Bayesian identification of energy models for industrial machinery controlled rotary axes

Abstract

A fair prediction of energy consumption in machine tools is an essential requirement to optimize work cycles, processes and equipments also in the light of energy efficiency, as promoted by Industry 4.0. In this work, we focus on the energy behavior of machinery controlled rotary axes commanded by brushless electric motors, and propose a framework to predict their energy consumption. It consists of a model-based energy representation of these axes, combined with experimental procedures and statistical techniques to characterize the model. In particular, we select a proper model from a number of candidates, and estimate its coefficients including their variability based on specific movement conditions. Model fitting is performed through a Bayesian approach, while the dependency of the coefficients on movement conditions is characterized with a K-Nearest Neighbors technique. The overall framework has been validated on the six axes of the pick and place anthropomorphic Comau NS16 robot. The results of this application confirm the effectiveness of the approach, which can therefore be considered as a valid tool for assessing the energy consumption of a machinery axis during the execution of a specific work cycle.