Fault Detection Of Systems With Markov Jump Factors

Fault detection(FD)is the essential and fundamental problem for practical systems.FD based on model has been investigated deeply in the past years.Networked control systems(NCSs)are widely used in wireless communication and internet,thus time delays,data dropout and quantization,which are frequently encountered in NCSs,result in new problems and challenges in FD.Meanwhile,a series of phenomena including switching among subsystems and network packet loss of NCSs can be described by Markov random process models.Therefore,the issue of FD for systems with Markov jump is both theoretical and applied important.Based on the above motivations,this thesis is concerned with the FD of systems with Markov jump properties.By Luenberger filter and full-order filter,H_?performance analysis and residual signal of the fault information are considered,then the FD filters are designed via convex optimization method.This thesis handles with FD filter design for uncertain Markov jump systems with time delays.Based on adaptive filter,dynamic residual systems are established by considering the relationship between systems residuals and disturbances.Then H_?performance analysis and filter design method are proposed in terms of linear matrix inequalities(LMIs),such that the faults can be detected efficiently.In the next two parts of this thesis,the problems of FD of NCSs with Markov packet dropouts are concerned.Different from the first part,the systems parameters under consideration here are not Markovian,while the packet dropout properties are governed by Markov jumps.In order to solve the problem of constrained bandwidth,the effects of Markov packets dropouts and constrained bandwidth on filter inputs are considered.Based on Lyapunov method,a sufficient condition for existence of full-order filter is derived to detect the fault.A numerical example is given to verify the algorithm.Furthermore,since there are few results on FD for networked control systems with both Markov packet dropouts and quantization,Markov chain is employed to describe the packet dropouts and a logarithmic quantization is used.The H_?filter can be obtained by solving LMIs to detect the fault signal.A numerical example is provided to show the effectiveness of the method.