curious residuals values

[Olivier Geoffroy]

Hello,

I am a little septical about residuals value given by code saturne. (tested rc1 and rc2)

In a lot of cases, residuals value are in range of 1e-2 (ok) to 1e4 (!), and never goes down let say, 1e-3.

I give some picture from a simple 3D pipe with heat source inside, and plot residual and probes (inlet for pressure, outlet for the thermal effect).

Running more iteration doesn't change the slope of probe and residuals, and results are ok, but residuals for TempK is stil near 1e3.
 
What can be wrong ?
 
Olivier

[David Monfort]

Hello Olivier,

Could you attach the listing file, so that I can have a look at the different values? And by the way, which value exactly are you extracting from the listing to plot the residuals? Code_Saturne does not output (yet) "true" residuals to enable the user to track the convergence. The residuals values are actually related to the linear system (iterative) solvers.

David

[Olivier Geoffroy]

Hello David,
 
I give you my listing, and also the quick and dirty python script which plot the residuals (need mathplotlib to run). To run the script, just do CS_plot <Path to the listing file>).
 
The value i am extracting and plotting from the listing is the "Rhs norm", not the Norm residual (related to the linear system), neihter the "derive" value.

But the "Rhs norm" don't give "good" residuals value.
 
Olivier