Mesh size requirement
 

Hi all,

I'm a beginner of LES. I tried to perform an LES of turbulent flow in a pipe. However, I observed that the fluctuations in velocity disappear after a substantial period of time. I googled about this and read that this might be due to the size of the mesh being too large. I was wondering if anyone can confirm this and also wondering what will be a good rule of thumb for determining mesh size. Thanks!

If your numerical diffusion is high then flow can re-laminarize (but you'd have to be doing really poor LES). What's more common is people don't know that you have to add initial perturbations to the velocity field in order to get a turbulent result. So did you remember to do this?

For pipe flows grid resolution is fairly straight forward. You use the usual 16 pts across the boundary layer (32 across the diameter). Or you can increase to 32/64, etc. What's severely limiting is near wall resolution (i.e. y+ <1). If you achieve your wall y+ requirement for whatever model you are using, and use a small stretch ratio then your grid resolution should be at least in the right ballpark.