Cluster statistics based normalization for online fault diagnosis of self-powered neutron detectors

Self Powered Neutron Detector (SPND) is a widely used sensor for measuring neutron flux in a nuclear reactor. In this work we propose a novel cluster statistics based normalization scheme to normalize SPND measurements. These normalized measurements are subsequently used in a recursive Principal Component Analysis (PCA) based approach for detecting faults and identifying faulty SPNDs in an online manner. The motivation behind cluster statistics based normalization is that faults effect only individual sensors, while simultaneous variations in multiple sensors are usually caused by dynamic variations in the reactor operation. The proposed normalization approach is applied on SPND data obtained from an operating nuclear reactor and results compared with existing sensor statistics based normalization approach. The results demonstrate the utility of the proposed normalization approach.