[LuckyTran]

Originally Posted by hiteshlande

Hello
Can anyone tell me that how to find exact mesh size for LES and how to do mesh independence study for it? As I am working on Solid Rocket Motor flow dynamics using LES....

Thank You...

You need at least 2 points across the structure you are trying to resolve (preferably 4 or more). Mesh independence is same as for any numerical computation. Look-up grid convergence index by Roache if you are totally lost.

[Andrea1984]

There is no such a thing as mesh independence for a LES performed with an implicit filter (which is the most common approach when it comes to LES), since the size of the filter is directly related to the mesh size the more you refine your grid the more you tend towards DNS (for a well posed LES), so in this case is more correct to talk about grid convergence rather than grid independence.

It is possible to obtain a mesh independent LES by switching to explicit filtering, and thus eliminating the direct relationship between filtering and discretization, as explained here: https://web.stanford.edu/group/ctr/S...0/3_04_you.pdf

[farouksdz]

hi, according to fluent a mesh corresponds to y+=1 (or less) is sufficient for LES turbulence model. to see the independance mesh simulation you could make several meshes correspond to different y+ values (ex: y+=1; 0.8; 0.6; 0.4; 0.2) and you could find the optimal mesh where there is no difference between two successive y+ values. good luck

[hiteshlande]

Thank You Very Much all of you for reply.....

[Andrea1984]

To complete my previous postI would suggest you to have a look at Celik et al, Index of resolution quality for Large Eddy Simulations, Journal of Fluids Engineering 127 (2005) 949-958

[LuckyTran]

For wall bounded flows (and wall-resolved LES) the y+ requirement near the walls is the most limiting factor and satisfying the y+ and some maximum cell aspect ratio (15,20,25 etc) usually dictates the largest cell size in the domain.

But if you wanted to, you could estimate the "turbulent length scale" of interest and use that as your base cell size, whether the length scale of interest is the geometric scale, integral scale, Taylor scale, or Kolmogorov scale is up to you.

I also want to correct myself.

Originally Posted by Andrea1984

There is no such a thing as mesh independence for a LES performed with an implicit filter (which is the most common approach when it comes to LES), since the size of the filter is directly related to the mesh size the more you refine your grid the more you tend towards DNS (for a well posed LES), so in this case is more correct to talk about grid convergence rather than grid independence.

Yes, implicitly filtered LES is by definition grid dependent. Grid convergence is much better way of talking about mesh dependency rather than grid independence. Technically your solution field is always different with any different grid.

But depending on what variables it is you are trying to resolve, those can become mesh independent or mesh insensitive (but I still prefer to refer to this as grid convergence rather than grid independent). While detailed local variables such as (velocity, temperature) may be grid sensitive, you can still achieve grid insensitive results for bulk quantities like overall pressure drop, pumping power, overall heat flux, etc.

[hiteshlande]

Thank You very much.
one more query is there. How to determine time step size for LES(is it using CFL criteria) and is this CFL criteria is important for implicit LES formulation?

[LuckyTran]

I think you meant implicit time-stepping and not implicit LES (implicit LES means LES with an implicit filter).

The CFL criteria is for determining stability which you generally do not have trouble with in implicit time-stepping schemes. But the Courant number does tell you a lot about temporal resolution. Stability also does not guarantee accuracy and you still need small enough time-steps to ensure accuracy. Hence, you should still target sufficiently small time-step (Courant number approximately 0-1) to ensure your results are accurate and ensure you are not temporally clipping the results.