kOmegaSST;
turbulence on;

/0/k
inlet
{
type fixedValue;
value uniform 0.006;
}
walls
{
type fixedValue;
value uniform 1e-10;
}
outlet
{
type zeroGradient;
}


/0/omega
inlet
{
type fixedValue;
value uniform 27.2;
}
outlet
{
type zeroGradient;
}
walls
{
type fixedValue;
value uniform 1e-10;
}

/0/nut
inlet
{
type calculated;
value uniform 0;
}
outlet
{
type calculated;
value uniform 0;
}
walls
{
type nutUSpaldingWallFunction;
value uniform 0;
}

Gwenael February 11, 2013 04:42

Hi -mAx-,

I have just a question about the behaviour of your flow when the y+~0.005. Usually the y+ rules for SST is y+~1 but in practice, it is very difficult to have on every walls the y+~1. Did you see some strange behaviour of your flow when y+ goes down, y+~0.005 ?

Thanks

-mAx- February 11, 2013 05:59

Bonjour Gwenael,
I don't think the problem may come from y+~0.005, since it is << 1.
But the problem may occures if you have cells with y+~30.
But as far as I read, then you can use nutUSpaldingWallFunction which selects automatically the right wall function depedantly on your y+.
In my case, I know that I have always y+~1 (or <<1, but never ~30), so I enforced (I think) solving directly sublayer by setting k and omega as 1e-10 at walls (instead of zeroGradient)
Bonne journee

fumiya March 9, 2013 23:54

B.C. for omega @wall
 
Hi -mAx-,

I understand that the appropriate boundary condition for omega is the "omegaWallFunction".

Did you get good results by fixing omega to very small values(1e-10) at the walls instead of using the wall function?

Best regards,
Fumiya

-mAx- March 11, 2013 07:14

Hello fumiya,
I set k and omega with very low value, since I know that my y+ are below 1. That's why I used nutUSpaldingWallFunction and not omegaWallFunction.
And I got good results

hayes March 13, 2013 09:25

Hi -mAx-

so I understand this methods works only if you can ensure having y+~1 or y+<1 everywhere.

Otherwise with higher y+ you would have to use continuous wallfunctions for k and omega in combination with the nutUSpaldingWallFunction. I am being correct?

Best regards,
Chris

-mAx- March 13, 2013 09:59

Hello Chris,
I would rather say, with y+ >>1 you may use nutUSpaldingWallFunction, but k and omaga at walls should be zeroGradient.
Turbulence gurus may correct me

hayes March 13, 2013 10:58

Hi -mAx-
yes I agree wallfunctions really being for a higher y+ range.

But lets consider your case where you have walls with y+~1 and y+<<1 and now in addition also walls with y+ going up to 50. I think using nutUSpaldingWallFunction with continueous k and omega wallfunctions could be beneficial with varying y+ at walls.

-mAx- March 14, 2013 02:59

Yes, if I have also walls with y+ up to 50, then I would use nutUSpaldingWallFunction with k and omega set as zeroGradient.
I set both them to very small value (not zeroGradient), if I am sure that max y+ at walls is O(1)

RodriguezFatz January 7, 2014 09:24

Hi Maxime,
Are you still sure about your boundary conditions? As far as I know omega goes to infinity at the walls. But you set it to zero. Why did you do so? I see you also set a wall function for nut, which will overwrite the nut at the boundary anyways, but setting a wrong boundary condition for omega will give wrong results in the whole domain !?

-mAx- January 7, 2014 10:19

I am pretty sure I have mixed settings for epsilon in Low-Re Model with settings for omega in k-omega SST.
Thanks for pointing me this out!


All times are GMT -4. The time now is 19:41.