Reliability Evaluation And Life Prediction Of Vehicle Axle Box Bearings

As one of the most vulnerable key components in the train transmission system,axle box bearings have extremely complex and harsh operating conditions.Once a fault occurs,it will seriously affect the safety of train operation.Therefore,it is of great significance to determine the operating status of axle box bearings through accurate operational reliability assessment and residual life prediction methods,in order to reduce the occurrence rate of train operation faults and avoid catastrophic accidents.In this regard,this thesis takes the axle box bearing life test data of a vehicle as the research object,adopts the signal processing and feature extraction technology,and based on the surrogate model method and intelligent optimization algorithm,the following work is carried out around the key issue of how to establish an accurate operational reliability assessment and residual life prediction model:(1)Vibration signal processing and performance degradation feature extraction of axle box bearings.Using singular spectral decomposition to decompose the vibration signal of each measurement period,several singular spectral components are obtained,and abnormal components with high noise are removed by calculating the arrangement entropy of each component.Combining kurtosis and information entropy criteria,a new screening index is established to determine the optimal singular spectrum component that is most sensitive to vibration and shock and has the lowest noise content.Based on the optimal singular spectrum component,the high-dimensional features in time domain and frequency domain are extracted,and the characteristics that can sensitively and stably reflect the growth of bearing degradation are determined through monotonicity and robustness indicators.The performance degradation feature set is constructed,which lays a foundation for improving the accuracy of subsequent bearing operation reliability assessment and life prediction.(2)The bearing operation reliability evaluation based on IDWPSO-FSVDD.The initial SVDD hypersphere model is trained with the bearing performance degradation feature set as the input variable to obtain the sample spherical centroid distance reflecting the bearing performance degradation.A fuzzy affiliation function is set with the sample spherical center distance as the variable,and the FSVDD model is constructed by assigning fuzzy affiliation to the sample,and the operational reliability calculation index is determined.The dynamic adjustment of inertia weights and differential evolution operations are used to improve the basic particle swarm algorithm,which is prone to the problem of falling into local extremes and the problem of insufficient population diversity in the late iteration.Based on the model proposed in this thesis,the operational reliability of the bearing is evaluated,and the performance change pattern of the bearing during service is obtained,so that a more accurate and reasonable life prediction interval can be determined for the subsequent work of remaining life prediction.(3)The remaining life prediction of bearings based on GA-MKL-SVR.The typical kernel functions of the SVR model are tested for single and mixed characteristics.After analysis,the RBF and polynomial kernel functions with the best mixed characteristics are fused to construct a multi-kernel learning SVR model that combines the advantages of different types of kernel functions.Taking the key parameters of the model as variables and the minimum mean square error as the goal,the optimal parameter combination is determined based on the CV-GA method,and then the mixing factor is determined to establish a mixed kernel matrix to construct a multi-kernel learning SVR model.Finally,based on the life prediction interval determined by the evaluation results of bearing operation reliability,the residual life prediction of bearings is carried out by using this method and single-core SVR method respectively.The relative error and variance of the obtained results are compared and analyzed to verify the superiority of this method in prediction accuracy and stability.This thesis proposes an accurate and reasonable axlebox bearing operational reliability assessment and life prediction method,constructs a set of performance degradation characteristics reflecting the axlebox bearing,reveals the performance change characteristics of the axlebox bearing in service,and obtains a more accurate life prediction result of the axlebox bearing.It provides a certain reference value for the condition monitoring and fault warning of axlebox bearings in operation to further improve the safety of train operation and make more reasonable maintenance and replacement decisions,so as to improve the current situation of low effective utilization rate and high maintenance cost of axlebox bearings.