How do I choose the refinement criteria and the value of per cent of cells being refined?

experience.

Are you refining because you have a converged solution and you want better resolution, or it is not converged?

allan

I need to perform a grid-independence check.

well!

double the number of cells in each direction then

allan

... but even a converged and grid independend solution might be completely wrong.

You should look for the correct solution

search for Richardson extrapolation. It is quite old stuff (around 1920, I guess) but it is a mathematical procedure to obtain a grid independent solution.

you have to start with a coarse mesh and then refine it twice by facor 2 in each dimension (2x2x2 for 3D case). then you to put the results (pressure drop, temperature increase/decrease or whatever into an equation and you will get the so called grid-independent solution. but keep in mind: this does not mean, you get the 'right' solution!

thomas

This is a respond to Joern and Thomas. If you have a grid-independend solution, you will THE solution for the problem that you have specified.

You probably have the solution for the mathematical problem you have specified but not for the psysical problem which you want to solve.

Yes, I agree with you. But this was not the question in the first place...

BONJOUR il faut rafiner le maillage au niveau des parois et dans les zones de grandes perturbation "zones de recirculation et à l'entrée.