Fault Diagnosis of Joint Backlash

Abstract

The aim of this paper is to develop appropriate techniques to detect and classify the joint backlash of a robot by monitoring its vibration response during normal operating conditions. In this investigation, Wigner-Ville distributions combined with two-dimensional correlation techniques have been employed to diagnose the joint faults of multi-link robots. In the study reported here, signal detection based on the Wigner-Ville distribution is proposed as a tool for pattern differentiation. To evaluate the performance of different detection procedures, the detection of a simulated impact transient embedded in three simulated observed signals is presented. To assess the validity of the proposed approaches, they have been successfully employed in the fault diagnosis of link-joints on both a two-link mechanism and an industrial robot.