[luccy]

hi
someone can tell me that this run has converged or not? convergence criteria is 10^-6,why continuity is Fluctuating .is this mean that it is not converged?
i used mixture model,and piso,and skewness correction and neighbor correction=1 and
under relaxation factor is: pressure=.2, mom=.2,k=.5,e=.5 and mesh is triangular and 2d
can i reduce relaxation factor in piso from default value or not?
tanx

[hwet]

Doesnt seem converged to me

[luccy]

Originally Posted by hwet

Doesnt seem converged to me

please tell your reasons.
tanx

[hwet]

actually i just noticed that continuity has reached less than 10e-6 residual as well. what criteria have you set? usually at continuity residuals of 10-4, the iteration process should stop by itself

[CeesH]

Your continuity residual might be fluctuating - but remember that it's very small already. So what does this little bit of oscillation matter? Lowering underrelaxation factors may 'solve' it, but in the end, your Cd and mass flow look like they are already constant up to 4-5 digits, so will any more accuracy really be significant? I think this convergence looks fine. I wouldn't draw too many conclusions on the residuals anyways, having constant averages/coefficients is a better judgement, IMO.

[luccy]

Originally Posted by hwet

actually i just noticed that continuity has reached less than 10e-6 residual as well. what criteria have you set? usually at continuity residuals of 10-4, the iteration process should stop by itself

Originally Posted by CeesH

Your continuity residual might be fluctuating - but remember that it's very small already. So what does this little bit of oscillation matter? Lowering underrelaxation factors may 'solve' it, but in the end, your Cd and mass flow look like they are already constant up to 4-5 digits, so will any more accuracy really be significant? I think this convergence looks fine. I wouldn't draw too many conclusions on the residuals anyways, having constant averages/coefficients is a better judgement, IMO.

tanx for response.
what's your idea about hwet's response??should iteration process stop by itself in convergence?(my simulation is unsteady)

[CeesH]

It depends on what criteria you set. But never ever ever base these criteria solely on residuals, as they are very system dependent. I ran simulations where the solution was constant up to 0.01% in all relevant parameters, where the residuals were ~ 10^-3. I've also had simulations where the residuals were < 10^-5, and the relevant parameters oscillating by more than 1%.

You can also set convergence criteria in FLUENT that are based on average numbers or so - for example, the solution is deemed converged if the solution is stable within a certain predetermined range, for X iterations. These are more suitable criteria to judge by. If you set goals for the residuals only, make sure to check whether your relevant parameters actually reached a constant value before analysis. You can always opt to run some more iterations with stricter settings.

[luccy]

Originally Posted by CeesH

It depends on what criteria you set. But never ever ever base these criteria solely on residuals, as they are very system dependent. I ran simulations where the solution was constant up to 0.01% in all relevant parameters, where the residuals were ~ 10^-3. I've also had simulations where the residuals were < 10^-5, and the relevant parameters oscillating by more than 1%.

You can also set convergence criteria in FLUENT that are based on average numbers or so - for example, the solution is deemed converged if the solution is stable within a certain predetermined range, for X iterations. These are more suitable criteria to judge by. If you set goals for the residuals only, make sure to check whether your relevant parameters actually reached a constant value before analysis. You can always opt to run some more iterations with stricter settings.

tanx very much