Feedrate optimization based on part-to-part learning in repeated machining

Abstract

Cutting operations often involve machining parts of similar geometry repeatedly, offering opportunities for learning-based improvements. While past studies have focused on enhancing machining accuracy through part-to-part learning, this work shifts the focus to optimizing feedrate under servo error constraints. A data-driven model, trained online on prior machining data, predicts future servo errors and enables iterative feedrate optimization. Confidence in the model improves as more parts are machined, permitting progressively higher feedrates. Experimental results demonstrate significant speed gains within a few iterations, showcasing the potential of part-to-part learning for autonomously achieving faster machining without violating servo error constraints.