MSPI false indication probability simulations

This paper examines false indication probabilities in the context of the Mitigating System Performance Index (MSPI), in order to investigate the pros and cons of different approaches to resolving two coupled issues: (1) sensitivity to the prior distribution used in calculating the Bayesian-corrected unreliability contribution to the MSPI, and (2) whether (in a particular plant configuration) to model the fuel oil transfer pump (FOTP) as a separate component, or integrally to its emergency diesel generator (EDG). False indication probabilities were calculated for the following situations: (1) all component reliability parameters at their baseline values, so that the true indication is green, meaning that an indication of white or above would be false positive; (2) one or more components degraded to the extent that the true indication would be (mid) white, and "false" would be green (negative) or yellow (negative) or red (negative). In key respects, this was the approach taken in NUREG-1753. The prior distributions examined in this paper are 1) the constrained noninformative (CNI) prior used currently by the MSPI, 2) a mixture of conjugate priors, 3) the Jeffreys noninformative prior, 4) a nonconjugate log(istic)-normal prior, and 5) the minimally informative prior investigated in [1]. Results are presented for a set of base case parameter values, and three sensitivity cases in which the number of FOTP demands was reduced, along with the Birnbaum importance of the FOTP.