Research On Life Prediction Of EMU Battery

With the development of high-speed EMUs,information technology has been applied into all life cycle of one trainset.The information systems of EMU operation and maintenance have been established and improved constantly.With the reform of railway equipment maintenance theory,the current preventive maintenance is gradually transferring from planned preventive maintenance to condition-based maintenance.This requires the full use of historical and monitoring data from EMU operation and maintenance systems,with combination of prognostics and health management technology,to effectively analyze the performance status and remaining useful life of the key components,thus meeting the requirement of intelligent maintenance of EMUs.Capacity is one of the direct indicators of battery performance.The current value and changing trend of capacity correspond to the current health status and life trend of the battery,respectively.This paper takes the battery of the auxiliary power supply system,which is one of the key components of the EMU as the research subject.It aims to use the related monitoring data to predict the capacity and remaining useful life of the battery online,so that we can understand the battery status,realize the optimization of the maintenance strategy,reduce the maintenance cost,and find the balance between economy and reliability.Combined with the actual working condition of the EMU battery,this paper mainly studies the following problems:(1)Considering the battery capacity recovery effect and the external environmental influence,new indirect health factors are proposed to characterize the battery's life status,so that the problem that the capacity as a direct health factor cannot be measured is solved.In view of the shortcomings of the traditional RVM algorithm,the hybrid kernel function is used to improve the learning ability and generalization ability of RVM.The cuckoo search algorithm is used to optimize the RVM kernel parameters.The chaos theory and boundary constraints are used to dynamically improve the cuckoo search convergence speed.The designed ICS-HKRVM is used to improve the accuracy of online estimation of the battery capacity.(2)Predict the decline trend of battery capacity and remaining useful life online.The capacity data estimated by ICS-HKRVM is reconstructed as time series data.For the problem of long-term error accumulation in the process of iterative multi-stepprediction,ICS-HKIRVM,which is improved with incremental learning idea,is designed so that the model can be updated online with the newly added data,ensuring long-term prediction accuracy without adding burden to the calculation memory.(3)The application of the actual monitoring data of the EMU battery proves the effectiveness of ICS-HKRVM and ICS-HKIRVM,and can accurately predict the current state and remaining useful life of the battery.It provides theoretical support for the intelligent and economic optimization of the current EMU battery repair and maintenance process.