Fault diagnosis for turbine generator based on wavelet packet PCA-SVM

As to the existing shortcomings of the traditional method for fault diagnosis of turbine generator, A new method based on wavelet packet PCA-SVM is proposed in this paper. First of all, take a wavelet packet transformation of the fault samples to extract the energy of each frequency band, and use them as the initial samples, and then make a data compression and feature extraction of the initial samples using the principal component analysis(PCA), eliminating the correlation between data and extracting the principal components which contain sufficient information of initial samples. Finally, we take the principal components as the input vectors of the support vector machines(SVM), this will reducing the dimension of the sample space and computing complexity. Simulation results shows that this method can effectively improve the diagnostic accuracy.