Research On Sparse Diagnosis Method Of Rolling Bearing Faults Based On Dictionary Learning

Rolling bearing is a key support component in major equipment.Due to long-term operation under complex and severe working conditions,rolling bearing failures frequently occur,which seriously affects the safety and stability of equipment operation.How to quickly and accurately detect and diagnose from the vibration signal with a large amount of noise interference,perform accurate and timely diagnosis of the health status of the equipment,and implement the maintenance strategy of "early detection of early failures,early prevention",those are of great significance to improve the operating reliability of equipment and avoid major economic losses.The periodic impact feature in the vibration signal is an important symptom of the failure of the rolling bearing,but the impact feature is sensitive to noise and susceptible to random interference,which increases the difficulty of extracting weak feature information.Therefore,thanks to the strong representation ability of sparse representation for complex signal components,this paper aim at extracting rolling bearing fault features and achieving effective diagnosis of rolling bearing,based on dictionary learning carry out the sparse diagnosis method of rolling bearing faults.The main work of this paper is as follows:(1)Aiming at the problems of the classic K-SVD dictionary learning algorithm being susceptible to noise,the learned atoms are difficult to accurately match the impact characteristics,a fault diagnosis method of rolling bearing based on TVF-EMD and K-SVD dictionary learning is proposed.This method uses the TVF-EMD method to extract the sensitive components with periodic impact characteristics to weaken the influence of noise on the K-SVD algorithm.And uses cluster analysis to remove atoms in the K-SVD dictionary that have nothing to do with the impact shape.Thus,the accuracy of the sparse representation of the rolling bearing fault feature is improved,and the effective identification and diagnosis of rolling bearings is realized.Numerical simulations verify the proposed method,and apply it to the fault diagnosis of rolling bearing of an electric locomotive.The effectiveness and superiority of the method was verified by numerical simulation,and it was applied to the fault diagnosis of rolling bearing of a certain electric locomotive.(2)Aiming at the problem that the K-SVD dictionary learning algorithm is difficult to accurately obtain the weak impact characteristics in the vibration signal under strong background noise,the accuracy of the early weak fault identification of the rolling bearing is low,a fault diagnosis method of rolling bearing based on adaptive optimal Laplace wavelet dictionary is proposed.This method combines K-SVD dictionary learning with Laplace wavelet parameterized dictionary,the reference atoms with significant impact characteristics are selected from the dictionary learned from K-SVD algorithm.Based on this,the parameters of Laplace wavelet are optimized by using whale optimization algorithm,and a parameterized dictionary matching with the impact response of rolling bearing faults is constructed,so as to realize the coefficient extraction of weak features and fault identification of rolling bearing.The effectiveness and superiority of the method was verified by numerical simulation,and it was applied to the fault diagnosis of bearing outer ring of SQI test bench.(3)On the basis of the above research,developed a rolling bearing fault diagnosis system based on the Lab VIEW software platform,contains basic functions such as time domain and frequency domain analysis of vibration signals,fault warning and fault identification.The research method proposed in this paper was embedded in the developed system,and realized the fault identification of the rolling bearing of the wind turbine generator.