Application of principal component analysis for the diagnosis of neutron overpower system oscillations in CANDU reactors

An accurate fault modeling and troubleshooting methodology is required to aid in making risk-informed decisions related to design and operational activities of current and future generation of CANDU® designs. This paper attempts to develop an explanation for the unanticipated detector response and overall behavior phenomena using statistical methods to compliment traditional engineering analysis techniques. Principal component analysis (PCA) methodology is used for pattern recognition using a case study of Bruce B zone-control level oscillations.