A Bayesian ensemble of sensitivity measures for severe accident modeling

Abstract

In this work, a sensitivity analysis framework is presented to identify the relevant input variables of a severe accident code, based on an incremental Bayesian ensemble updating method. The proposed methodology entails: i) the propagation of the uncertainty in the input variables through the severe accident code; ii) the collection of bootstrap replicates of the input and output of limited number of simulations for building a set of Finite Mixture Models (FMMs) for approximating the probability density function (pdf) of the severe accident code output of the replicates; iii) for each FMM, the calculation of an ensemble of sensitivity measures (i.e., input saliency, Hellinger distance and Kullback-Leibler divergence) and the updating when a new piece of evidence arrives, by a Bayesian scheme, based on the Bradley-Terry model for ranking the most relevant input model variables. An application is given with respect to a limited number of simulations of a MELCOR severe accident model describing the fission products release in the LP-FP-2 experiment of the Loss Of Fluid Test (LOFT) facility, which is a scaled-down facility of a pressurized water reactor (PWR).

A practical sensitivity analysis method for ranking sources of uncertainty in thermal–hydraulics applications

“…Reference by the authors (Hoseyni et al, 2014) uses this approach and describes a systematic framework for characterizing important phenomena and quantifying the degree of contribution of each parameter to the output in severe accident uncertainty assessment. Another alternative is utilization of ensemble-based sensitivity analysis introduced recently by Di Maio et al (2014), and Hoseyni et al (2015) using finite mixture models by Carlos et al (2013). The idea relies on identification of the rank of parameters by an ensemble of sensitivity measures which are calculated based on the code output distribution estimation from the available data with no further required code computations.…”

“…The limitations of aggregation strategies are that ''majority voting" fails in the case of no agreement among the ranking orders provided by different sensitivity measures. Moreover, the ''sum" strategy fails when the evidence of the superior capability of one (or more) sensitivity measure(s) cannot be accommodated over the remaining ones within its rigorous assignment of equal weights (equal preferences) to the outcomes of the sensitivity measures considered (Hoseyni et al, 2015).…”

Probabilistic sensitivity analysis of offshore wind turbines using a transformed Kullback-Leibler divergence

“…Balesdent et al (2013) uses the divergence as a cross-entropy sensitivity search indicator in the DoE. Hoseyni et al (2015) applies it to characterize sensitivity in a finite mixture model analysis. Greegar and Manohar (2016) uses the KL in a comparison of sensitivity indices, using also the Hellinger, Wassertein (Kantorovic) and the I 2 metrics.…”

Three-loop Monte Carlo simulation approach to Multi-State Physics Modeling for system reliability assessment

“…In this work, we resort to moment-independent sensitivity measures, such as Hellinger distance and Kullback-Leibler divergence (Diaconis et al, 1982;Gibbs et al, 2002), for ranking the input variables most affecting the system reliability uncertainty (Di Maio et al, 2014b;Hoseyni et al, 2015).…”