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/)
-   -   Running without wall function (http://www.cfd-online.com/Forums/openfoam-solving/113088-running-without-wall-function.html)

batta31 February 12, 2013 05:38

Running without wall function
 
Hi to everyone! I'd like to run a case without wall function and the Spalart-Allmaras turbulence model. Searching around I didn't really figured out what to do regarding the boundary condition on the solid body for nut and nuTilda. I suppose this setting:

nut:
solidBody
{
type nutLowReWallFunction;
value uniform 0;
}

nuTilda:
solidBody
{
type fixedValue;
value uniform 0;
}

Am I correct or should I use also fixedValue for nut?
Any answer is really, really appreciated!

Cheers
Simone

batta31 February 12, 2013 06:26

Using a dummy turbulence model I found out the the availlable turbulence models are:

(
LRR
LamBremhorstKE
LaunderGibsonRSTM
LaunderSharmaKE
LienCubicKE
LienCubicKELowRe
LienLeschzinerLowRe
NonlinearKEShih
RNGkEpsilon
SpalartAllmaras
kEpsilon
kOmega
kOmegaSST
kkLOmega
laminar
qZeta
realizableKE
)

So maybe only those two schemes can be used without wall functions? In the sense that I have to use wall function with SpalartAllmaras model?

chegdan February 12, 2013 17:43

There are other turbulence models that can be used without wall functions, for example LamBremhorst. Similar to high Re turbulence modles, low Re turbulence models requirements for y+ in order to be valid....so more stringent than others. I believe that LaunderSharmaKE has one near y+<=1 and LamBremhorst is about a y+ of 5. You might want to search the forum for the y+ values, and or read through Wilcox's book on turbulence modeling.

Lastly, you do not need use wall functions for the SpalartAllamaral. Some may argue what the BC's for SpalartAllmaras should be, but from a phenomenalogical perspective...zeroGradient works because there is diffusion of turbulent viscosity down through the turbulent boundary layer in course meshes. Ideally, it should be zero once it reaches the wall (if the turbulent boundary layer is assumed to be maintained and if the mesh is fine enough). i think one may run into stability problems with a nuTilda= 1e-20 (don't use zero exactly) at the wall. try the thread http://www.cfd-online.com/Forums/ope...us-values.html for some more information. Good Luck

batta31 February 13, 2013 03:22

Hi chegdan and thank you for your answer..
So you are suggesting me to use a setup like:

nut:
solidBody
{
type nutLowReWallFunction;
value uniform 0;
}

nuTilda:
solidBody
{
type fixedValue;
value uniform 1e-20;
}

And for the inlet/outlet and the other patch for the farfiel? Is it correct to use freestream BC? With a value that is a fraction of viscosity?

I read that for nut one should use about 0.1nu and for nuTilda 0.3nu.

Cheers
Simone

chegdan February 14, 2013 11:22

For nut wall, you can use the Low Re wall function like you have stated. It sets nut to zero and provides access to calculate a y+. For SA model, use waht you have , but you can also try zeroGradient to see if the results are closer to any validation results.

For inlet values, you can set nut to type calculated and then define nutTilda with either what you have or

\tilde{\nu}=\sqrt{\frac{3}{2}}||\vec{U}||Il

where ||\vec{U}|| is the inlet velocity magnitude, I is the turbulent intensity, and l is the turbulent length scale, see here. Making your own SA inlet BC to do this for you is relatively easy, you can derive one from the k or epsilon BCs dealing with defining scalar values based on turbulent intensities.


All times are GMT -4. The time now is 01:55.