Estimating the small failure probability of a nuclear passive safety system by means of an efficient Adaptive Metamodel-Based Subset Importance Sampling method
Résumé
The assessment of the functional failure probability of a thermal-hydraulic (T-H) passive system can be done by Monte Carlo (MC) sampling of the uncertainties affecting the T-H system model and its parameters. The computational effort associated to this approach can be prohibitive because of the large number of lengthy T-H code simulations necessary for the accurate and precise quantification of the (typically small) failure probability. To overcome this issue, in the present paper we propose an Adaptive Metamodel-Based Subset Importance Sampling (AM-SIS) approach that originally and efficiently combines the powerful features of several advanced computational methods of literature: in particular, Subset Simulation (SS) and fast-running Artificial Neural Network (ANN) metamodels are coupled within an adaptive MC-based Importance Sampling (IS) scheme. The objective is to construct a fully nonparametric estimator of the ideal, zero-variance Importance Sampling Density (ISD) and iteratively refine it, in such a way that: (i) the accuracy and precision of the corresponding failure probability estimates are improved and (ii) the number of burdensome T-H code runs is reduced, along with the associated computational cost. The method is demonstrated on a case study of an emergency passive decay heat removal system of a Gas-cooled Fast Reactor (GFR). A thorough comparison is made with respect to several advanced MC methods of literature.
Origine : Fichiers produits par l'(les) auteur(s)
Loading...