H-Infinity Filtering Based Fault Detection, Estimation and Fault Tolerant Control

Abstract

This thesis investigates the integrated fault detection, estimation and fault tolerant control problem for linear systems and Lipschitz nonlinear systems. Faults and disturbances are taken into consideration in a unified formulation. A H-infinity observer-based fault detection filter (FDF) is applied to generate residual signals. The FDF is designed to minimize the influence of disturbances and maximize the sensitivity of faults at the same time. Then a new online fault estimation scheme is designed by applying H-infinity filtering to residual signals instead of system outputs. Compared with existing literature, in which system outputs are commonly adopted for fault estimation, the proposed fault estimation scheme based on residual signals can achieve more accurate fault estimation results due to the fact that the influence of disturbances is minimized in the residual signal produced by the H-infinity FDF. Finally a fault tolerant controller is designed to retain the system stability and performance by compensating for the faults. The integrated scheme consists of three essential steps that are centered around the residual signal, namely, generating the residual signal for fault detection, and filtering the residual signal for fault estimation and fault compensation. In this framework, fault diagnosis and fault compensation are designed simultaneously in a closed-loop system. This integrated scheme is considered to be one of the most important contributions of this thesis. To demonstrate the effectiveness of the proposed method, two examples are given and simulation results are presented.