Distribution in Calibration

Persalys usage
1

Hello,

I am performing a test case study with Persalys and the only module I’m struggling with is Calibration.
When I try to use BLUE or 3DVAR methods, I can’t change the standard deviation for the Calibration parameter, an error message is printed in red (see attached screenshot).
Also I don’t see where I can define the covariance matrix of the output observations.
My Persalys xml file is also uploaded. Étude_test.xml (3.3 MB)

PersalysBugCal
PersalysBugCal800×547 52.2 KB

2

Hi Cédric,

Welcome on this forum !

I was not able to load to the xml with my current version of Persalys: what version do you use?

According to the message, I would say that the physical model is not consistent with the observations selected from the data model. Notice that there are several kinds of variables in a calibration study:

  • observed input variables, which are read from the data model, that is \boldsymbol{x},
  • calibration input variables of the physical model, that is \boldsymbol{\theta},
  • observed output variables, which are read from the data model, that is \boldsymbol{z}.

The physical model must be:

\boldsymbol{z} = h(\boldsymbol{x}, \boldsymbol{\theta})

and the data model must contain both \boldsymbol{x} and \boldsymbol{z}.

This is perhaps better explained at:
http://openturns.github.io/openturns/master/theory/data_analysis/code_calibration.html

In the meantime, I try to test in with another version and let you know if I can reproduce the bug.

Regards,

Michaël