Robust Fault Detection Based on Nonlinear Analytic Redundancy Techniques With Applications to Robotics

Abstract

A new approach to sensor and actuator fault detection in the presence of model uncertainty and disturbances, and its application to a wheeled mobile robot (WMR) are presented in this paper. Robust fault detection is important because of the universal existence of model uncertainties and process disturbances in most systems. This paper proposes a new approach, called robust nonlinear analytic redundancy (RNLAR) technique, to sensor and actuator fault detection for input-affine nonlinear multivariable dynamic systems in the presence of model-plant-mismatch and process disturbance. The proposed RNLAR can be used to design primary residual vectors (PRV) for nonlinear systems to detect sensor fault that are completely insensitive to both the model-plant-mismatch and process disturbance. It is shown that the PRV for actuator fault cannot be made completely insensitive to these factors. In order to overcome this problem, a nonlinear PRV design method to detect actuator faults is proposed where the PRVs are highly sensitive to the actuator faults and less sensitive to model-plant-mismatch and process disturbance. The proposed robust fault detection methodology is applied to a WMR and the simulation results are presented to demonstrate the effectiveness of this new approach.