Application of quasi-Monte Carlo and importance sampling to Monte Carlo-based fault tree quantification for seismic probabilistic risk assessment of nuclear power plants

Kubo, Kotaro   ; Tanaka, Yoichi; Hakuta, Yuto*; Arake, Daisuke*; Uchiyama, Tomoaki*; Muramatsu, Ken

The significance of probabilistic risk assessments (PRAs) of nuclear power plants against external events was re-recognized after the Fukushima Daiichi Nuclear Power Plant accident. Regarding the seismic PRA, handling correlated failures of systems, components, and structures (SSCs) is very important because this type of failure negatively affects the redundancy of accident mitigation systems. The Japan Atomic Energy Research Institute initially developed a fault tree quantification methodology named the direct quantification of fault tree using Monte Carlo simulation (DQFM) to handle SSCs' correlated failures in detail and realistically. This methodology allows quantifying the top event occurrence probability by considering correlated uncertainties related to seismic responses and capacities with Monte Carlo sampling. The usefulness of DQFM has already been demonstrated. However, improving its computational efficiency would allow risk analysts to perform several analyses. Therefore, we applied quasi-Monte Carlo and importance sampling to the DQFM calculation of simplified seismic PRA and examined their effects. Specifically, the conditional core damage probability of a hypothetical pressurized water reactor was analyzed with some assumptions. Applying the quasi-Monte Carlo sampling accelerates the convergence of results at intermediate and high ground motion levels by an order of magnitude over Monte Carlo sampling. The application of importance sampling allows us to obtain a statistically significant result at a low ground motion level, which cannot be obtained through Monte Carlo and quasi-Monte Carlo sampling. These results indicate that these applications provide a notable acceleration of computation and raise the potential for the practical use of DQFM in risk-informed decision-making.