Fault Reconstruction Based On Sliding Mode Observer

Due to the increasing demands for higher performance, higher safety and reliabilitystandards, the approaches to fault detection and isolation(FDI) have received more andmore attention . The fundamental purpose of an FDI scheme is to generate an alarm whena fault occurs and also to identify the nature and location of the fault. The most commonlyused FDI methods are observer-based where the measured plant output is compared tothe output of an observer designed from the model of the system, and the discrepancy isused to form a residual. Using this residual signal, a decision is made as to whether afault is present and also an attempt is made to determine its location. In recent years, anuseful alternative to residual generation is fault reconstruction,which not only detects andisolates the fault, but also provides an estimation of the fault. This is the reason why faultreconstruction method has attracted more attention and has been developed sufficiently.In this dissertation, fault reconstruction method based on sliding mode ob-server(SMO) is discussed. The main content of the dissertation is outlined as follows.The fault reconstruction method for linear systems is reviewed and applied to theactuator and sensor fault reconstruction for electronic throttle control system. Then themethod is extended to sensor fault reconstruction based on sliding mode observer forT-S fuzzy system. The design algorithms of sliding mode observer based on iterativelinear matrix inequality(ILMI) is given. The method is applied to the fault reconstructionscheme of yaw rate and lateral acceleration sensor for vehicle electronic stability controlsystem.For a class of descriptor systems with actuator faults, fault reconstruction based onsliding mode observer is considered. The design problem is transformed into a slidingmode observer design problem for linear systems. Sufficient conditions for the existenceof the observer are given. The method is extended to the case of the general descriptorsystems with actuator faults and the existence conditions of the observer are also given.A new canonical form for descriptor systems is given and a design method based oncanonical form is presented. The design algorithms can exploit all the possible degreesof freedom of the parameterand be resolved by linear matrix inequality(LMI). By theintroduction of suitable filters, the method is extend to the descriptor systems with sensor faults and sufficient conditions for the existence of the observer are given.Fault reconstruction methods of actuator faults based on sliding mode observer foruncertain descriptor linear systems and Lipschitz nonlinear descriptor systems are consid-ered and the design algorithms are given. The method uses H∞control theory to designthe observer gain such that L2 gain from the uncertainty to reconstruction error of faultis minimized. The design is formulated as a convex optimization problem subject to LMIconstraints.For a class of descriptor linear systems with actuator faults, fault reconstructionbased on cascaded sliding mode observer(CSMO) is considered. By importing a newcanonical form, the design method is given and the existence conditions of the observerare obtained. The method is extended to the general descriptor linear systems with actua-tor faults and the existence conditions of the observer are given.