S-D22.3 Meta-Modeling uncertainty assessment Y4M10

[KZzizzle]

22.3. An uncertainty assessment module exists that permits to propagate parameter uncertainty through a modeling pipeline and to determine the resulting uncertainty of modeled quantities of interest. Y4M10

• support for efficient linearized uncertainty propagation of uncorrelated factors based on S-D2
• support for Monte-Carlo-based analysis of correlated, non-linear factors

Definition of Done & User story:

• An uncertainty analysis service exists. It is possible to select whether a linearized or a monte-carlo based uncertainty analysis is performed (alternatively, have two variants of that service). For monte-carlo, some additional parameters exist (e.g., number of iterations, initial equilibration runs)
• In the uncertainty analysis service, a number of variables can be defined and named (see S-D22.1), along with their typical value and the associated probability distribution (normal/uniform with stddev/range, linear/log scale)
• As a result corresponding output ports (or attached parameter nodes) are created that can be connected to input ports of a pipeline
• Output ports of that pipeline that have the type scalar can be connected back to the sensitivity analysis service
• When running the uncertainty analysis service, one reference pipeline execution and 2*n variations are run for the linearized variant and nriterations variations for the monte-carlo one (see the jupyterlab prototype for detail)
• The uncertainty analysis displays and provides as an output (at a port):
  • The uncertainty contributions
  • The combined uncertainty
  • For linearized UQ: The sensitivities (in [-] and [dB] for linear perturbations, and as [1/dB] or [dB/dB] for relative ones) along with the R^2 value of the 3-point fit.

(see the jupyterlab prototype for detail on both the linear and the MC uncertainty analysis)
A corresponding jupyterlab can be found here.

Note: In the above user story, the uncertainty analysis service acts as both an iterator and a collector object. See 22.2 for alternatives.

[esraneufeld]

todo:

• use technology developed for sensitivity analysis to also implement two uncertainty analysis iterators and corresponding analysis scripts (based on EN implementation)