Demonstration of a Pattern Recognition System at the HFIR

摘要

The noise signals from the sensors in a system, such as a nuclear power plant, contain information concerning its current operating condition. Changes in these noise signals may reflect a change in operating condition. Interpretation of these changes may provide the operator with timely detection and diagnosis of anomalous conditions. A pattern recognition system was developed at ORNL to continuously monitor these noise signals on-line in unattended operation. The system was designed to provide a detailed record of the power spectral density values (PSDs) of the signals and their statistical characteristics. Through statistical comparison of current and past PSDs (pattern recognition), the system can detect changes in the noise signals. The system was demonstrated at the High Flux Isotope Reactor (GFIR) by continuously monitoring two neutron signals through four consecutive reactor fuel cycles from April through July 1979. (ERA citation 05:011903)