Rolling Intelligent Fault Diagnosis Based On Wavelet Theory Methods

Rolling bearing is the common component in machinery. Its running state can influence the performance of the whole machine directly. According to statistics, in the rotating machinery fault at the scene, 30% of faults are caused by the rolling bearing. It is of great realistic significance to diagnosis the fault of rolling bearing accurately and reliably. This paper proposed a new fault diagnosis method of rolling bearing using improved BP neural network.Firstly, systematically expounded the vibration of rolling bearings and the vibration characte of their typical failure. According to the different failure caused by different forms of vibration, bulding vibration signal model with single defect on each component and analysis fault charactera of the vibration signal. And then, the use of bearing vibration signal to diagnose rolling bearing is feasible.Secondly, study and analyze the wavelet transform at different aspects of fault diagnosis of rolling bearing applications: denoising; singularity detection. Rolling element bearing vibration signal is vulnerable to random noise pollution, how to denoising becomes one of the key issues about rolling element bearing fault diagnosis. The traditional denoising methods may treated the useful signals which have small energy as noise, and eliminated it. Then this paper expressed the wavelet denoising method. In the area of denoising, the wavelet denoising improved the signal noise ratio perfectly and provided a reliable basis of fault diagnosis decision-making.In addition, wavelet singularity detection can overcome shortcomings that the other methods are not sensitive to detect the rolling faults. It can be effectively used for rolling bearing state detection and fault diagnosis.Finally, aimed at shortcomings in BP algorithm, such as convergence rate, the smallest partial paralysis and so on, a new rolling fault diagnosis method was produced. Wavelet transformation was used to extract feature of dynamic vibration information, and then input to BP neural network to train the network with conjugate gradient algorithm. Moreover, appling the BP neural network to identify the failure of the rolling bearings. By analyzed on the rolling bearing vibration signals, the experiment results showed that, compared to other methods, the new method is faster to converge and has a higher precision in fault diagnosis applications, and can accurately realize the rolling fault diagnosis.