Dear all, I am doing some LES simulation on a circular cylinder with Reynolds Number of 10,000. this is run on CFX software with unstructured mesh. I get confused on how to create the mesh of the computational domain. should I create the mesh based on the Filter Width, or find out a filter width first and create the required mesh later.

all your help is appreciated.

Kai

1. What kind of filter do you use?

2. If you use implicit filter (i.e. grid filter), the filter size is the length scale of grid.

3. In the case of implicit grid-filter, the grid size should be decided depends on the small scale flow structures.

4. Usually you can estimate the Kolmogorov length scale, and make mesh based on that scale. The smallest cells might have scales of 10-100 times larger than the Kolmogorov length scale.

5. If your numerical schemes are very stable and standard Smagorinsky model (only positive eddy viscosity), you may get reasonable (?) results even though your cell size is bigger than 100 times of Kolmogorov length scale.

6. Be careful close to the cylinder wall. What is the distance from the cylinder wall to the first grid point in your case? (In my case, square prism, 2% of D). 0.5% to 1% maybe enough in your case.

7. Not only the normal distance from the wall but also the spanwise length scales are also very importnat. If you want to investigate streamwise coherent structures, please read other literatures on the scale of the streamwise vortex (0.1D in square prism) and distances between the structures in spanwise direction (4D in square prism).

Kim.

Thanks a lot for your guidance Kim,its helpful.

Catch you soon..

Kai

How can one estimate the Kolmogorov scale (and hence the grid size) for a given flow.

Thank you.

hi kim i have a question. how have u decided the distance of the first grid point from the wall of the square cylinder to be 2% of D? I mean on what basis i can calculate it. also how will you know the kolmogorov scales b4 calculations are made?

Thanx in advance. senthil

hi senthil, did u attend any turbulence course ?? if 'yes' then u should know how to calculate y+ and the first y grid point position before solving the prob. refer turbulence course lecture notes or any standard books.

CFD group

IITM

who is this.Please have in mind that this is not the place to play...o.k... sorry if this hurt some ones feelings. senthil.k.b

[triple_lor]

hi Kim, if i want to edit eddy viscosity formula in LES model, how can i specify the grid scale in the formula? because the mesh in ununiform, fine near the wall and coarse in the center. are there any variables in cfx i can use?

Many thanks.