Based on Hilbert-Huang Transform and BP Neural Network Fault Diagnosis of Rolling Bearing

Rolling as a mechanical device one of the most common parts, it runs the state will directly affect the function of the whole machine, so the technology of rolling bearing fault diagnosis is very important. Currently used for bearing fault detection methods can be divided into many types, but the most commonly used diagnostic method or vibration, this is done as a basis for some research. In this paper, gearbox fault diagnosis of rolling bearing as the background, the neural network of strong nonlinear mapping ability, and self-learning, adaptive and self-organization ability of studies of the Hilbert-Huang Transform and the combination of BP neural network rolling element bearing fault diagnosis method. Paper describes the bearing fault diagnosis technology development status at home and abroad; introduced bearing fault-related failure mechanism and the performance of several forms of the normal vibration and bearing vibration signal and failure to conduct a comprehensive analysis of the signal; discussed the use of EMD and wavelet packet transform fault feature vectors are two ways to extract the parameters and as a neural network input vector; designed based on the Hilbert-Huang transform and BP neural network for rolling bearing fault diagnosis system. Finally, the collected data acquisition system contains different working conditions of vibration signals measured data, were constructed bearing vibration parameters based on the normal and fault state vectors of training samples, using the neural network has already been built for training and use trained neural network running on each bearing an intelligent diagnosis. Experimental results show that the use of the diagnostic system can be very good for effective bearing vibration analysis, it can effectively identify the bearing of the different working conditions. 

Keywords: bearing, fault diagnosis, BP neural network, Hilbert-Huang Transform