In-process Detection of Low and High Frequency Chatter in Robot Machining

Abstract

Chatter is one of the most complex phenomena that are exhibited during a milling process, while having the most detrimental effects on surface quality, part accuracy, tool wear and machine tool health. Thus, in-process detection of chatter is of utmost importance to enable real-time suppression algorithms. When robot milling is considered, the effects of chatter are even more prominent than CNC milling, since robots inherently have significantly lower stiffness than machine tools, due to their design as open kinematic chains. Additionally, two types of regenerative chatter, namely Low Frequency Chatter and High Frequency Chatter can be observed in robot milling. The former is a product of the low eigenfrequencies of the robot and is solely dependent on robot dynamics, while the latter is a product of the dynamics of the cutting tool and spindle assembly. This work presents an algorithm capable of fast, in-process prediction of both chatter types. The energy shift between the harmonic signal towards chatter-related frequencies, as well as the cyclostationarity of the process are explored and used as features to feed a Support Vector Machine algorithm for chatter detection, enabling a very high detection performance.