Intelligent feedrate optimization using a physics-based and data-driven digital twin

Intelligent manufacturing machines envisioned for the future must be able to autonomously select process parameters that maximize their speed while adhering to quality specifications. Accordingly, this paper pro-poses a framework and methodology for using a physics-based and data-driven digital twin of a feed drive to maximize feedrate while respecting kinematic and contour error limits. To correct for inaccuracies intro-duced by unmodeled dynamics and disturbances, the data-driven model is updated on-the -fly using sensor feedback. Experiments on a 3-axis CNC machine tool prototype are used to demonstrate up to 35% cycle time reduction without violating error tolerances compared to the status quo.& COPY; 2023 CIRP. Published by Elsevier Ltd. All rights reserved.