Fault analysis of in-core detectors in a PWR using time-series models

Abstract

Instrument fault detection and estimation is important for process surveillance, control, and safety functions of a nuclear power plant. In this study, a method incorporating the dual-hypotheses decision procedure and system characterization using time-domain models of process variables representing a subsystem is presented. The multivariate time-series models can be developed on-line, and can be adapted to changing system conditions. The method can be made more effective by considering specific subsystems of a pressurized water reactor (PWR), and by signal selection such that a strong cause and effect relationship exists among the measured variables. The technique is applied to the reactor core subsystem of the Loss of Fluid Test (LOFT) reactor using in-core neutron detectors and core-exit thermocouples. Thermocouple anomalies such as bias error, excess noise, changes in response characteristics are detected and estimated using appropriate measurement models.