Experiments of observer-based fault detection for an industrial robot

Different techniques to fault detection and isolation (FDI) for robot manipulators have been proposed in recent years, but the problem of applications of such techniques on industrial robot has not been extensively investigated. In this paper an observer-based approach to FDI is applied to a conventional industrial manipulator with open control architecture. A nonlinear discrete-time observer is designed to generate the residuals sensitive to both sensor and actuator failures. Then, the effectiveness of the resulting FDI scheme is tested in a number of experimental case studies. The results show that clear and distinct fault signatures can be obtained in the presence of modeling uncertainties and external disturbances. Also, the potential of nonlinear FDI techniques applied to conventional industrial machines is demonstrated.