Developing a Monte Carlo-based fault tree quantification module for SAPHIRE targeting seismic probabilistic risk assessments

久保 光太郎  ; 藤原 啓太*; 村松 健

Kubo, Kotaro; Fujiwara, Keita*; Muramatsu, Ken

During the probabilistic risk assessments (PRAs) of nuclear power plants for external events, such as earthquakes and tsunamis, accurately evaluating system, structure, and component damage correlations is crucial for analyzing risk profiles. This study develops a module called Module Using Random Monte Carlo Sampling (MURAMASA), which integrates the direct quantification of fault trees using Monte Carlo simulation (DQFM) method into System Analysis Programs for Hands-on Integrated Reliability Evaluations. MURAMASA is designed to enhance the treatment of correlated failures and to improve integration with internal event PRAs, thereby providing a practical tool for seismic PRA analysts. The DQFM method enables accurate consideration of correlations of failures in seismic PRA by introducing correlation coefficients into the random sampling of seismic responses and capacities. In conventional DQFM, many trials are required to estimate low-probability outcomes, such as in cases with low seismic acceleration levels. To address this issue, we propose partial DQFM method, which combines DQFM with the minimal cut set upper bound approximation to improve conditional core damage probability accuracy in such cases. A case study involving two integrated fault trees confirmed that MURAMASA can accurately and efficiently obtain fragility curves. MURAMASA demonstrated usability and computational efficiency in seismic PRA analyses.