Degradation curves integration in physics-based models: Towards the predictive maintenance of industrial robots

Predictive maintenance has been proposed to maximize the overall plant availability of modern manufacturing
systems. To this end, research has been conducted mainly on data-driven prognostic techniques for machinery
equipment individual components. However, the lack of historical data together with the intricate design of
industrial machines, e.g. robots, stimulate the use of advanced methods exploiting simulation capabilities. This
paper aims to address this challenge by introducing a generic framework for the enhancement of advanced
physics-based models with degradation curves. The creation of a robot’s simulation model and its enrichment
with data from the degradation curves of the robot’s components is presented. Following, the extraction of in-
formation from degradation curves during the simulation of the robot’s dynamic behaviour is addressed. The
Digital Twin concept is employed to monitor the health status of the robot and ensure the convergence of the
simulated to the actual robot behaviour. The output of the simulation can enable to estimate the future behaviour
of the robot and make predictions for the quality of the products to be produced, as well as to estimate the robot’s
Remaining Useful Life. The proposed approach is applied in a case study coming from the white goods industry,
where it is investigated whether the robot will experience some failure within the next 18 months.