UNCERTAINTY ANALYSIS OF SYSTEM RELIABILITY AND AVAILABILITY FOR NUCLEAR POWER GENERATION STATIONS

Uncertainty analysis of systems' reliability assessments with particular emphasis on commercial nuclear power generation stations has been investigated. A thorough review of reliability theory concepts which pertain to assessments of uncertainty is presented. Sources of uncertainty in reliability assessments are identified, and methods to account for their existence are discussed. Available methods for determining uncertainties in system characteristics (such as unavailability) by synthesizing the known or assumed uncertainties in component characteristics (propagation of errors) are presented and compared.; Two new computer programs (RANGE and COSMOS) for uncertainty analysis are described which implement Monte Carlo simulation and second-order moments method, respectively. These new codes are shown to be superior to several existing codes which have been developed for uncertainty analysis. The COSMOS program significantly extends the applicability of the moments method for error propagation to systems comprised of as many as 100 components (or modules). Comparison with other uncertainty analysis programs indicates that results obtained from COSMOS are consistent with those obtained by existing codes. RANGE provides a very fast, and flexible Monte Carlo simulation program for uncertainty analysis. Comparison with existing programs of this type indicates that RANGE provides superior efficiency and versatility.; Methods for minimizing the uncertainty in systems' reliability assessments are discussed. A reliability data base has been developed which utilizes a data base management system for efficient storage and retrieval of failure data. Several software interfaces for fault tree analysis programs are also described. The contribution of common cause failures to the uncertainty in system unavailability is discussed, with particular emphasis on the evaluation of common cause failure probabilities. A new computer program (CCPE) has been developed which models common cause failures for nominally identical components by a specialization of a multivariate exponential probability distribution which is characterized by a binomial failure rate. Estimates and confidence intervals for the model parameters are provided which can be used to obtain error bounds on unavailability values for use in uncertainty analysis codes such as RANGE.; The influence of noncoherent structures on the uncertainty in calculated system characteristics is examined. Measures of probabilistic importance for noncoherent structures are developed. These measures are used to determine an optimal restoration sequence for failed components in noncoherent systems.; A modular approach to uncertainty analysis is discussed. The general scope and applications of uncertainty analysis within the system design process are described as well.