Multi-Hazard Risk Assessment Using Bayesian Network and Fault Tree Analysis Considering Effects of Structural Damage

Abstract

Recently, South Korea experienced two strongest earthquake events in its modern history, i.e. 2016 Gyeongju (Mw 5.4) and 2017 Pohang Earthquakes (Mw 5.5). In the region generally considered as a low or moderate seismic zone, the occurrences of such earthquakes and their socio-economic consequences alarmed the general public. Moreover, those earthquake events featured a number of main- and after-shocks, which raised a significant concern about potential major catastrophes caused by multi-hazard effects. This paper presents a probabilistic framework being developed to assess such multi-hazard risk of nuclear power plants (NPPs). First, a ground motion prediction equation is represented by a Bayesian Network (BN). The relationship between main- and after-shocks, e.g. the modified Omori law is incorporated into the BN. Second, to overcome limitations in existing Probabilistic Risk Assessment (PRA) of NPPs, which often employs event tree and fault tree analysis, the BN representing the multi-hazard is connected with the fault trees constructed for the NPP. Finally, to address the impact of structural damage caused by earlier shocks on later events, the fragilities of NPP components are updated. These updated fragilities are incorporated into the fault trees connected with the BN for accurate after-shock risk assessment. The proposed methodology integrates our knowledge on the multi-hazard (BN), reliability of NPP (fault tree) and inter-hazard effect (system identification). The proposed framework is demonstrated by an NPP under main- and after-shock scenarios. Potential applications to other types of multi-hazards and future research needs are also discussed.