Fault Detection Based on Multi-objective Observer and Interval Hull Computation

Abstract

The design of a novel sensor fault detection scheme for discrete-time linear time-invariant systems with unknown but bounded uncertainties is proposed in the paper. To make the residual sensitive to fault and robust against the unknown disturbance and measurement noise, we propose a multi-objective fault detection observer method based on the criteria of H-index and P-radius and further present an iterative LMI method to solve the observer design problem. In the design of residual evaluation, we propose a threshold computation method via the interval hull approximation of the residual reachable set. The merit of the proposed method is that its threshold computation exhibits less conservatism and higher computational efficiency than the zonotope-based method. Simulations are conducted to demonstrate the superiority of the proposed method.