Fault Detection and Isolation for Uncertain Systems Based on Unknown Input Set-Membership Observer

Abstract

This paper proposes an actuator fault diagnosis method for a class of discrete-time linear systems subject to parametric uncertainties, unknown but bounded disturbance and noise. The actuator faults of the systems are taken as unknown inputs, and the proposed unknown input observer has a new structure that provides more robust and more design degrees of freedom. A zonotope-based method is used to estimate the threshold of residual generated by the unknown input observer. Based on the proposed method, a fault detection and isolation strategy using a set of unknown input set-membership observers is presented. The effectiveness of the proposed method is verified by numerical simulation of a flight control system.