Research On Health State Assessment And Remaining Useful Life Prediction Of Rolling Bearings

Rolling bearings act as the key parts to convert motion and transfer power in high-speed trains,urban rail vehicles,heavy-haul trains and engineering vehicles.The performance of rolling bearing directly affects the safety and reliability of mechanical equipment.The fatigue damage of rolling bearings may cause mechanical system failure,reduced operation stability,and even serious safe accidents.Therefore,it is necessary to study advanced health state assessment and remaining useful life(RUL)prediction methods,which is of great significance to ensure the safe operation of carrying tools.Taking the accurate assessment of rolling bearing health state and the reliable prediction of RUL as the research goal,this paper studies the rolling bearing health state assessment method and RUL prediction method.The specific works are as follows:(1)To solve the problem that the traditional sparse measures are weak in evaluating the health state at the original scale,a health state assessment method based on multi-scale sparse measures fusion indictor is proposed to realize the timely warning of early fault and accurate health state assessment of rolling bearings.First,multi-scale analysis combined with sparse measures is proposed as the feature of bearing health state,and adaptive weighted signal preprocessing technique(AWSPT)is introduced to suppress signal noise,which strengthens the early warning of weak faults and improves the representation ability of features.Then,multidimensional representation feature is fused by diversity to obtain one-dimensional indictor.Aiming at the uncertainty of scale parameters in the difference degree fusion,the scale parameters are determined based on 3 a criterion,and the difference degree is converted into the interval[0,1],which realizes the quantitative evaluation of the rolling bearing fault severity.Two cases verify the effectiveness and superiority of the proposed method.(2)The prediction of rolling bearing RUL is based on the performance of the constructed degradation indictor.A monotone indictor construction method considering the removal of characteristic burr is proposed to improve the monotonicity and trend of the fusion indictor.In view of the characteristic burr which deviates from the expected degradation trend in the rolling bearing degradation state characterization feature,a burr correction technique is developed to detect and remove the burr in the characterization feature and improve the performance of the degradation indictor.Then,principal component analysis(PCA)is used to fuse multi-dimensional feature features,remove the redundant information in the original state feature space,maintain the global structure of bearing degradation state data,and further use the exponentially weighted moving average(EWMA)algorithm to smooth the fusion indictor to obtain high-quality degradation indictor.The run-to-failure experimental results verify the superiority of the proposed method in monotonicity and trend,which lays a solid foundation for follow-up RUL prediction.(3)In view of the interference of random noise and not considering the degradation characteristics of rolling bearings in the original data-and-model-driven method,an improved data-and-model-driven RUL prediction method is proposed to realize the accurate RUL prediction of rolling bearing.Firstly,considering the degradation characteristics of rolling bearings,the first prediction time(FPT)is determined based on 3σ criterion,which is used to trigger the prediction process of rolling bearings.Then,aiming at the random fluctuation of the degenerate model,the process equation and observation equation are constructed based on the exponential model structure of the Wiener process.The RTS smoothing filtering algorithm is embedded to suppress the process noise and measurement noise,reduce the influence of the random fluctuation of the degradation curve,and improve the prediction accuracy of the RUL.The superiority of the prediction ability of the proposed method is verified by simulated random data and run-to-failure experiment.In summary,based on the accurate evaluation of the health status of the rolling bearing and the reliable prediction of RUL,this paper takes the research on the improved health state evaluation method and RUL prediction method.Effectiveness and superiority of the proposed method have been well verified in rolling bearing run-to-failure experiments,which has important theoretical significance and practical value for ensuring the safe operation of rail vehicles and realizing intelligent operation and maintenance.