Nuclear Reactor Process Monitoring in Time-Domain Using Stochastic Signals

摘要

Present-day operating-, safety- and monitoring systems usually mainly payattention to the large process changes or anomalies. By focussing on small changes it may be possible to detect certain process disturbances or malfunctioning of the nuclear power plant in an early stage. In this manner unwanted disturbances can be dealth with before they cause any serious problems. In this work techniques and methods are presented that are designed to detect and analyze these small process changes. Chapter 1 presents an inroduction to the process. In chapter 2, several signal processing techniques (sometimes also called signal analysis techniques) are discussed. Chapter 3 deals with anomaly detection methods of which three are discussed, namely the extremes method, the distribution method and the sequential probability ratio test. In chapter 4 each anomaly detection method is optimized and the three methods are compared using results from theory and simulation. The signal processing techniques presented in chapter 2 can be applied simultaneously to one signal. A method for combining the outcomes of several techniques, which were applied to a synthetic neutron noise signal, is presented in chapter 5. This is done using an approach based on fuzzy logic. In chapter 6 the new experimental setup SIMBOL and the measurements that were performed with this facility are presented. In chapter 7 the first step towards an application of a process-monitoring system is done and presented. The techniques and methods presented in this thesis have a broader area of application than just reactor noise signals.