Health Monitoring Of Hybrid Mechatronic System Based On Ckf And Wiener Degradation Model

With the rapid development of industrial technology,the level of intelligence,integration and automation of industrial production is constantly increasing.Hybrid mechatronic systems are widely used in industrial production and play an important role in the process of industrial modernization.In recent years,numerous serious accidents have occurred due to the fault of mechatronic systems which not only caused huge economic losses,but caused casualties.Therefore,how to ensure the reliability and safety of hybrid mechatronic systems under various conditions is an important issue.In this thesis,researches on hybrid bond graph modeling,fault detection and isolation,fault identification,and remaining useful life prediction(RUL)are carried out for health monitoring of hybrid mechatronic system.First,hybrid bond graph is used to model hybrid mechatronic system and parameter identification method is applied to determine model parameter for describing dynamic behavior of actual system.Besides,global analytical redundancy relation(GARR)and fault signature matrix(FSM)are constructed from hybrid bond graph model,which are respectively used for fault detection and fault isolation.In addition,the adaptive threshold based on linear fractional transformation theory is proposed to improve sensitivity of fault detection.Next,in order to determine the true faults from the suspected fault candidates(SFC),a fault identification method based on cubature Kalman filter(CKF)is applied,which solves the problem of multi-fault isolation when faults with same characteristics occur simultaneously.Meanwhile,results of fault identification can reveal the information about severity of the fault.Finally,the Wiener process model is applied to establish degradation model of incipient fault,which can not only describe the evolution trend of fault parameter,but also reflect various uncertainties in degradation process.Least square method and improved krill herd algorithm are used to estimate the unknown parameters in degradation model.After that,the probability density function of RUL is derived based on identified degradation model,which realizes the RUL prediction of incipient fault.