Page 1 of 1

monitoring convergence

Posted: Thu Apr 26, 2012 10:47 am
by d.charreron
Hello,

I would like to understand the different terms of the convergence on the listing files. Their physical meaning and the avarage value where we can suppose the calculation converged.
Rhs norm
N-iter
Norm residual
Derive

Is there a way to insure the calculation has converged at 95% for example? Or that the residual are less than 5%?


Thanks,


damien

Re: monitoring convergence

Posted: Thu Apr 26, 2012 7:31 pm
by Jacques Fontaine
These criteria are not useful to determine convergence of your calculation. They are printed in listing for rhs sub iterations and linear solvers monitoring.

In case of steady solution, to control the calculation convergence, you need to add probes in your domain (at suitable locations: not near the boundary or at a place where there is no physic...) and plot these probes (or averaged probes - see Time averages in the gui - ) during the calculation. Your calculation is converged when your ALL your probes are stagnant.

Best regards,
JF