Research On Performance Degradation Evaluation And Remaining Life Prediction Method Of Rolling Bearing

As the core parts of rotating machinery,rolling bearings play an important role in many fields,such as intelligent manufacturing,transportation,and so on.Any degree of failure damage will directly affect the normal operation of equipment,and even lead to the paralysis of the whole mechanical system,causing huge economic losses.Therefore,timely understanding the performance degradation status and remaining useful life of rolling bearings and formulating reasonable preventive maintenance strategies are of important engineering significance for guiding industrial production.Therefore,this thesis selects rolling bearings as the research object and conducts in-depth research on performance degradation evaluation and remaining useful life prediction methods from a data driven perspective,mainly including the following contents:(1)Aiming at the problem of weak fault characteristics caused by a large amount of noise in the full life data of rolling bearings collected by sensors,this thesis studies a noise reduction algorithm based on iterative symplectic geometry packet decomposition(ISGPD).The core of ISGPD is to optimize traditional symplectic geometric mode decomposition using dual iterative algorithms to improve its decomposition defects in noisy environments.In symplectic geometric packet decomposition,singular value decomposition(SVD)is used instead of traditional QR decomposition to avoid decomposition errors,and wavelet packet decomposition is combined to perform inner iteration noise reduction to obtain bearing fault features;In order to avoid feature loss caused by single decomposition,an outer iteration algorithm is proposed,using the difference between the original signal and the acquired component as the iterative input to ensure the integrity of fault features.The research results show that the proposed method can effectively filter out background noise and achieve feature extraction of weak bearing faults.(2)In order to solve the problems of difficult evaluation of rolling bearing degradation performance and identification of degradation status,this thesis studies a new method for identifying rolling bearing degradation status based on hierarchical timeshifted multiscale fluctuation dispersion entropy(HTMFDE)and improved convolutional neural network.In this thesis,HTMFDE is used to obtain bearing degradation characteristics at different scales and levels,and the Jensen-Renyi divergence(JRD)between health and fault state characteristics is calculated to comprehensively evaluate the performance degradation of bearings;In addition,this thesis combines the idea of local amplitude to obtain sample data with different degrees of degradation,and trains an improved convolutional neural network model to identify different degradation states of bearings.The experimental results show that the method proposed in this thesis can effectively evaluate bearing degradation performance and accurately identify bearing degradation status.(3)Aiming at the problem of difficult to accurately predict the remaining service life of rolling bearings,this thesis studies the application of slow-varying damage-assisted two-stream neural networks(SD-TSNN)to predict the remaining service life of rolling bearings.In this thesis,phase space warping(PSW)theory is used to obtain auxiliary data characterizing the slow-varying damage of rolling bearings.A two-stream network SDTSNN was established to extract deep degradation features from raw data and slowvarying damage data,respectively,to comprehensively reflect the health status of bearings.Then,the fused features are used to estimate the remaining useful life of the rolling bearing.The experimental results show that the prediction model proposed in this thesis has higher accuracy and better robustness compared to other prediction methods.