grid size
 

Hi, Can any one tel me (in more details) how can I see the the influence of the grid size on the computed solution. I'm traying to use a vof code. thanks.

Re: grid size
 

It strongly depends on properties of the flow and type of convection scheme you used to discretize convection terms in solution to NS equations. One of the most common way to see this influence is IMHO testing, testing, testing. I may be wrong - which will be corrected by next posts probably :) Do you use your own vof code? -- Maciej Matyka my home page

Re: grid size
 

What do you mean with IMHO testing? I want to now how to make difference between two descritizations and what is their influence on solution. for exemple, in 2D if I take (nx x ny) and (n1x x n1y), how can I take the best grid.

About vof, I'm coupling several versions of vof issued from open sources. I want to verifiy if my criterian of "basculement" is correct, then I started by elimination of the sources of spurius currant step by step, which first is the grid dependancy.

Re: grid size
 

In addition to above comments, as a rough estimation, to get enough resolution for VOF calculations, if you are dealing with a jet or droplet, it's recommended to put at least 10-15 grids per radius to get kind of grid-independent solution.

Re: grid size
 

thanks John, but I need more details. If you tel me to use 10-15 grids per radius, is it ghost cells or what?

Re: grid size
 

VOF, by itself does not need ghost cell. As you may know, ghost cells are only necessary (in general in MAC formulation of incompressible flow) in implementing some boundary conditions (BC) and they're used only in staggered arrangemnent (I admit 99% of VOF methods use staggered arrangement). So if you don't have BC near the liquid phase, you don't need to be concerned about ghost cells and even in case of having BC, you won't need more than 1 ghost cell for every BC. So, my suggestion is use 10-15 cells per radius not including ghost cell, although 1 ghost cell doesn't make any difference. This is an empirical estimate that gives good resolution. About spurious current, remember that you can't get grid-independent spurious current magnitude, since spurious current is error associated with CSF (and other methods of) implementation of surface tension. So, if you want to get grid-free solution and do grid study, I think you can use a case with minor effects of surface tension or a dynamic case in which inertial effects dominate the flow (not surface tension-dominated flows). Like rayleigh jet breakup case or drop collision or ....

Re: grid size
 

Thanks a lot John but if you can give me your email. yours MER