Research On Total Life And Residual Useful Life Prediction Method Of Freight Coupler Based On Data Driven

Because of its forward-looking,high maintenance efficiency,safety and reliability,condition-based maintenance has gradually replaced planned preventive maintenance as the main maintenance mode of railway freight cars at home and abroad.The fatigue life prediction of key parts of railway freight cars is the basic element of condition-based maintenance research.As one of the key parts of heavy haul railway freight cars,couplers are easy to produce fatigue cracks and seriously affect the safety of train operation due to the influence of road conditions,train load,structural technology and other factors in practical service.Therefore,it is of great significance to study the total life and residual useful life of heavy haul freight couplers,and propose the safety limit of coupler service mileage and residual useful life prediction model for improving train safety and reducing maintenance costs.In this thesis,in view of the fatigue life prediction method of heavy haul freight couplers,the following researches were done:Firstly,the failure analysis of the coupler was carried out based on the combination of field investigation,simulation analysis and bench test.At the same time,the data about failure mode,total life and crack propagation of coupler were obtained,which provides a data basis for the prediction of the total life and residual useful life of the coupler based on data driven.The failure analysis shows that the dangerous parts of coupler knuckle are the root of the upper and lower traction flange and the middle part of the inner wrist surface,the maximum damage is located at the root of the lower traction flange.The service life of coupler body is usually much longer than that of the coupler knuckle.Then,the total fatigue fracture life of the coupler knuckle was studied based on the fatigue applied statistics theory,and the total fatigue failure life of the coupler body was studied based on the reliability analysis method for the small sample failure data.The total fatigue fracture life of the coupler knuckle satisfies the lognormal distribution.When the confidence is 95% and the survival rate is 50%,the total life of coupler knuckle is 1.224 million km.The fatigue fracture life in the case of inner wrist surface fracture is 1.04～1.75 times longer than that of the lower traction flange fracture,and the difference in life is derived from accidental error but also conditional error.The reliability curve based on the analysis method for small sample is closer to the empirical reliability curve derived from original large sample.Based on the analysis method,the total life of the coupler with 50%reliability is about 2.1 million km.Finally,three residual useful life prediction methods of couplers based on data driven were proposed,and their prediction effects were compared and analyzed taking the coupler body as an example.For structures with large life dispersion,the residual life prediction method only considering the common characteristics of the research object has a poor prediction accuracy(the average relative error is 57.56%),but the operation is convenient and simple.The residual useful life prediction method based on delay-time concept and hypothetic distribution check combines the current state,and the accuracy is improved(error is 34.52%),but the prediction value of residual useful life varies sharply with the mileage in service.The residual useful life prediction method based on delay-time concept,support vector regression and Kalman filtering takes into account the process of performance degradation,so the change of prediction value of residual useful life is smoother,but the prediction accuracy is slightly lower(error is 43.67%).In this thesis,the whole life prediction methods for coupler fatigue fracture based on classical statistical theory and SMOTE-Bootstrap-Bayes method were proposed,and the residual useful life prediction methods of coupler based on condition monitoring data were proposed.The research results have a good theoretical support for the whole life and residual useful life prediction of railway freight car’s key structures,and provide corresponding application reference for train operation safety,economy and the implementation of condition-based maintenance.