Detection and diagnosis of abnormal transients in nuclear power plants

We describe a simulation based algorithm that combines fuzzy logic with macroscopic conservation equations to diagnose multiple-failure events subject to uncertainties in transient data. Clusters of single-failure data points of similar characteristics are obtained through a pattern recognition algorithm and the cluster centers are combined in the space of macroscopic inventory derivatives to generate multiple-failure clustercenter. A fuzzy membership function is used to represent the likelihood of a data point belonging to a cluster, and failure estimates are obtained through solution of a fuzzy matrix equation. The algorithm has been sucessful in detecting simulated malfunctions in the pressurizer of a pressurized water reactor. (author)