Research On Fault Diagnosis Method Of Wind Turbine Based On HHT And SVM

The research on fault diagnosis of wind turbine is mainly about the analysis ofvibration signal of the equipment. Most vibration signals of the mechanical faults arenonlinear and non-stationary，therefore the key part of fault diagnosis is processingnon-stationary signal.Firstly, this article introduces the theoretical basis of wavelet analysis, WaveletPacket (WPA) transform and Hilbert-Huang transform (HHT), and the fault featureextraction algorithm that establishes on the base of the theoretical basis. By means of thewavelet analysis, the energy of frequency bands generated by wavelet decomposition andreconstruction of the vibration signals in different fault states is normalized aseigenvectors; Wavelet packet transform which uses the adaptability of wavelet packetdecomposes the vibration signal into different frequency bands, and normalizes thecalculated energy of each frequency band and treats it as a eigenvector; Hilbert-Huangtransform is using the EMD decomposition method to decompose the signal into a seriesof IMF contained single feature scale, extract the first few IMF components and treatthem as eigenvector of the fault signal. Taking MATLAB software as simulation platform,this article through programming achieves extraction algorithm of three fault features,and to a fault signal, for example, the three algorithms were simulated.Secondly, construction fault diagnosis model of wind turbine based on LeastSquares Support Vector Machine(LS_SVM) is applied to the device state recognition.Finally, this article that takes the vibration signals of the three kinds of fault state ofa low-speed gear as a example, uses the aforementioned three feature extractionalgorithm for extracting the fault feature of each signal and uses these extractedeigenvectors as three sets of training and test samples. Get corresponding LV_SVM faultdiagnosis model by using three sets of training samples, and use the corresponding testsamples to test the diagnostic accuracy of the LS_SVM model. The results prove that thefault diagnosis method based on HHT and SVM has good diagnostics effect, provides agood reference for the practical application of wind turbine fault diagnosis.