Research On Observer-based Robust H_∞ Fault Detection

Model uncertainty and time delays are usually unavoidable in most industrial processes and computer control systems,such as chemical processes, paper industry,electric power transmission,and networked control systems.In order to ensure the system safety and reliability,researches on robust fault detection for systems with model uncertainties and systems with time delays are very important.The main purpose of this dissertation is to investigate problems related to observer-based robust H_∞fault detection for linear systems with Polytopic model uncertainties,linear systems with multiplicative stochastic noise,linear systems with parametric faults,and linear systems with constant time-delays,respectively.In Chapter 2,two approaches to observer-based H_∞fault detection are briefly introduced.One is the H_∞optimization approach.Using the H_∞norm or nonzero minimal singular value of transfer function from fault to residual as a sensitivity performance index and the H_∞norm of transfer function from unknown input to residual as a robust performance index,the design of fault detection filter can be formulated as a minimization problem and a unified solution has been obtained by solving a matrix Riccati equation.The other is the H_∞filtering formulation of fault detection,which can be solved by applying the technique of linear matrix inequality.In Chapter 3,the problem of robust fault detection for linear discrete-time systems with both unknown inputs and Polytopic uncertainties is studied.The main contributions include the H_∞filtering formulation of robust fault detection,the extension of an H_∞filtering approach to linear system with Polytopic uncertainties,the derivation of sufficient conditions in terms of linear matrix inequalities,and the parameterization of parameter-independent robust fault detection filter.Chapter 4 deals with the problem of fault detection for a class of linear systems with multiplicative stochastic noise.Under the circumstances that some prior knowledge about the possible fault can be obtained,the dynamic behavior of the fault is proposed and used to solve fault estimation problem.While,if there is no extra knowledge available about the possible fault other than the condition of l₂ bounded,the residual generation is formulated into a problem of observer-based H_∞filtering problem in the sense of stochastic,and the linear matrix inequality approach to solve this problem is proposed.The problem of residual evaluation is also considered.With given residual evaluation function and threshold,the occurrence of fault can be detected and the false alarm rate can be estimated;with given residual evaluation function,an available false alarm rate can also be guaranteed by suitably choosing threshold.In Chapter 5,the problem of observer-based parametric fault detection system is studied.The core of this chapter is to first describe the faults occurred in system actuators,sensors and components in a form of additive parameter deviations,then to transform the parametric fault detection problem into a similar additive fault setup,based on which an observer-based H_∞optimization approach to parametric fault detection is proposed.The definition of parametric fault detectability is also introduced.Chapter 6 investigates the design problem of observer-based H_∞fault detection filter for a class of time-delay systems.The key of this chapter is the introduction of a generalized coordinate change such that in the new coordinates all the time-delay terms are injected by the system input and output.Based on the new obtained systems expression,an observer-based fault detection filter is considered as the residual generator,while the dynamics of the residual is changed into a general linear time-invariant form with unknown input and fault time-delay terms.Then the fault detection filter can be designed by applying approaches to linear time invariant systems.As an example,an H_∞optimization approach is introduced.