[dinhanh]

Hello,

I am trying to do simulation with the LES model for the nozzle flow. For LES, that requires a large of number of grid. However, if I understand correctly, there is no rule for the number of grid for LES where the turbulent length scale is compared to the local grid size. it means that the domain grid is depended on how detail of turbulent effect you want to know, doesn't it?

I read in some papers, the number of grid for LES is listed as Nx, Ny, Nz. What are those number?

In case I make the grid for Fluent, how small of cell, and how many cells I need to make?

In case using wall function for LES (in Fluent), the first grid cell should be in the log law region, shouldn't it?

Thank you.

[LuckyTran]

LES is a filter-dependent simulation and hence (usually grid-size dependent). The spirit of LES is to resolve only the largest scales in the flow, but some people will run a near-DNS style LES. It's up to you how fine you want to run your LES.

Originally Posted by dinhanh

I read in some papers, the number of grid for LES is listed as Nx, Ny, Nz. What are those number?

Check the definitions in the paper. Usually Nx,Ny,Nz means # of cells in each direction (for a cartesian grid) but I've also seen Nx,Ny,Nz used as resolution.

Originally Posted by dinhanh

In case I make the grid for Fluent, how small of cell, and how many cells I need to make?

In case using wall function for LES (in Fluent), the first grid cell should be in the log law region, shouldn't it?

Thank you.

If you know the turbulent length scale that you want to resolve. Then your grid size should be smaller by some factor. e.g. L/3, L/10, L/16, L/32. This factor is a measure of the LES resolution but is very similar to grid requirements for RANS in the mean velocity profile. You need some # number of grid points to resolve the velocity profile. In LES you need # number of grid points to resolve the eddy.

For wall functions, yes your grid should be coarse enough for the wall function to be appropriate.