Feature Extraction Of Rolling Bearing Based On Singular Value Decomposition

Rolling bearing is one of the most important parts in mechanical devices.The key operating components such as inner ring,outer ring and rolling element are complex in structure and bear large load.Whether the running state of the bearing is normal or not is directly related to the working state of the whole mechanical equipment.Once the bearing fails,it may cause the entire unit to run down and cause unnecessary losses.Through the effective rolling bearing fault diagnosis technology,the actual working state of the rolling bearing can be evaluated,the service life of the bearing can be predicted,unnecessary maintenance and maintenance can be reduced,and the production economic benefit can be improved.Based on the singular value decomposition technique,this paper studies the bearing fault feature extraction.The main contents are as follows:Firstly,according to the actual application status,important role of the rolling bearing and the characteristics of the singular value decomposition(SVD)algorithm,the background and significance of the topic are introduced in detail.Then,the domestic and international research status of singular value decomposition(SVD)and the future development trend were comprehensively expounded.Lastly,the signal processing flow of SVD was introduced through experiments,and the characteristics of SVD algorithm were explained in detail.Secondly,aiming at the formation method of component signals after SVD,the difference of signal processing effect and fault feature extraction effect between simple method and average method was studied.The loss signal rate and the signal-to-noise ratio were used to evaluate the effect of the simulated signal processing of the two component formation methods.The fault feature extraction capability of the two modes was verified by the measured signal.The research results show that the sinusoidal signal and the amplitude modulation signal have the same effect;for the variable frequency signal,the average method is obviously due to the simple method;both methods can effectively extract the fault characteristic frequency.But the computation amount of the average manner is much larger than that of the simple one.It is recommended that except frequency change signal,for other common signals the simple method should be prior used.Thirdly,in order to the problem that that the effective order of singular value is difficult to determine in SVD noise reduction and the energy operator method is susceptible to noise interference,a bearing fault feature extraction method based on singular value fluctuation difference spectrum and Teager energy operator was presented.Firstly,the vibration signal was constructed as a Hankel matrix,and then the matrix was decomposed by using Singular Value Decomposition,the singular value obtained by the decomposition was used to obtain the fluctuation difference spectrum.Finally,the Teager energy operator is used to demodulate the reconstructed signal after noise reduction.The simulation data and the actual bearing fault data analysis showed that the method can effectively improve the signal-to-noise ratio and extract the fault characteristics.Finally,aiming at the problem that the resonance demodulation method was susceptible to noise interference and the parameter of band-pass filter in traditional resonance demodulation was difficult to determine and depended on the subjective experience of human,a new method based on singular value decomposition and resonance demodulation was proposed.Firstly,the bearing vibration signal was decomposed into four components by using the singular value decomposition algorithm,and the kurtosis value of each component was recalculated.The component with the largest kurtosis value was selected and the center frequency and bandwidth of the bandpass filter were automatically determined by using the spectral kurtosis algorithm.The filtering and envelope demodulation were conducted for the component signal.The results show that the extraction performance test and robustness test on this new method improves that this method can adaptively determine the filter frequency band,reduce the influence of noise interference,and has good stability in the case of band-pass filter failure.The comparison with the variational mode decomposition(VMD)algorithm shows the superiority of the method.