Prediction of robot milling deformation error based on process feature dimensionality reduction and RVM

Abstract

Industrial robots are increasingly serving as a key manufacturing paradigm alongside traditional machine tools, largely due to their expansive workspace and operational versatility. However, their inherent lack of high structural stiffness often leads to system deformation when subjected to milling forces, thereby severely limiting overall machining precision. Consequently, understanding the complex relationships between the deformation errors at the robot’s end-effector and various operational parameters, such as spindle speed, tool feed rate, joint configurations, and cutting forces, is critical. In this study, we introduce a novel approach for predicting robotic milling deformation errors. Our technique leverages process feature dimensionality reduction combined with a Relevance Vector Machine (RVM), utilizing empirical position and force data. We validated this approach through comparative trials involving diverse machining settings within various robot workspaces, demonstrating its enhanced performance compared to conventional theoretical models. Findings indicate that more than 88.8% of the test instances yielded prediction errors of less than 0.005 mm. Ultimately, this methodology establishes a robust modeling framework for future error compensation strategies in robotic machining operations.