Rolling Bearing Fault Feature Extraction And Separation Based On Sparse Optimal Decomposition Theory

Rolling bearings are key components of rotating machinery,and have a wide range of applications in industrial scenarios.However,it often operates under non-ideal conditions,and as a result,various failures often occur.Any potential problem can lead to equipment failure,bring economic losses,or even cause safety accidents.Therefore,it is crucial to research and develop monitoring methods that can diagnose bearing health status information in a timely and accurate manner.Based on the theory of sparse optimal decomposition,this dissertation carries out research on weak fault feature extraction and composite fault feature separation in rolling bearing fault diagnosis,and the core work is as follows:(1)The research status and development trend of weak fault feature extraction method and compound fault separation technology of rolling bearings at home and abroad are summarized.The application of sparse optimization decomposition core theory in rolling bearing fault diagnosis is described and the main research content of this paper is summarized.(2)The common failure forms of rolling bearings are discussed,the characteristics of the law and vibration signal when the failure occurs are studied,and the inter-group sparse intra-group clustering qualities of the rolling bearing failure period shock signals are revealed,indicating the natural adaptability as well as the practical significance of using the sparse characterization method to extract the shock components in the vibration signals.The overcomplete dictionary construction method and the sparse coefficient solution theory are also highlighted.(3)For the problem of weak fault feature extraction of rolling bearings under low SNR conditions.To inherit the excellent feature extraction ability of the classical sparse diagnosis model and at the same time improve its problems such as more inconsistent penalization of sparse coefficients,large computational effort,and difficulty in selecting the representation dictionary,a rolling bearing fault diagnosis method based on the sparse representation of period weighted wavelets is proposed.The method explores the weighting theory of sparse coefficients and proposes a period enhancement strategy based on the tuning Q-wavelet dictionary and nonconvex regular penalty function.Firstly,the average kurtosis index is introduced as a weight to distinguish the contributions of wavelet coefficients of each component of the signal,aiming to balance the penalty term on the penalty degree of sparse coefficients;Secondly,from the perspective of fault signal period mining,a period enhancement strategy is proposed based on the improved envelope harmonic product spectrum period estimation method and embedded into the sparse representation model to achieve the period enhancement of the filtered signal;Finally,the obtained sparse signals are envelope-demodulated and the corresponding features are extracted to achieve fault identification.The numerical simulation signals of bearing faults and two engineering simulations of the measured signals are used to verify that the proposed method has good diagnostic efficiency and feature extraction performance.(4)For the problems of weak early composite fault characteristics and difficult separation of feature coupling in rolling bearings,a parameter improvement-based RSSD-OCYCBD rolling bearing composite fault separation method is proposed.Firstly,the RSSD resonance sparse component separation algorithm is used for noise reduction pre-processing to achieve the separation of high-vibration harmonic components and the acquisition of low resonance shock components containing the main transient shocks of the bearing.The obtained low resonance components are filtered out the invalid subbands below the threshold value according to the proposed subband multiscale energy screening principle,and the remaining subbands are sparsely reconstructed to obtain the optimal low resonance components containing the most information about the fault shocks;Secondly,the optimized OCYCBD algorithm is used to filter the obtained optimal low-resonance components to achieve the fault shock feature enhancement;Finally,the obtained feature-enhanced signals are envelopedemodulated to extract and separate each fault feature for composite fault detection.The numerical simulation and the analysis of the results of two composite fault engineering signals show that the proposed method can perform nonlinear decoupling of the composite fault signals,which is conducive to extracting and separating the fault features in the composite fault signals and realizing early composite fault diagnosis of rolling bearings.