Uncertain data and expert opinions in the assessment of the unavailability of suppression systems

Abstract

The probability distribution of the demand unavailability of water-based special suppression systems (e.g., sprinklers) for nuclear power plants is developed from direct nuclear experience and indirect evidence from tests and non-nuclear applications. The indirect evidence is treated as expert estimates, and is combined with the direct evidence using Bayes' theorem. The credibility of the estimates is explicitly modeled in the likelihood function. The sensitivity of the results to the major assumptions of the model is investigated.