The fault tendency analysis of hydro-generator based on WNN

In the research of fault diagnosis of the machinery, it is necessary to analyze the fault tendency of machinery. An approach is designed that predicts and controls the futural running status of machine in time by wavelet neural network. By this approach, feature analyzes the tendency of machine fault visually with histogram. Information is first extracted from original machinery signal by wavelet packet transform. Then, feature information is put into Kohonen self-organized mapping neural network to be clustered and form some different classification spaces of running status of machine. Finally, analyzing correlation between feature information during different running periods to predict the futural running status of machine. Experiment shows that the method works well in the fault diagnosis and tendency analysis of hydro-generator.