[prom]

Dear Fellas,

How do you convince yourself as the solution is converged? When there is significant variation in mass residual and your velocity residual shows constant line, does this mean I have to let it solve further or change the grids to redo simulation? Or is my solution converged completely? I have attached residual graph. Kindly help me out. I am lost here.

[LuckyTran]

I convince myself it is converged when I say that there is nothing I can do to make it better. Usually this means I run one case out to an absurd number of iterations (like 100k). And then I compare the solution at the very end with intermediate solutions and look for where I think I could have stopped it. I look for asymptotic solution behavior. I also look for asymptotic residual behavior. Some cases will reach that final solution very quickly, like 1000 iterations, some will take a little longer (10k) and so on.

[Gerry Kan]

Dear Prom:

From the looks what what you presented, I would say your solution is converging nicely. The bounces in mass imbalance is not pathologic, as long as it is generally decreasing. Your monitoring point results also suggest the solution has stabilized, at least at the region of interest.

Personally I would let it run longer until the residual goes below 10e-6 or lower, but this is a matter of preference. In "real life" situations, however, with complex geometries and less than optimal mesh quality, it is not uncommon to see the solution stuck at a high residual (say, 10e-3 / -4).

Gerry.

[prom]

Thanks LuckyTran and Gerry for your reply. Appreciate your help.