Finite mixture models for sensitivity analysis of thermal hydraulic codes for passive safety systems analysis

Abstract

For safety analysis of Nuclear Power Plants (NPPs), Best Estimate (BE) Thermal Hydraulic (TH) codes are used to predict system response in normal and accidental conditions.

Invariant methods for an ensemble-based sensitivity analysis of a passive containment cooling system of an AP1000 nuclear power plant

“…As a comparison, it is worth noticing that ranking order T inlet ; G, α 1 and D 3 is the same that can be obtained with N=1000 within a FMM-based SA (Di Maio et al 2014a) and with N=1000…”

“…First, a limited number N of simulations of the TH code are performed and a Finite Mixture Model (FMM) is used to reconstruct the pdf of the model output (Carlos et al, 2013). The natural clustering made by the FMM on the TH code output (McLachlan et al 2000, Di Maio et al 2014a) is exploited to estimate global sensitivity measures using a given data approach (Plischke et al, 2013). As shown in (Borgonovo et al (2014)), in fact, variance-based and distribution-based sensitivity measures rest on a common rationale that allows them to be estimated from the same design of experiments.…”

A dynamic probabilistic safety margin characterization approach in support of Integrated Deterministic and Probabilistic Safety Analysis

“…Statistics and Finite Mixture Models (FMMs) [Carlos et al, 2013;Di Maio et al, 2014 b ] are used for the quantification of the uncertainties of the outputs. FMM provides a natural "clustering" of the TH code outputs, by reproducing them providing information pertaining to the most important input variables which affect the output uncertainty, whereas OS focuses on characterizing the PDFs of certain percentiles and providing approximate estimation of safety limits.…”

Hybrid reliability analysis of structural fatigue life: Based on Taylor expansion method

“…After obtaining the statistical moments of the structural fatigue life performance function given in equations (20)- (27), the cumulative distribution function of the performance function can be fitted. To do so, the following methods are available: the maximum entropy method, 25 the polynomial method, 26 the finite mixture density method, 27 the Pearson system, 28 and the saddle point approximation method. 29 These methods cannot avoid singularities, and the obtained numerical solutions may become unstable during the calculations.…”