first, you did not provide details about your problem so that discussing BC.s in your setup is not possible...

In general:

1) Use dynamic SGS model, the turbulent viscosity will be determined by the flow conditions

2) Walls:Use natual BC.s for velocity if you have 3-4 grid nodes within y+

This can help

http://www.cfd-online.com/Forums/main/1126-inlet-outlet-boundary-conditions-les.html

HI LES boundary conditions. Generally speaking the reason we use LES is because it can begin to address issues of transition and turbulence through the simulation of the large eddies: the smaller ones are modeled sub-grid-scale. So yes you will need a "good grid" maybe y+1 for the boundary layers. However, this region is well enough resolved by y+1 URANS solutions: this is the basis of detached eddy simulations where RANS is used in the boundary layers and LES in the free stream. This might be a good compromise for you. The inlet boundary condition is very important to get a good prediction of inlet turbulence structures, these will influence the transition prediction. This is very demanding, you really need to compute the upstream flowfield in order to have reasonable inlet turbulence. People have tried random number perturbation of the inlet velocity, this "fake" turbulence gradually mixes to become a reasonable turbulence spectrum but with LES the eddies smaller than the grid are not resolved of course.

cheers Martin

Hi Martin,

Thank you for the advice. Regarding the the inlet BC, I have found in OpenFoam an implemented BC for the inlet that would allow mapping the internal field of another layer of the same field downstream of the inlet. I think that this is very helpful as it allows you to have a fully developed flow at the point of interest without having to have a very long domain upstream of that point. So I think that by setting a fixed  initial  value at the inlet and then mapping data from downstream could be a good compromise.

Best,

Karim

first, you did not provide details about your problem so that discussing BC.s in your setup is not possible...

In general:

1) Use dynamic SGS model, the turbulent viscosity will be determined by the flow conditions

2) Walls:Use natual BC.s for velocity if you have 3-4 grid nodes within y+

We're working a lot in LES modelling with OpenFoam. First of all you should understand if you're going to use a resolved mesh, i.e. first mesh grid point at y+ < 1, and 8-10 in laminar region, or you need to use a wall function. 

With RANS it is very difficult to cath the flow separation, indeed you will not be able to get because this is they fail at. However, LES is very difficult to set up if you do not have the initial conditions. There are a lot of resources fr the initial condition using synthetic approach, or introducing random variation (note recommended).  But, as professor Denaro mentioned they depend on your computational resources!!

Have a look into this paper

Wu, X. (2017). Inflow turbulence generation methods. Annual Review of Fluid Mechanics, 49, 23-49.