Research On Diagnosis Of Bearing Based On Double Tree Complex Wavelet And Broad Learning System

As a kind of standardized parts,the roller bearing is widely used in various kinds of mechanical equipment,so it is of great practical significance to diagnose the fault diagnosis of the bearing.In this paper,the paper discusses the characteristics of the characteristic extraction and fault identification of the rolling bearing fault diagnosis,and puts forward the fault characteristics extraction theory based on the multi-dimensional relative energy representation method based on the double tree complex wavelet transform.The following work is done in this article:First,the theory of double tree complex wavelet transform(DTCWT)is studied in this paper.The signal is divided into the real and virtual parts,and the filter operation is divided into different subbands of different frequencies.At the same time,in the practical application,the frequency mixing of the frequency is overcome,and the characteristics of translation invariant and complete decomposition are carried out,and the double tree complex wavelet method has broad application prospect in the field of signal processing.Second,in order to solve the problem of comparison classification between different lengths and different sampling points,a multi-dimensional relative energy representation method is proposed,and the signal is compared in the same dimension.The definition of RGB image in digital image processing is compared with the signal of the double tree complex wavelet,and the multidimensional representation method of the signal is demonstrated.The energy of each segment signal is calculated from the energy Angle of the signal,and the relative energy of each string is relative to the original signal.The relative energy is extracted when the signal characteristics are extracted,and the original signal is carried out,and the noise reduction and compression are carried out.Finally,compared with the existing fault pattern recognition method,the width learning system(BLS)is used to build the network model by using an incremental algorithm of the horizontal expansion of the network,which can effectively solve the problem of the low success rate of the low success rate of the network.Compared with the high precision and fasttraining,the depth neural network has the advantages of simple and easy to operate in the practical application.Based on the theoretical analysis and experimental data,the multidimensional relative energy failure representation method based on the multi-dimension of double tree complex wavelet has the characteristics of translation invariant,dimensional unity,noise reduction and data normalization,and has broad application prospect in the field of fault diagnosis.The width learning system is fast training,and the value matrix of the network structure expansion is adjusted for different data,and the value of further research is studied.