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/)
-   -   Nearwall treatment for the kOmegaSST turbulence model (http://www.cfd-online.com/Forums/openfoam-solving/58070-nearwall-treatment-komegasst-turbulence-model.html)

johnb January 20, 2009 12:58

I am currently doing vehicle a
 
I am currently doing vehicle aerodynamics simulation with OpenFoam v1.5 and using simpleFoam with the incompressible implemented kOmegaSST turbulence model. My main concern is when this model is valid in terms of y+. Since it uses wall functions I guess the y+ should be in the log-law region (30<y+<300) and hence it would be improper to have a finer mesh with lower y+ (in my case y+ ~7) together with this model?

Would it be an idea to try a Low-Re turbulence model like Launder-Sharma or QZeta, or would that require a further refined mesh (y+~1)?

Thanks in advance!

kati January 21, 2009 07:02

If you want to use low-Re mode
 
If you want to use low-Re model, you should see that all walls have y+~1 (4-5 should still be okay in some places, but 7 is quite a lot). If I were you doing vehicle aerodynamics, I would use SpalartAllmaras. It is a hybrid model, so if you select nutWallFunction as boundary condition for nut on walls (mut/mutWallFunction if compressible) you don't need to worry so much about y+, because the model uses different treatment depending on the y+ value.

The low-Re models can be tricky, I've heard. There are many of them i.e. none is very general, and epsilon equation is very sensitive to grid on the wall.

One more point with standard wall functions: the theory is valid only for flat plates (straight ducts maybe too - I don't remember), so that all pressure gradients and especially separation or attachment are not well predicted. So even if your y+ is in the correct limits, you may not get good results with wall functions.

Regards,
Kati

ivan_cozza January 21, 2009 12:22

Hi John, I'm facing the same
 
Hi John,
I'm facing the same questions, I'm trying to solve a Re 1.3e5 flat plate with a tickness equal to 3% of the chord. As I'm interested in the k and eps fields, I would like to resolve well the boundary layer with something like a low Re model, and I know that k-Omega SST as some low Re version, but I know that now in OpenFOAM there's only the wall function version...
Has anyone the low Re version for SST?

johnb January 22, 2009 02:52

Hi Kati and Ivan, Thanks for
 
Hi Kati and Ivan,
Thanks for the advice, I will try using SpalartAllmaras (and post how it goes here) and I guess you use nutWallFunction with a zero (or say 1-e10) value of nut at the wall boundaries.

What would be very interesting for the KOmegaSST model (as well as other turbulence model using wall function approach) is to implement an enhanced wall treatment. In such way I guess that a coarse mesh region will correspond to a wall function approach (zeroGradient) while a fine mesh region will use an appropriate Low-Re boundary condition (fixedValue). Does anyone know if there are plans on releasing such option?

It seams like these three posts is handling modification to the existing kOmegaSST model, very interesting:
http://www.cfd-online.com/OpenFOAM_D...tml?1231869877
http://www.cfd-online.com/OpenFOAM_D...tml?1186414688
http://www.cfd-online.com/OpenFOAM_D...tml?1220457716

Regards,
John


All times are GMT -4. The time now is 11:41.