Equipment Health Monitoring for Industrial Robotic Arms

Abstract

The topic of equipment health monitoring (EHM) for robotics, including condition monitoring (CM), process monitoring (PM) and predictive maintenance (PdM) is of great interest within the literature, however, there is a significant lack of historical datasets for techniques to be tested upon. Commercial offerings in this area are often manufacturer specific, meaning that fleets that include robots from multiple suppliers cannot easily have performance/condition compared across the fleet. To address this, the work presented within this paper includes an accelerated wear test (AWT) conducted on an industrial robotic arm whilst being monitored using a suite of retrofitted sensors. The resulting data from the AWT is then analysed through a variety of techniques, including regression models, classification models, and a long short-term memory (LSTM) autoencoder, to demonstrate the potential for such methods to be utilised for robot EHM.Additionally, the associated dataset captured during the AWT is to be made openly available through the University of Sheffield’s online research data repository, ORDA, to allow further research to be conducted.