Research On Fault Diagnosis For A Class Of Singular Systems Based On Filtering Methods

Researching on singular systems is theoretical and realistic significant as they not only have more complicated structures but also have better representation characteristics than the state-space ones.Existing results are more focusing on the controlling problems of the singular systems,and the fault diagnosis problem of the singular systems needs to improve.As these reasons,filtering-based methods are implemented which focuses on the fault diagnosis problem for a class of singular systems.The main contributions of this thesis are as the following aspects.Firstly,the problem of the fault diagnosis based on robust observer for a class of singular systems is studied.For a class of linear singular systems,a novel observer with non-singular structure is proposed and is easy to implement.The existence conditions of the observer parameters are given,the solving matrices of the parameters are given in the forms of Linear Matrix Inequality(LMI).The designed observer not only guarantees the augmented error system exponential convergence,but also the robustness of the fault against the augmented disturbance.Further,this method is extended to a class of Lipschitz non-linear singular systems for solving the fault diagnosis problem.Secondly,the problem of the robust fault diagnosis filter design for a class of singular systems with constant delay is studied.For a class of linear singular systems with constant delay,the observer-based robust H_? fault diagnosis filter is designed.This filter has a non-singular structure and is easy to implement.The existence condition of the filter parameters is given,the solving matrices of the parameters are given in the form of LMI.The residual evaluation function and the threshold are introduced in for judging if the fault occurs.This method is extended to a class of Lipschitz non-linear singular systems with constant delay for solving the fault diagnosis problem.Further,for a class of linear singular systems with constant delay which has a tiny fault occurring in a short time,the observer-based finite-time robust H_? fault diagnosis filter is designed.This filter not only guarantees the dynamic model of the augmented error system is Finite-Time Boundness(FTB),but also makes sure the augmented residual meet the robust H_? performance in finite-time.This method is extended to a class of Lipschitz non-linear singular systems with constant-delay for solving the fault diagnosis problem in finite-time.Thirdly,the problem of the robust fault diagnosis filter design for a class of singular systems with time-vary delay is studied.For a class of linear singular systems with time-vary delay,the observer-based robust H_? fault diagnosis filter is designed.This filter has a non-singular structure and is easy to implement.The existence condition ofthe filter parameters is given,and the Cone Complementarity Linearization(CCL)iteration algorithm is introduced in for solving the non-convex problem.The residual evaluation function and the threshold are introduced in for judging if the fault occurs.This method is extended to a class of Lipschitz non-linear singular systems with time-vary delay for solving the fault diagnosis problem.Further,for a class of linear singular systems with time-vary delay which has a tiny fault occurring in a short time,the observer-based finite-time robust H_? fault diagnosis filter is designed.This filter not only guarantees the dynamic model of the augmented error system is FTB,but also makes sure the augmented residual meet the robust H_? performance in finite-time.This method is extended to a class of Lipschitz non-linear singular systems with time-vary delay for solving the fault diagnosis problem in finite-time.Then,the fault diagnosis problem based on the robust Kalman filtering for a class of non-linear singular systems is studied.For solving the problem of the error by the linearization of the non-linear models,a Robust Extended Kalman Filter(REKF)algorithm based on lower bound is proposed.The robust performance of REKF is analyzed,and an accurate fault estimation is implemented.For a class of non-linear singular systems with unknown statistics of noise,an Adaptive Robust Extended Kalman Filter(AREKF)algorithm is proposed,and the fault estimation is implemented.Further,the robust fault diagnosis Kalman filtering problem for a class of non-linear singular systems with measurement time-delay is studied.The process noise of the transforming augmented system is seen as the unknown noise,and by using the AREKF algorithm,the fault diagnosis problem is solved.Finally,fault diagnosis problem based on modified Unscented Kalman Filter(UKF)algorithm for a class of non-linear singular system is studied.Multiple-model fault diagnosis method based on UKF algorithm for a class of non-linear singular systems is proposed,and the fault estimation is implemented for a class of non-linear singular systems.For a class of non-linear singular systems with state uncertainties,a multiple-model fault diagnosis method based on strong tracking unscented Kalman filtering(STUKF)algorithm is proposed,the influence of the uncertainties on the filtering accuracy is decreased by adjusting the predicted convariance matrix,and an accurate fault estimation is implemented.For a class of non-linear singular systems with unknown statistics of time-vary noise,a multiple-model fault diagnosis method based on Adaptive Unscented Unscented Kalman Filter(AUKF)algorithm is proposed,and the the fault estimation is implemented.Further,the fault isolation problem of a class of non-linear singular system is studied based on the AUKF algorithm.Combining the AUKF with the Multiple-Model Adaptive Estimation(MMAE)algorithm,the specific location of the fault is identified,and a fault isolation for a class of non-linear singular system is implemented.