Dear All,

I'm running trials of various Fluent RANS models on strongly 3D flow, testing their performance (not very good) in predicting separation over a hill shaped obstruction. Re 130000, fluid - air, inlet BL & turbulent properties specified by inlet profile.

I'm using grids of 1M & 2M cells and a number of combinations - order of discretisation, grid, turbulence model & so on. I have a problem with the KO-SST model on the 2M cell grid which I suspect is related to a mistake in my set-up, but I can't find the problem. It doesn't converge and the coefficient of drag is in hundreds/thousands, when all other models are around 2 or 3. (Spalart Allmaras doesn't converge either at any order of discretisation, but the cofficient of drag is more sensible, 12 was the highest figure). On the 1M cell grid, KO-SST drag coefficient converges to 6 significant figures with 3rd MUSCL, but is wildly off when applied to the 2M grid. My first thought was scaling, but the dimensions are correct, I also checked the inlet velocity, turbulent properties etc, but no problem there. Could anyone make any suggestions? If it's just a bad choice of model, why does it give sensible CD on the 1M grid but 1000 times higher on the 2M ?

Any suggestions/comments would be much appreciated!

try decreasing time step

Hi Renato thanks for the reply,

I've had to progress - I'm using KO std, but I'll try reducing time step in future.

Regards,

Ianto.