Remaining Useful Life Estimation For Electric Scooter Based On Stochastic Filter

With the development of society and the improvement of people's awareness of environmental protection,electric vehicle,as a green and convenient means of transportation,has become more and more popular and loved by people.Especially for the electric scooter,it is specially designed for the elderly and disabled people with mobility diff-iculties.Fault diagnosis and prediction technology has experienced decades of development,and has achieved fruitful research results.How to apply the fault diagnosis and prediction technology to the electric scooter system to improve the reliability and safety of the electric scooter has important research significance and practical value.In this dissertation,a fault diagnosis and remaining useful life prediction technique based on the hybrid bond graph and stochastic filter is proposed for the electric scooter.Firstly,based on the hybrid bond graph theory,the hybrid bond graph model of the electric scooter system is established,and the global analytical redundancy relations is derived for fault detection according to the hybrid bond graph model.Then,the nominal values of the system parameters are identified by the system identification method.Based on the global analytical redundancy relations,the mode change signature matrix and the mode dependent fault signature matrix are established to track the working mode and isolate the fault candidates of the system.In order to update the fault candidates to locate the real fault,it is necessary to identify the magnitudes of the fault candidates by means of fault identification method,and compare the results of fault identification with the nominal values of the parameters to determine the real fault.In this dissertation,the extended Kalman filter and unscented Kalman filter are introduced.According to the obvious difference between the evolution speed of system parameters and system states,a double scale unscented Kalman filter is proposed.Finally,the experimental results show that the proposed algorithm can not only ensure the accuracy of fault estimation,but also improve the efficiency of fault estimation.On the other hand,this dissertation studies the remaining useful life prediction of the electric scooter under the intermittent fault.The influence of the change of using condition on the remaining useful life is analyzed and the using condition based remaining useful life prediction scheme under the intermittent fault is conducted in this dissertation.Then,a nonlinear curve fitting based degradation model selection scheme is proposed.Compared with the traditional method,this method does not need the assumption that the degradation model type and fault type are known,which makes the remaining useful life prediction algorithm have better flexibility and adaptability.Finally,the feasibility and effectiveness of the method are verified by the experimental results.