Detection for non-stationary vibration signaland fault diagnosis ofhydropower unit

In view of the traditional method is difficult to accurately detect non-stationary vibration signal of hydro-generator units and the low accuracy of existing vibration fault diagnosis methods, this paper introduced the permutation entropy algorithm for detection and analysis. And then realized feature extraction of non-stationary vibration signals based on multi-dimensional permutation entropy, so as to construct fault data samples; The diagnosis model of support vector machine (SVM) based on genetic algorithm is established, and the sample data is the input of the model, then the fault diagnosis and identification is completed. The simulation results show that permutation entropy can effectively detect the mutations of non-stationary vibration signals, and the fault diagnosis method based on MPE and SVM can effectively identify abnormal situation of the unit and achieve higher diagnostic accuracy. © 2016, China Water Power Press. All right reserved.