Characterizing Fire PRA Quantitative Models: An Evaluation of the Implications of Fire PRA Conservatisms

Conservative bias may be present in fire PRAs due to limitations in data or methodologies. An evaluation was performed to characterize the current situation with fire PRA models and the implications regarding perceived risk associated with the degree of conservative bias. The principal areas of the fire PRA data and modeling that may be subject to such biases were identified and the impacts these biases have on the reported point estimate CDF and the contributors were quantified. These biases were assessed using a number of sensitivity studies where in a set of modeling approaches or assumptions were varied from the NUREG/CR-6850 guidance that are considered to be conservatively biased. Three point estimates were developed using NUREG/CR-6850 guidance and by incrementally removing biases by crediting more realistic approaches supported in part by revised guidance or in-progress industry and NRC efforts. The conclusion from the evaluation is that reasonable (realistic) approaches to the assessment of the fire hazard will result in a reduced estimate of the fire risk, will likely change the primary risk insights, and could greatly influence the priority that is assigned to possible plant changes resulting from a re-characterization of the causes of risk significant fires and fire zones.