Information content of integral experiments data: A Bayesian approach for quality assurance. (15th December 2021)

• Record Type: Journal Article
• Authors: Ivanov, Evgeny; Bess, John
• Is Part Of: Annals of nuclear energy. Volume 164(2021)
• Publication Date: 2021-12-15
• Subjects: Data assimilation -- ICSBEP -- IRPhE -- Inferred experimental data -- Informativeness; Nuclear energy -- Periodicals; Nuclear engineering -- Periodicals; 621.4805
• DOI: 10.1016/j.anucene.2021.108657
• Languages: English
• ISSNs: 0306-4549

Abstract

Highlights: Characterization of Integral Experiments data in terms of information theory. Quantified metrics of uncertainties built up via Data Assimilation procedure. An evaluation of experiments data addressing uncertainties and lack of knowledge. Unfolding of inferred experimental data and combination of partial experiments. An extension of representativity factor on correlated integral experiments. Abstract: Nuclear technological science involves predictive calculations in different fields requiring all concepts and models having a solid basis in reality. In this context, the only observations and experiments could provide an objective data. This is why, we consider calibration and validation against high-fidelity and well-characterized experiments as two major directions liaising the simulations and the real world. It seems clear that the needs in experiments depend on their intentions – either for calibration or for validation. For example, calibration would require a few maximum representative benchmarks, while validation should not ignore anyone unless ones proven as of low-fidelity or erroneously evaluated. We are focusing, mainly, on an evidence-based background needed in a validation process, including an extension of the internationally recognized standards by an involvement of data assimilation to provide entirely assessor-independent benchmarks if, even, they are based on an inferred but not on a direct measurement. At the level of a consideration of each individual experimental case we are following an idea that the characterization of experimental data comprises two macroscopic categories, such as the experimental data credibility and the similarity of the physics behind the experiments and a given application. In such a context, we are discussing some recent advances in an establishment of the indicators of informativeness. The reasonings are illustrated by several practical cases related to an experimental data evaluation and to criticality safety assessment.