Abstract:In order to improve the diagnosis performance of bearing fault signal, wavelet analysis and RBF neural network were combined to classify the bearing vibration signal. Firstly, wavelet transform was applied to the bearing vibration signal, and the soft threshold denoising method was used to filter out the vibration signal noise. Then, the vibration signal was matrix processed. Then the RBF neural network was constructed, and the bearing vibration signal eigenvector was input, and the weight and threshold were initialized. Finally, the stable RBF neural network fault diagnosis model was obtained through continuous reverse iteration. Experimental results showed that, by setting the number of hidden layer neurons and determining the appropriate scale of RBF neural network, wavelet denoising can effectively improve the accuracy rate of bearing fault identification. Compared with the common bearing fault classification algorithm, the algorithm in this paper had a higher accuracy rate of fault identification.