Enhancing Industrial Collaborative Robots Inverse Dynamic Model for Application in Fault Detection: A Hybrid Approach

An accurate dynamic model of industrial collaborative robots is needed for condition monitoring applications, such as in the detection of early faults or anomalies. Due to the complex and coupled nature of cobots, their modelling is challenging and easily prone to errors. These errors can be linked to simplifying assumptions made during the model formulation, like unchecked kinematic errors, the linearization of the dynamic equations, unaccounted load, incorrect friction model, non-exciting trajectory, etc. This paper presents different techniques of modelling a UR10e cobot offline, ranging from physics-based models to data-driven methods and model enhancement through a hybrid approach. The prediction performance of these techniques was compared, and the hybrid model yielded the best overall accuracy for the two test trajectories examined. Moreover, the hybrid model developed has the potential to be deployed in model-based condition monitoring for the early detection of cobot faults.