Predictive Decision-Making in Support of the Selection of Degree of Realism in Probabilistic Risk Assessment (PRA)

Due to the large number of items and events in Probabilistic Risk Assessment (PRA) of commercial nuclear plants, risk analysts need to carefully select the degree of realism (DoR), or level of analysis, for the required modeling and simulation (M&S) analysis. This selection is influenced by various aspects such as: i) the level of background knowledge that is pertinent to the PRA item(s) under analysis, and ii) the required resources since conducting M&S imposes financial requirements. In existing literature, the impacts of chosen DoR on PRA outputs are usually analyzed after the baseline risk quantification, as demonstrated in evaluating model uncertainty. In contrast, this research pioneers the development of a predictive decision-making approach to inform the selection of the DoR before the actual M&S runs are conducted. The proposed decision-making approach is developed based on two predictive decision-making attributes: the predicted differences in safety risk estimate (∆SaRi) and the cost of analysis (∆CAN). A qualitative-quantitative process is developed to provide a systematic and structured prediction of the two decision-making attributes. The prediction process is based on causal modeling of the underlying factors that impact the two decision-making attributes, ∆SaRi and ∆CAN. This allows for accounting for the uncertainties associated with subjective prediction and allows for uncertainty propagation into the decision-making analysis. Compared to sole reliance on expert judgment, causal model-based prediction helps reduce the trial-and-error involved in selecting the required DoR in PRA applications. This research is demonstrated on a case study in fire PRA of NPPs, where an adequate DoR needs to be selected from two fire models: an engineering correlation and a zone model. Each fire model represents an alternative for different DoR in the decision-making analysis. The practicality and feasibility of the proposed predictive decision-making approach based on the experience obtained through the case study demonstration.