Robust Detection and Isolation of Sensor Faults for Disturbed Nonlinear Systems

In this paper, a sensor fault detection and isolation scheme based on discrete unknown input multiobservers for a disturbed nonlinear systems is proposed. The proposed multi-observers are based on an uncoupled state multimodel. Sensor faults can be isolated using the generated structured residuals. Nevertheless, the presence of disturbances make the isolation task difficult. Then, a robust sensor fault detection is investigated to make it easier. A bank of unknown input multiobservers is designed to generate structured residuals to accomplish the isolation task. The multiobserver gains are provided in terms of Linear Matrix Inequalities (LMI)s ensuring the convergence of the estimation error. A numerical example is presented to prove the effectiveness of the proposed approach.