Research On Bearing Fault Diagnosis Based On Wavelet Packet And Extreme Learning Machine

Rolling bearings are the key components of industrial machinery systems,and are applied to more and more areas,the working status of the merits of the mechanical system determines the overall performance,in order to ensure the normal operation of machinery and equipment to be real and effective rolling bearings Fault detection.Common fault extraction methods are Fourier transform,wavelet analysis,set empirical mode decomposition,wavelet packet analysis as a new fault analysis method is also more and more applications to them.Wavelet decomposition decomposes the low frequency part of the signal frequency band,and the wavelet packet decomposition is the decomposition of the upper frequency band.All the frequency bands are decomposed further.It can be seen that the wavelet packet transform is the perfect complement of the wavelet transform and improves the signal The resolution of the high frequency band.At present,wavelet packet analysis has been widely used in mechanical fault diagnosis,power system fault diagnosis,image processing and many other fields.Based on the analysis of the existing fault diagnosis methods,this paper focuses on the fault diagnosis method based on wavelet packet and limit learning machine.The work is as follows:(1)This paper introduces some common methods of fault diagnosis of rolling bearing,and introduces the basic structure and characteristics of rolling bearing.It includes the vibration type,natural vibration frequency,vibration model and fault characteristic frequency of rolling bearing.(2)This paper mainly introduces the basic principle of wavelet and wavelet packet analysis,and introduces the time domain analysis Rolling bearing fault feature extraction method,and the fault feature extraction method of rolling bearing based on frequency domain analysis,this paper analyzes its different characteristics in time domain and frequency domain,and describes its realization method in detail.(3)This paper introduces the basic theoretical knowledge of the limit learning machine and describes the concrete realization method of the limit learning machine in the field of fault diagnosis of rolling bearing.The time domain analysis of the original signal of the rolling bearing is carried out,and the time domain characteristic parameter is obtained.The corresponding time domain eigenvector is constructed as the input parameter of the limit learning machine,so as to realize the faultclassification of the rolling bearing.The original signal of the rolling bearing is decomposed by wavelet packet,and the characteristic values of each frequency domain of the subband are extracted and analyzed.Then,the corresponding frequency domain eigenvector is constructed as the input parameter of the limit learning machine to realize the fault classification.Finally,the results of the fault diagnosis in the two cases are analyzed and compared.Through MATLAB simulation results,this paper considers that the fault diagnosis method based on wavelet packet and limit learning machine can extract the frequency domain characteristic information well,and it can be used for accurate classification and judgment,and has good theoretical research value and application value.