Fault Diagnosis of Rolling Bearing Based on Information Fusion

Abstract

The rolling bearing is widely used as the mechanical rotating parts, whose state directly determines the performance of the whole machine. Since there exist the problems of low accuracy in traditional single sensor detection and lack of stable diagnostic system, acceleration and acoustic emission can be combined to conduct the study on the fault diagnosis of rolling bearing. The wavelet technology should be used to reduce the noise of the signal obtained by two kind of sensors. Hilbert technique is used to demodulate the signal after noise reduction, and the frequency domain envelope will be got. The eigenvector is obtained by calculating the frequency energy of the fault signal. Then Multi-sensor information fusion system is established by BP neural network in which the suitable samples is selected to train until the error meets the requirement, and the diagnosis of rolling bearing fault is realized.