A Dual-Model Fault Detection Approach with Application to Actuators of Robot Manipulators

Abstract

Multiple-model (MM)-based methods have been successfully applied to many fault detection schemes; however systematic design of the associated model set remains an open question. The difficulty comes from the fact that using a large model set reduces the risk of undetected faults, but also increases the computation load drastically. In this paper we propose a dual-model fault detection (DMFD) algorithm aiming at solving the model set design problem, and apply it to detect actuator faults of robot manipulators. The DMFD algorithm is able to detect various types of unexpected actuator faults, including abrupt faults, incipient faults, and simultaneous faults, in a computationally efficient way. To evaluate the performance of the DMFD algorithm, upper bounds of the false alarm and missed detection probabilities are explicitly presented in terms of the tunable variables. Furthermore, experiments are conducted to demonstrate its ability in immediate detection of faults.