KPCA-based fault detection and diagnosis model for the chemical and volume control system in nuclear power plants

Abstract

To study the fault intelligent detection and diagnosis method of nuclear power plant systems and improve the detection and diagnosis effect of internal fault of nuclear power plant Chemical and Volume control System (CVS), this study presents an intelligent Fault Detection and Diagnosis model for the Chemical and Volume control System (FDD-CVS) in nuclear power plants (NPPs). The model is based on failure mode and effects analysis of the CVS system and is implemented by combining kernel principal component analysis (KPCA) with decision tree and support vector machine (SVM). FDD-CVS can rapidly and visually recognize faults in CVS based on independent time-point system parameters, and it is capable of diagnosing fault types and specific fault locations. The model is characterized by clear principles, hierarchical diagnostics, fast diagnostic speed, and visualized results. The model is trained and tested by using the data of the passive nuclear power simulation analyzer. The fault detection rate of FDD-CVS is 96.38%, the false alarm rate is 4.34%, and the average accuracy rate is 98.40%. Overall, the fault monitoring and diagnostic method proposed in this article is innovative and provides valuable references for fault diagnosis research in nuclear power plants.