Research On Fault Diagnosis And Prognosis Method For Nonlinear Mechatronic System Based On Adaptive Evolutionary Filter

With the rise of high-tech industries in last few decades,electrical machines become more and more critical in industrial application.Production processes have urgent demands on safety and reliability of electrical machines,hence,timely and effective fault diagnosis and prognosis techniques for mechatronic systems are reasonably essential.This dissertation presents a real-time model-based fault diagnosis and prognosis method for a nonlinear mechatronic system.For monitoring of complex systems which includes multiple energy domains like electrical,mechanical and hydraulic,etc,bond graph serves as a powerful modeling method that can describe causalities of the monitored system through intuitive graphics.In the thesis,a bond graph model of mechatronic system considering the influence of Stribeck friction is established.Firstly,fault diagnosis of the system is conducted based on the bond graph model.By analyzing the constitutive relation of all junctions and the causality path connected to the junctions in the model,the analytical redundancy relations(ARR)are generated.A combinative fault signature matrix is constructed by combining dependent ARR with independent ARR to promote the isolation capability of the monitored system under multiple faults condition.After fault isolation,an adaptive evolutionary filter(AEF)technique is developed for fault estimation and remaining useful life prediction.The AEF can adaptively determine the genetic parameters according to the fitness of the particles,by which the particle impoverishment problem in general particle filter can be effectively mitigated and hence the accuracy of fault estimation and remaining useful life prediction can be improved.At last,a test bed of the nonlinear mechatronic system is established to validate the developed bond graph model.The effectiveness of the proposed method is verified by comparison studies through simulations and experiments on the nonlinear mechatronic system.