[ArathoN]

I searched a lot about how to do a preliminary analysis of the grid sizing in the 3 directions (normal, spanwise and streamwise), but I've found only info about the normal direction (y) with all the possible consideration about yplus and wallfunction.

Now do any of you know some good report and lectures/books where they give some insight on how to consider the grid size in the other two directions?

I'm trying to study simple cases like channel or backward facing step, but I found myself completely clueless about this aspect of CFD, I searched all the time about some info with no success. All the papers I've found they only specify the grid sizes without explaining their choice.

P.S. as a side note how do you refine a non-uniform mesh? I saw in the papers that they halve the first grid size (in one direction), so do they maintain constant the ratio and increase the grid points number? or is there a better technique?

[ArathoN]

I'll write my conclusions:
- first you should know if the problem is 1d,2d or 3d. It is important because you'll have the direction where you need to refine mesh (e.g 1D means only one direction to refine the other can be considered uniform)
-cell-to-cell expansion max 1.2-1.15
-if you define yplus=<1 you should at least have 10 grid point in boundary layer (RAS models)
-max aspect ratio around 20 (can be higher in some cases)

these are the guidelines when using RAS Models. I hope it will help someone.

A nice utility to define grading is this.