Development of network-based probabilistic safety assessment: A tool for risk analyst for nuclear facilities

Abstract The probabilistic safety assessment (PSA) methodology has been developed and utilized to assess the overall risks to a nuclear facility. However, PSAs are challenged when it comes to accurately describing relations among events, or to accommodate newly observed data, or to consider severe accident scenarios within a current framework. To overcome such challenges and take advantage of the merits of recent systems analysis concepts, this paper develops an improved PSA approach, by integrating the current fault tree-based PSA framework with a Bayesian network. The proposed approach enables one to account for event relations beyond logic gates, to incorporate additional field observations and to conduct vulnerability assessments in an accident condition. To demonstrate the proposed Bayesian-based method, it is applied to a nuclear research reactor recently constructed in JUST, Irbid, Jordan. Several case studies are conducted to demonstrate how realistic information about events and from field inspections changes the core damage risk. In addition, critical scenarios are investigated for an accident, to perform vulnerability assessment beyond a design-basis event. Consequently, it is shown that the proposed approach provides an enhanced framework for risk assessments at nuclear facilities. This framework is ultimately expected to improve decision support for risk-informed designs.