Research on Fault Diagnosis Method of Steam Turbine Generator Rotor Abnormal Vibration Based on Probabilistic Neural Networks

This paper proposes a fault diagnosis method for rotor abnormal vibration of steam turbine generator based on probabilistic neural networks by using historical monitoring data from the Distributed Control System (DCS) of the steam turbine generator. Firstly, the historical DCS data is subjected to Laplacian Eigenmaps for extracting key variable information that affects the vibration status of the generator in the DCS system, which can be served as the input of the probabilistic neural network. Then, by incorporating the Grey Wolf Algorithm (GWA) to optimize the key parameters of the probabilistic neural network model, the probabilistic neural network model is built for timely warning of early abnormal vibration of generators. Finally, the effectiveness of the proposed method is verified by using the historical DCS monitoring data collected from a 1000 MW large turbine generator unit. The results show that the proposed method can greatly utilize historical data from DCS and prior knowledge of faults, and perform advanced diagnosis of abnormal vibrations in steam turbine generator rotors.