Mirror milling chatter identification using Q-factor and SVM

Abstract

Mirror milling is an effective approach to improve large-scale monolithic thin-walled parts machining quality through ensuring the mirror relations of cutter and supporting head. However, the introduction of supporting head influences the dynamic characteristics of the tool-workpiece system. Essentially, the measured raw signal contains more coupled components and shows more oscillatory and aperiodic behaviors. Therefore, it is difficult to identify the mirror milling chatter using the monitoring signals. Comparing with traditional indicators, Q-factor can be used to describe the machining state in the view of signal oscillatory behavior, which is suitable for chatter-related component extraction in thin-walled part machining. In this paper, chatter-related signal component identification and diagnosis of thin-walled parts based on signal Q-factor and support vector machine (SVM) is proposed. The frequency band with maximal variation of Q-factors is taken as the chatter-related signal component. Using the feature vector constructed by Q-factors and power spectrum values of the determined frequency band, the SVM is used for milling state diagnosis. The prediction accuracy is much higher than the other frequency band and traditional indicators. It indicates the effectiveness of the proposed mirror machining chatter identification method.