Research On The Evaluation Of The Health Status And The Prediction Of The Remaining Life Of Rail Transit Bearings

As an efficient and environmentally friendly public transport,rail transit has become an indispensable part of urban transportation.With the increase of operating mileage and passenger volume,the safety of rail transit becomes more and more important.As one of the key components of rail transit vehicles,rolling bearing is widely used in the running parts of vehicles,motors and other equipment,and its operation status will directly affect the safe operation of rail transit.Therefore,it is of great significance to evaluate the health status and predict the remaining life of rail transit bearings to ensure the safe operation of rail transit.In this paper,rail transit bearing is taken as the research object,and the fault mechanism analysis,degradation state feature extraction,health state evaluation and residual life prediction of bearing are systematically studied.The main contents are as follows:(1)The failure mechanism theory of rail transit bearing is studied.This paper analyzes the basic structure,failure form and reason of the bearing and the characteristic frequency of the bearing theory,and briefly describes the state monitoring method of the bearing and the whole life experimental data set,which provides the theoretical basis for the later research.(2)The feature extraction method of bearing degradation state based on multi domain is studied.In order to solve the problems of insufficient evaluation ability of single or single domain degradation state feature in bearing health state evaluation and insufficient accuracy in remaining life prediction,a multi domain degradation state feature extraction method is proposed.Firstly,the time-domain and frequency-domain degraded state features are extracted respectively,and the time-frequency-domain degraded state feature extraction method based on the variational mode decomposition,envelope demodulation and singular value is mainly studied.By analyzing the life-cycle change curve of each degraded state feature,it is pointed out that it is difficult for a single degraded state feature to evaluate the health state of the bearing.(3)The evaluation method of bearing health based on t-SNE and nuclear markov distance is studied.In the process of bearing health assessment,it is difficult to select the characteristics of degradation state and build the health index.Firstly,random forest algorithm is used to select the degradation state characteristics with high importance,and to build the high-dimensional relative degradation state characteristics.Then,in order to prevent the degradation state feature redundancy from affecting the evaluation results,t-SNE is used to reduce the dimension of high-dimensional relative degradation state feature set,and the degradation state features are fused.Then the health index is introduced to evaluate the health status of the bearing,and the validity of the proposed method is verified by the vibration data of the bearing life cycle.(4)The prediction method of bearing remaining life based on PSR and ELM＿AdaBoost is studied.In view of the low accuracy of the data-driven remaining life prediction model,based on the extreme learning machine,a remaining life prediction model based on phase space reconstruction(PSR)and ELM＿AdaBoost is proposed.Firstly,the influence of different model structure parameters on the accuracy of prediction results is analyzed,and the best prediction model structure parameters are determined.Then,taking the full life cycle vibration data as an example,the remaining life of the same bearing and different bearings are predicted.Finally,the comparative analysis of different prediction models verifies the generalization ability and effectiveness of the method in this paper.