Spanwise Length and Grid Refinement for SAS-SST, DES, and LES
 

Good day, all -

The CFX documentation claims that symmetry and periodicity conditions are either impossible or impractical for SAS-SST, DES, and LES simulations since the asymmetrical, 3D turbulent structures must be resolved. Furthermore, the geometry of interest must be 3D and be refined in the spanwise direction in order to sufficiently capture, at the very least, the largest turbulent structures.

Despite this, I have read countless papers from DES and LES experts that have performed 2D or quasi-3D (i.e., a small span with minimal refinement, such as an airfoil of 0.1*chord span with 4 spanwise nodes) simulations and obtained excellent results.

1) Can the SAS-SST, DES, and LES models be used for some 2D simulations, such as low Reynolds number flow over an airfoil?

2) Can quasi-3D simulations be effective in some cases?

3) If a 3D simulation is required for the case of an airfoil in low Re flow (for example), does the airfoil have to stand alone (i.e., have a finite span that doesn't touch the sidewalls) or can it extend to the sidewalls and include a periodic condition?

4) CFX documentation notes that the SAS-SST model follows "most" of the setup guidelines of the DES model (e.g., y+ < 1). However, are there some guidelines it doesn't follow? For example, DES is susceptible to grid-induced separation but SAS is not.

Thanks for any and all help. I am very interested in the SAS-SST model but our computational resources are limited. I'm just wondering if the guidelines are as stringent for SAS as they are for DES. I can run a 3D simulation, but I'd prefer minimal spanwise resolution, if possible.

I agree that symmetry boundaries are not appropriate for LES (and related model) runs. But periodic boundaries are very commonly used and providing your domain is wide enough to describe the full turbulence length scales it should be good enough for most cases.

You cannot run LES (or the others) 2D. They are inherently 3D and if you constrain them to 2D you will get rubbish results.

As stated above, as long as your domain is wide enough to capture the largest turbulence length scales (or maybe a small multiple of it, need to do a sensitivity check to find out) you should be OK. At least to engineering accuracy anyway.

I cannot help you with details on the setup of DES and SAS, sorry. I don't have much experience there.

Thanks so much, Glenn. You've been more than helpful.

Others are welcome to chime in, as well. I know the theory behind the models as I've read the papers they were derived in. However, I'm unfamiliar with the practicalities, so any and all input is welcome and appreciated.

Hi
 

I have done a study on the effect of span width & span wise resolution for DES on thick airfoil sections. We have presented it in European Wind Energy Conference-2009. I guess it might be helpful to you, you can find the paper in
EWEC-2009 conference website.:)

Title "INVESTIGATION OF GRID RESOLUTION REQUIREMENTS FOR DETACHED EDDY SIMULATION OF FLOW AROUND THICK AIRFOIL SECTIONS".

Thanks, Pavitran! Looks like a good read.

Anyone had any experience with this?
In particular I'm trying to model a compression ramp flow (20degrees) at Mach 2.85 using the SST-SAS model.

An update from Josh (via email):
"It seems that LES, DES, and SAS all require similar resolution. I've read reports that have domains as small as 0.1c in span with only 4 periodic nodes of refinement. Typically, the domain tends to be 0.1c with z+ < 20, from what I've seen."

re: the 0.1c span: "It seemed to be a recurring span for DES/LES methods in my field (low Reynolds number external aerodynamics). I forget where I read that resolution requirements for the DES and SAS studies were similar. That's only from what I've read, though. I've never validated any DES/SAS/LES."