outlier variable residual norms

320 views
Skip to first unread message

Andrew....@csiro.au

unread,
Apr 8, 2014, 9:17:33 PM4/8/14
to moose...@googlegroups.com
Hey,

after some time steps i now get

Outlier Variable Residual Norms
variable1: number1
variable2: number2

etc. I must say this is pretty cool, but i'm not sure what the numbers actually mean. Eg, my Nonlinear |R| is 1E-4, but the Outliers are 1E-1. I would have expected the latter to be a lot smaller than the former. So what are these norms?

a

Permann, Cody J

unread,
Apr 9, 2014, 9:28:53 AM4/9/14
to Andy Wilkins, moose...@googlegroups.com
The outliers are those variables whose square of their individual L2-norm  is "too far" from the square of the average norm.  Right now "two far" is some fixed value (currently 2.0) determined empirically by a computer scientist ;)  If you want to see the actually calculation of that value, it's in NonlinearSystem.C.

We print them because they might be too big to reliably pass a tolerance check.  It's sort of like a yellow flag if you see a bunch of those coming out.  You might try tighter tolerances or different solver options, etc.

Cody



a

--
You received this message because you are subscribed to the Google Groups "moose-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to moose-users...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Slaughter, Andrew E

unread,
Apr 9, 2014, 9:34:14 AM4/9/14
to Permann, Cody J, Andy Wilkins, moose...@googlegroups.com
Correction, the code you should examine is Console::writeVariableNorms() in framework/src/outputs/Console.C. I am in the process of removing the code in NonlinearSystem, although they are nearly identical. 

Andrew....@csiro.au

unread,
Apr 9, 2014, 6:01:13 PM4/9/14
to andrew.s...@inl.gov, cody.p...@inl.gov, moose...@googlegroups.com

Thanks Cody and Andrew.  I suggest the Outliers be scaled in the same way as the Nonlinear |R|.  It seems strange to me that |R|=1E-4, while the outliers are 1E-1.

 

a

 

 

Ph: +61 7 3327 4497.  Fax: +61 7 3327 4666
Queensland Centre for Advanced Technologies
PO Box 883, Kenmore, Qld, 4069 
 

Permann, Cody J

unread,
Apr 9, 2014, 6:32:06 PM4/9/14
to Andy Wilkins, Andrew E Slaughter, moose...@googlegroups.com
Andy,

We don't scale anything here.  We simply do a comparison.  Your output doesn't make sense to us, can you paste the actual output of one of your solves with the norm and outliers?

Thanks,
Cody 

Andrew....@csiro.au

unread,
Apr 9, 2014, 6:35:21 PM4/9/14
to cody.p...@inl.gov, andrew.s...@inl.gov, moose...@googlegroups.com

Sure:

 

Time Step  2, time = 18400219.8

                dt = 119.800000

0 Nonlinear |R| = 9.484776e-01

1 Nonlinear |R| = 6.886460e-03

2 Nonlinear |R| = 6.095817e-05

Nonlinear solve converged due to CONVERGED_FNORM_ABS iterations 2

Solve Converged!

 

Outlier Variable Residual Norms:

  pwater: 9.484776e-01

  pgas: 7.285548e-05

  conc: 9.178726e-05

 

a

 

 

Ph: +61 7 3327 4497.  Fax: +61 7 3327 4666
Queensland Centre for Advanced Technologies
PO Box 883, Kenmore, Qld, 4069 
 

Reply all
Reply to author
Forward
0 new messages