Guiding the Formulation of Surrogate Values of Modeling Input Parameters in Screening Analysis of Probabilistic Risk Assessment (PRA)

In the screening analysis in Probabilistic Risk Assessment (PRA), precise plant data collection is avoided as much as possible due to time and resource constraints. These constraints mainly arise from the large number of events or underlying factors, which can make even simplistic modeling and simulation (M&S) efforts economically unviable. As a consequence, modeling assumptions in screening analysis are needed to determine the appropriate level of realism that can lead to efficient screening, i.e., conservatively screening out PRA items while minimizing the associated costs. In this work, the progress on the development and application of a novel methodology that aids PRA analysts and practitioners in the formulation of modeling assumptions is provided. The methodology focuses on the modeling assumptions associated with the determination of the appropriate surrogate values for the simulation input parameters. Here, “surrogate values” refer to those values of the M&S input parameters for which modeling assumptions have been utilized. Such modeling assumptions introduce epistemic uncertainties that can substantially influence the outcome of the analysis, and therefore, impact the decision-making process, such as leading to false negative results in the screening analysis. The proposed methodology is based on backward uncertainty propagation using large simulation sensitivity dataset. This dataset is generated using simulation models and used to inform the choice of the surrogate values of modeling input parameters. The application of the methodology is demonstrated in multi-compartment fire scenarios screening in fire PRA. The case study addresses the modeling assumptions associated with ventilation conditions in fire analysis.