Robust fault detection in hybrid systems using set-membership parameter estimation

Abstract

Hybrid systems exhibit continuous and discrete dynamics and are encountered in many complex and safety-critical systems. Due to their complex nature, the fault diagnosis task becomes very challenging. In this paper, we present a method to perform fault detection with a nonlinear hybrid system using set-membership parameter estimation in a bounded-error framework. Our method relies on a consistency test between the feasible parameter set as computed in normal fault-free operation or given as nominal, and the feasible parameter set as estimated during on-line operation over a given time horizon. Hence, a fault is detected if the feasible set for the parameter vector estimated online is inconsistent with nominal values. An illustrative example is presented.