ASSESSING CONFORMANCE TO SAFETY GOALS USING NONPARAMETRIC EMPIRICAL BAYES METHODS - A NUCLEAR-REACTOR APPLICATION

Nonparametric empirical Bayes methods are used to develop decision cri teria for use in deciding whether the risk of a given facility is comp atible with a corresponding specified quantitative safety goal. The cr iteria utilize the uncertain results of a probabilistic risk assessmen t (PRA) and are derived from an empirical Bayes point of view in which the results from a set of similar facilities are used to estimate the population variability curve (PVC) for the parameter of interest. The PVC is estimated nonparametrically in the sense that the distribution al family to which the PVC belongs is completely unknown and unspecifi ed For the assumed model, the method guarantees that all facilities ul timately accepted as being compatible with the goal have a prespecifie d exact assurance probability that the goal is not exceeded. The metho d also accounts for two possible biases in the PRA results. Criteria a re developed for use in assessing the compatibility of nuclear power p lant PRA-produced severe core damage frequency estimates with a corres ponding subsidiary objective.