Fault detection for nonlinear systems in presence of input unmodeled dynamics

This paper presents the problem of fault detection in the presence of input unmodeled dynamics. There are mainly two sources of uncertainties for a system in addition to external disturbances: parametric uncertainties and unmodeled dynamics. For nonlinear systems, the problem of fault detection and isolation has been studied under the assumption that modeling uncertainty can be bounded by an a priori known function. Many robust algorithms have been designed based on this assumption. But, the problem of fault detection in the presence of dynamic uncertainties for nonlinear systems has not been addressed by many researchers. The goal of this work is to focus on the problem of actuator fault detection in the presence of input unmodeled dynamics. We use state-estimation errors as residuals for fault detection and monitoring the system for any off-nominal behavior. A rigorous analysis is done to understand the effects of unmodeled dynamics on the process of fault detection. Based on the analysis, we derive a threshold function for the purpose of fault detection, which can decouple the effects of fault and unmodeled dynamics. Finally, simulations are performed to show the effectiveness of the proposed method.