2021 journal article

Evidence-based background for constrained uncertainty quantification in a core transient analysis


By: E. Ivanov*, A. Sargeni*, K. Ivanov n & G. Bruna

author keywords: Multi-physics modeling; Validation and uncertainty quantification; UAM-LWR; Data assimilation
Source: Web Of Science
Added: October 26, 2021

The paper discusses some topics related to a validation of multi-physics modeling. Since validation belongs to a category of decision-making processes, it should be prepared by dedicated scientific researches. In particular, the validation means a kind of characterization of the predictive capability maturity of given tools, libraries and calculational models. The compliance criteria might be expressed in terms of achievable accuracy or, inverse, in terms of somehow identified and quantified uncertainties. Despite the uncertainties, as such, might not be measured or compared with something measurable all the judgments should rely on reality, i.e., be supported by an evidence-based background. In practice it does require to eliminate subjective statements, if any, replacing them with something inferred from objective observations, including representative integral experiments. Unfortunately, in many fields like, among others, multi-physics simulations, we have not statistically sufficient number of high-fidelity and confident experiment-based benchmarks. In addition, because of technological and safety constraints, what is needed lies, largely, beyond the experimental domain. This is why, assessors have to rely on numerous, but partially representative experiments. These data could be treated using one or other Data Assimilation techniques to provide correction factors and uncertainties to single- and few-physics modules all having an evidence-based background. Then, coupling these pre-validated modules and their uncertainties, we could estimate uncertainties (and accuracies) in an application domain using one of wide range error propagation techniques. Thus, combining experiments-grounded and calibrated uncertainties, we are providing consistent, evidence-based background for validation. The last phase – constrained uncertainty propagation – was illustrated with an example of one international standard problem on transient initiated in LWR core by inadvertent Control Rod ejection.