CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Spalart Allmaras (http://www.cfd-online.com/Forums/main/15343-spalart-allmaras.html)

CFD - NewBe June 26, 2008 07:59

Spalart Allmaras
 
Hey buddys,

got a question to turbulence modeling in general.

There are the several ways of dealing with turbulence. Lets talk about RANS. The T-Models try to predict the turbulent shear stress. My question is, do they try to predict the elements of the stress tensor over the whole boundary layer thickness or do they start to be valid after a special wall distance? And does this gap been closed by the logarithmic wall law in general?

I'm specially interested in Spalart Allmaras, because the damping functions, specially for the production term of the viscosity, are nearly zero till y+<=4, which means that there is no production in viscosity or that the logarithmic wall law is valid there instead.

I'm wee bit confused. Can somebody help me out my confusion, please?


CFD - NewBe June 26, 2008 08:13

Re: Spalart Allmaras
 
...right now I proved myself wrong. The logarithmic wall law is valid for y+ > 30. But what about the gap between 5<y+<30? Is the SA-Model valid till the wall?...

Praveen. C June 26, 2008 09:07

Re: Spalart Allmaras
 
SA model can be integrated up to the wall. You will need a grid of y+ = O(1).

Paolo Lampitella June 26, 2008 17:08

Re: Spalart Allmaras
 
As stated by Praveen the SA model has Low-Re capability and can be used down to the wall (if the grid is fine enough to let the model works properly)

However there is no attempt to predict the stress tensor elements (this is what happens in 2nd order models like "Reynolds-stress models") because this is a so-called "turbulent viscosity model"

Also the K-epsilon model has a low-Re extension which works properly down to the wall (again with the right conditions) and maybe there are some few other models with this capability.

But, actually, the biggest part of the models doesn't have this feature and there is no reason (if not totally wrong) to have any grid point inside the viscous sub-layer. Usually the first grid point is put inside or at the end of the log-layer and a properly chosen boundary condition is selected (which uses the log-law, so this approach is only valid where this law is valid) to let everything works fine enough (that is to have the right stress at the wall).

Even with the LES approach, some very popular models have the same problems at the wall (e.g. smagorinsky) if not corrected

I hope this is what you were looking for

CFD - NewBe June 27, 2008 02:15

Re: Spalart Allmaras
 
Thanks Paolo for this comprehensive answer.

I understand the difference.


All times are GMT -4. The time now is 20:40.