[Tapoky]

Hi friends,

I need help. My calculation is very simple. But I have one question.
How can I automatically stop calculation when values of MFR in last 1000 iterations below 0.00001? Maybe I need UDF? Or maybe it can be made easier?

[RaiderDoctor]

It's actually really easy to do this. Simply go to the Tree-Solution-Monitors and double-click Convergence Conditions. This will bring up a new window. Click the "Add" button, and est the Report Definition to your Mass Flow Rate Report Definition. I'd double-check on what some of the options within the window means, just for completeness, though.

[Tapoky]

Originally Posted by RaiderDoctor

It's actually really easy to do this. Simply go to the Tree-Solution-Monitors and double-click Convergence Conditions. This will bring up a new window. Click the "Add" button, and est the Report Definition to your Mass Flow Rate Report Definition. I'd double-check on what some of the options within the window means, just for completeness, though.

I thought about it, but the MFR has not only positive value, but also negative. And in Convergence Conditions I can determine only condition when my value below some value. Well all negative values will satisfy that condition too. But it will be wrong.

[LuckyTran]

In newer versions of Fluent it is straightforward to setup this type of asymptotic convergence check for lots of variables. It's v16 it is cryptic and you have fewer options, but doable since you already have a working monitor.

Originally Posted by Tapoky

I thought about it, but the MFR has not only positive value, but also negative. And in Convergence Conditions I can determine only condition when my value below some value. Well all negative values will satisfy that condition too. But it will be wrong.

The stop criterion is not a threshold, it is a type of normalized absolute variation. If m(i) is the monitor, the variance after 1,2,...p iterations is:

R(1) = abs(m(n)-m(n-1))/m(n)
R(2) = abs(m(n)-m(n-2))/m(n)
R(p) = abs(m(n)-m(n-p))/m(n)
It checks to make sure all the R's are less than the stop criteria over p iterations. You should note that the variation is normalized the latest value of the monitor. Which means that you need to specify the stop criterion as a % variation rather than as absolute threshold. An example is you could use the the lift coefficient as a convergence monitor (which should converge to some finite positive number).

[RaiderDoctor]

Looks like Lucky beat me to the punch. Go ahead and give it a try and check back in to make sure it works.

[Tapoky]

LuckyTran, Thank you very much for your help. It really helped me.