Application Research Of Wavelet Neural Network Technology In The Shield Machine Fault Diagnosis

In recent years, the rotating machines are becoming larger, more intelligentand more complex. These devices for expanded state monitoring and faultdiagnosis technology have become even more important. Due to its blend ofnonlinear mapping ability of neural network technology, the self-learning and theadaptive ability, the generalization and fault tolerance, and the wavelet analysistechniques in border processing and filtering, and the time-frequency analysis, andthe signal to noise separation and advantages of detection of a weak signaladvantage for the fractal index, the identification and diagnostic signals and themulti-scale edge detection areas, and the fault diagnosis based on wavelet neuralnetwork technology and so on, all of them have become major focus of researchuniversities,and the research institutes.How to use the wavelet neural network onengineering applications has become the focus on wavelet neural networkstudy.This article based on the wavelet neural network study the real-time signalscollected from shield machine main drive motor vibration,to explore theapplication of wavelet neural network in the shield machine fault diagnosis.Themain work is done as follows:View book about the constitute of shield machine,the classification and theway of shield machines work,learn about the history and development of theshield machine.Focus on understanding the working principle of the maincomponents of the shield machine, find related fault type, fault characteristics andsummarized into a table.Do the groundwork for the next follow-up study.Analysis the difference between wavelet analysis and Fourier transform,raised the advantages of wavelet transform, compared with the Fourier transform,on dealing with nonlinear signal data in the frequency domain and time domain.Processing vibration signal gathering from the shield machine, demonstrated thefeasibility of the wavelet processing shield machine vibration signal.View the development and application of neural network,try to understandthe principle and its application,learn about the lack of the neural network and use the model on the data collected from the shield machine,found the weaknessand lack of the BP neural network model in dealing real-time acquisitionGiven the lack of BP neural network, think of the two combined, usingmorlet wavelet instead the BP neural network activation function, and builtwavelet neural network to processing vibration signal acquisition. The speed oferror convergence and computing are much better than BP neural network.The result shows that wavelet neural network can be used for the shield faultdiagnosis. In the meanwhile, this article still shows the lack of the study andmakes the work left clearly.