Development and Applications of Reactor Noise Analysis at Ontario Hydro's CANDU Reactors

Abstract

In 1992 a program was initiated to establish reactor noise analysis as a practical tool for plant performance monitoring and system diagnostics in Ontario Hydro's CANDU reactors. Since then, various CANDU-specific noise analysis applications have been developed and validated. The noise-based statistical techniques are being successfully applied as powerful troubleshooting and diagnostic tools to a wide variety of actual operational I&C problems. The dynamic characteristics of critical plant components, instrumentation and processes are monitored on a regular basis. Recent applications of noise analysis include (1) validating the dynamics of in-core flux detectors (ICFDs) and ion chambers, (2) estimating the prompt fraction ICFDs in noise measurements at full power and in power rundown tests, {3) identifying the cause of excessive signal fluctuations in certain flux detectors, ( 4) validating the dynamic coupling between liquid zone control signals, (5) detecting and monitoring .mechanical vibrations of detector tubes induced by moderator flow, (6) estimating the dynamics and response time of RTD temperature signals, (7) isolating the cause of RTD signal anomalies, (8) investigating the source of abnormal flow signal behaviour, (9) estimating the overall response time of flow and pressure signals, ( 10) detecting coolant boiling in fully instrumented fuel channels, (11) monitoring moderator circulation via temperature noise, and (12) predicting the performance of shut-off rods. Some of these applications are performed on an as-needed basis. The noise analysis program, in the Pickering-B station alone, has saved Ontario Hydro millions of dollars during its first three years. The results of the noise analysis program have been also reviewed by the regulator (Atomic Energy Control Board of Canada) with favorable results. The AECB have expressed interest in Ontario Hydro further exploiting the use of noise analysis technology.