A feature-based approach to monitor motor-operated valves used in nuclear power plants

Degradation and failure of motor-operated valves (MOV) compromise operational readiness of the safety related systems of a nuclear power plant. Motor current signature analysis has been found to be a selective and early indicator of a number of mechanical and electrical failures/abnormalities related to the MOV. We present an unsupervised and fully automated method for the extraction of the motor current signature and an analysis to diagnose possible failures in MOVs. The reference line frequency is obtained by sampling the line voltage, and is used to demodulate the current waveform. Having obtained the current signature, a set of features are extracted from the signature. A discriminant analysis is performed on these primitives to detect and classify various types of failures. The proposed method is nonintrusive, computationally efficient and yields good results. It can be easily installed as a part of an expert system for preventive maintenance of MOVs in nuclear power plants