Reduction of experimental efforts for predicting milling stability affected by concept drift using transfer learning on multiple machine tools

Abstract

Due to complex interrelations between the characteristics of the machine tool, spindle, tool wear and the stability of milling processes, the design of stable machining operations is challenging. Concept drift resulting from, e.g., tool wear and different dynamic behaviours often require fundamental experimental investigations on each machining centre. This paper presents a methodology for modelling process characteristics with respect to resource constraints by transferring insights from extensive experiments conducted on a reference machine to other machine tools in a process-informed manner. This methodology was exemplarily applied to predict wear-dependent process stabilities with a significantly reduced number of required cutting tests.