CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   "Wallscale" variable with SST (http://www.cfd-online.com/Forums/cfx/19112-wallscale-variable-sst.html)

Louwrens August 8, 2002 03:41

"Wallscale" variable with SST
 
What does the wallscale variable mean that's plotted when using the SST turbulence model with 5.5? Why is it only computed for the first few iterations? How critical is its convergence for the accuracy of the solution?

I'm modelling flow over a rankine halfbody, and ending up with a wallscale global imbalance around 5% where all momentum and mass imbalances are at 0%...

Cheers

Louwrens

Robin August 8, 2002 09:10

Re: "Wallscale" variable with SST
 
Hi Louwrens,

The wall scale is not a conserved quantity, so don't worry about it not balancing. It is simply used to provide the SST model with a distance from the wall in order for it to know how to blend from a k-omega formulation near the wall to a k-e formulation in the free-stream.

It is a simple Poisson Equation and is only calculated for a few iterations because it is not flow dependant (only geometry). A few iterations is all that is needed to get it converged. If you wish, you can modify the number of calculations by using the appropriate expert parameters.

Robin

Louwrens August 9, 2002 09:17

Re: "Wallscale" variable with SST
 
Thanx Robin

On the topic of the SST model: From the manual and Menter's initial paper, its supposed to be more accurate than k-e and standard k-w models, but from validation cases I've run, I get the same (or better!) results with the k-e model. It almost seems as if the sst model is more dependent on a properly resolved boundary layer, which makes things difficult for me, since I use inflation with my first grid point at y+=11 (at laminar - log interface) and going even finer would probably cause problems. I know its forte is in flow with negative pressure gradients, and predicting separation (which will be my last validation test) but for pressure drop and friction factors (both pipe and flat plate flow) as well as pressure distributions over a rankine halfbody, the k-e model gave consitently better results. Any thoughts and / or suggestions?

Louwrens

Robin August 9, 2002 10:39

Re: "Wallscale" variable with SST
 
Hi Louwrens,

If you do not create an appropriate mesh, you will not see the benefits of SST. For the model to perform properly, you should have a near wall resolution of y+ <= 2.

The main feature of the SST model is the ability to integrate through the boundary layer to the wall, rather than use a wall function. Since it is not always possible to create a fine mesh in the boundary layer, the automatic wall treatment will blend-in a wall function as the y+ values get larger, in the end behaving much like a standard k-e model. The automatic near wall treatment thus allows you to refine the mesh where you need it and have a wall function treatment in less critical areas.

Regarding your comment, I use inflation with my first grid point at y+=11 (at laminar - log interface) and going even finer would probably cause problems, this is not the case for any of the k-omega based models (SST, BSL, etc) since there are no wall functions, or any of the other turbulence models when the scalable or automatic wall function is used. If you are using k-e with the scalable wall functions, it is insensitive to y+ being too small. So the point is: don't be afraid to inflate closer to the wall.

Best regards, Robin


Louwrens August 9, 2002 11:04

Re: "Wallscale" variable with SST
 
OK... lots of info to digest, probably gonna need the weekend and a few beers for that! ;-)

A near wall resolution of y+<=2 basically implies using the sublayer model, as explained in the solver manual (pp280-298) (Hmm... if I read more accurately I would've seen "The k-w models do not accept coarser grids, due to the automatic near-wall treatment for these models." -p.295- moral: RTFM CAREFULLY!)

Still, at the moment y+=11 lies in the order of 0.1mm and y+=2 would be at 0.008mm... which brings up the problem of high AR elements, and very fine surface meshes to avoid this. From a point of view of computational time vs. improvement in results gained, would this really be worth it?

TGIF

Louwrens

Robin August 9, 2002 11:19

Re: "Wallscale" variable with SST
 
Hi Louwrens,

High aspect ratio elements may be a problem in the free-stream where the flow is not aligned with the elements. If the flow is aligned with the elements, such as near the wall or in a duct etc., you need not worry. Stretch away!

Regards, Robin


All times are GMT -4. The time now is 08:19.