Research On The Defect Diagnosis For Rolling Bearing Based On BP Neural Network

Rolling bearing is one of most ordinary parts in mechanical machine, and its running status is directly related to the safety of machine operation. Whether there is a defect is a standard to assess the quality of bearing. Defect is one of the important factors resulting in bearing fault, so it is significant to study the technology of defect diagnosis for rolling bearing.Vibration come into being while rolling bearing rotated, and the vibration signal is characterized by its steadiness. When rolling bearing has defect, the impulsion power is produced because of other elements striking, and the vibration signal become unsteadiness. Based on the characteristics of vibration signal, we can diagnose the defect of rolling bearing. The vibration signal can be represented in the field of time and frequency, and only the combination of time and frequency characteristics can fully represented the condition of rolling bearing. Most traditional defect diagnosis methods use one field of the vibration signal. Neural network has the capacity to learn, self-organize and self-adapt. A defect diagnosis method based on the BP neural network is built up in this paper.Firstly, as the necessary preparation, the parameters of time and frequency characteristics are analyzed, and the relationship between each parameter and the bearing defect is expounded in the forms of quantity and quality. And then, based on the basic theory of neural network technology , defect diagnosis system for rolling bearing based on BP is built up. Some time field parameters and some frequency field parameters are chosen as the input variables of neural network according to the choosing principle for input variables of it. At the same time, the structure of network is optimized and the learning algorithms are comparatively studied. After that, using the measuring data of two former research subject ,the defect and undefect samples are built based on the vibrating characteristic of rolling bearing. Finally, the BP neural network is trained on MATLAB, and then a experiment is conducted to validate the validity of neural network for diagnosing the defect of rolling bearing. The result shows that BP neural network used for diagnosing the defect of rolling bearing is effective and practicable.