Grid Independence Test
 

Hi all,

I need to perform a grid independence test on a simple ish geometry flow (a 0.1m diameter circular pipe expanding into a 0.3m diameter circular pipe). I'm using k-epsilon with standard wall functions.

Should I first do a sensitivity analysis to determine the optimum y+, near-wall distribution and boundary conditions on a standard size mesh, say 200,000 cells? Following this then change the No. of cells while keeping the above optimum conditions constant when trying to find independence?

My first approach was to do the opposite, change cells and then play with the near wall distribution and boundary condition etc. but I can't seem to achieve independence this way.

Thanks :)

Hi,

1. If wall function are not used with the k-e model use a y+ of 1 or smaller
2. If wall function are used consult your theory guide usually you can go for up to 50 y+ even greater. Now if you have adverse pressure gradients, separation etc, please avoid wall functions
3. Before you do grid convergence (grid independent) analysis you need to insure that you set the correct boundary conditions. Keep in mind that with this study you are trying to quantify the discretisation error of your numerical schemes with respect to the grid you are using. Just google NASA GCI and you find the info you need.
4. If your simulation is axi-symmetric you can consider running it first in 2d as you are solving RANS, set-up your CFD configurations (initial boundary conditions solver setting etc) and then extend to 3d.

Hope it helps

Cheers

A