Beyond-design-basis Accident Analysis using Bayesian Network: Application to Research Reactor

The research reactor can be subjected to an internal accident during the design lifetime. Especially, a loss of primary coolant system (LOPCS) flow is one of significant threats to the safety of pool-type research reactor. This failure is possibly extended to the direct core damage (CD) of the reactor. Under this background, the objective of this paper is to perform the beyond-design basis accident analysis based on a Bayesian network (BN) instead of utilizing the standard fault tree (FT) analysis. For this purpose, the BN for LOPCS leading to the CD is developed based on an event tree (ET) and fault trees (FTs) related to the LOPCS. Then, the BN analysis is conducted under an assumption that the CD accident occurs. The analysis results reveal the updated risks of all events, and the most vulnerable scenarios and basic events. Finally, the BN is expected to be utilized for identifying the real-time risk status and to aid in making risk-informed decisions. This study focuses on the beyond-design-basis accident analysis of LOPCS of research reactor. The concept of a BN is applied for probabilistic safety assessment method instead of utilizing the standard FT analysis for a systems analysis. Unlike the standard FT analysis, the BN enables an analyst to explore scenarios subjected to beyond-design-basis accident. This is because the Bayesian inference facilitates the updating of prior probability information about all events based on the evidence. The numerical result shows that the BN finally assists in the identification of important events that lie on an updated critical path.