Application of RBF Neural Network Based on Wavelet Packet denosing and EMD method in fault diagnosis for turbine generator

Because the process of fault diagnosis for turbine generator usually contains noise and has a characteristic of strongly non-linear and non-stationary. In this paper, to overcome the deficiency of existing methods, a new approach for fault diagnosis of turbine generator based on wavelet packet denoising ,EMD method and RBF neural network is proposed. Firstly, the fault data of turbine generator is analyzed using wavelet packet to remove the noise; Then the denoised data is disposed by EMD method to extract the frequency eigenvectors of the IMF components, and these eigenvectors were used as the training samples of the RBF network. Finally, use the well-trained RBF network to identify the faults. The simulation experiments show that the proposed method of fault diagnosis for turbine generator is effective and the denosing using wavelet packet transform is essential.