[2G14-17] PRA Methodology

[2G14] Development of Dynamic PRA Using Multi-Fidelity Models

*Xiaoyu Zheng¹, Hitoshi Tamaki¹, Tomoyuki Sugiyama¹, Yu Maruyama¹ (1. JAEA)

Keywords:Dynamic PRA, Core damage frequency, Multi-fidelity models, MELCOR 2.2, Support vector machine, RAPID

By tightly coupling probabilistic methods and deterministic simulations, dynamic probabilistic risk assessment (DPRA) can be used to estimating risk metrics such as core damage frequency, by means of generating a large number of accident sequences that generally cause an overwhelming computational cost. JAEA has developed a new DPRA approach using multi-fidelity models. By flexibly applying multi-fidelity simulators, it can provide more reliable risk estimation comparing to traditional PRA, but with relatively low computational cost comparing to other DPRA methods.