Joint Stress Estimation and Remaining Useful Life Prediction for Collaborative Robots to Support Predictive Maintenance

Abstract

Anticipating the maintenance needs of lightweight robotic manipulators at precise future instances represents a significant challenge within the automation domain. This letter introduces an innovative and comprehensive method to estimate the severity of stress imposed on a robot joint at any given time. Additionally, we present a knowledge-based predictive model aimed at approximating the End of Life (EoL) for a robotic joint, enabling the prediction of its Remaining Useful Life (RUL) with respect to the designated load case. This predictive model is rooted in a baseline derived from empirical data covering the entire Universal Robots (UR) e-series and is trained using synthetic data. Subsequently, it undergoes evaluation with a real-world dataset and is further validated in a case study. The model demonstrates a high level of accuracy, with worst-case performance reaching <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\text{90.3}{\%}$</tex-math></inline-formula> as the lower limit.