[A.Devesa]

Good morning,

i'd like to ask following questions about LES in Foam:

1) forces and turbulentInlet: using forceCoeffs combined with a turbulentInlet implies huge oscillations (order of magnitude 1) of the drag coefficient of a body, even sitting far enough from the inflow, so that the inlet perturbations do not reach it.

2) directMappedPatch and parallelism: could someone inform me how to use this BC in parallel? I tried special decompositions, so that both inlet and mapped patch remain on the same processor, without success...

3) LES and numerical schemes: are there some more better schemes than linear to perform a least dissipative LES discretization?

4) LES and SGS-models: what would be your recommendation to approach the best results for an external turbulent flow (e.g. around an Ahmed body, an airfoil...): oneEqEddy, Smagorinsky, ...?

Looking forward to reading your answers and advices!
Antoine

[cedric_duprat]

Hi Antoine,

2- What's the problem with this BC ? I did parallel computations using this inflow BC and, it worked fine. however, did you check your boundary condition after the decomposition step ? What I have done with OF 1.4.1 is that I defined by hand the inflow BC after the decomposition process.

3- There is (at least was) a fourth order centred scheme but, usually second order is enough. Notice that OF is an unstructured mesh solver (even is you have a structured one). And High order scheme in unstructured mesh is still no basic.

4- oneEqEddy is supposed to be better in flow with large scale (which will be your case) and "poor" mesh resolution. By poor don't mean a RANS mesh but a coarser mesh compare to a fine LES one. However, in external flow, the DES method family (DES, DDES, IDES, ...) could be a good way to avoid LES cost.

I hope this will help you

Cedric

[A.Devesa]

Thank you Cedric for your answers. It allows me to think that i'm not quite doing everything wrong, since I'm already using the 2nd order centered scheme, and modeling Sgs through onEqEddy.This clears at least points 3- and 4-, since i'd like to achieve a LES before switching to DES...

Concerning 2-, the simulation just stops without warning or error, just after "starting time loop". Anyway, thanks for your advice, i'll retry it and check the processors after decomposition... Perhaps it's just as simple as that!

Any hint about 1-. Or would rather suggest me to use directMappedPatch instead of turbulentInlet, in order to get the best inlet condition?

Antoine

[cedric_duprat]

I don't really know what's behind "turbulentInlet" but, I think it's not a "LES inflow BC", at least for internal flow.
directMapped has already been used as turbulent inflow by different OpenFOAM LES users such as G. Tabor, E. de Villiers or myself.
The negative point is that the cost of your calculation will increase because you'll need a longer inlet.
On the other side, for your type of configuration, maybe you don't need a "turbulent inlet". So maybe fixing the rate of turbulence at the inlet is enough.

I can't tell you more, you've to check with the existent literature of LES of external flows to know what is the way of doing.

I hope this will help you

Cedric