Global sensitivity analysis of passive safety systems of FHR by using meta-modeling and sampling methods

Abstract

Global sensitivity analysis (GSA) aims at quantifying the individual effect of input variables and the main idea is to decompose the variance of a physical model response into fractions that can be attributed to inputs. The Sobol’ indices are obtained by adopting different algorithms analytically. Those methods can be divided into two main parts: the sampling-based method part includes Latin Hypercube Sampling (LHS) and Sobol sequence; the meta-modeling method part includes Polynomial Chaos Expansion (PCE), Stochastic Expansion (SC), and Arbitrary Polynomial Chaos Expansion (APC). This work employs these methods to perform the global sensitivity analysis of Fluoride-salt-cooled, high-temperature reactors (FHRs). The General Flow (GenFlow) code is coupled with DAKOTA or SALib to analyze the loss of force circulation (LOFC) accident of FHR. This study focuses on the effects of five key thermal parameters on safety parameters. It discusses the characteristics of different methods and the accuracy using different orders of the method. Statistical results show that the power load factor would affect safety of FHRs significantly, as it has significant effect on several outputs. The PCE and SC method have significant advantage, as the sensitivity analysis results are obtained from few samples without sacrificing too much accuracy. If the simulation model has jump discontinuities, the results could be significant uncertainty for both meta-based and sampling-based method.