A novel methodology to scientifically justify the formulation of modeling assumptions in screening analysis of Probabilistic Risk Assessment (PRA)

In Probabilistic Risk Assessment (PRA) of nuclear power plants (NPPs), there is a growing reliance on modeling and simulation ( M&S). Due to time and resource constraints, PRA analysts do not conduct M&S for every PRA event or its underlying factors; instead, they selectively, through screening analysis, determine the required level of realism for M&S. Since plant data for estimating the values of input parameters may not yet be fully collected, modeling assumptions would be required. This paper proposes a new approach for systematically selecting a set of proper modeling assumptions that minimize a false negative result in the screening analysis. The proposed methodology conducts uncertainty quantification and sensitivity analysis and generates a sensitivity dataset, which can then be utilized to guide and justify modeling assumptions to generate surrogate values of M&S input parameters. The proposed methodology is applied to the screening analysis of a multi-compartment fire scenario at an NPP.