Abstract: Robotic milling exists both mode coupling chatter and regenerative chatter. Effective chatter types detection in robotic milling is the basis for accurate chatter suppression and guarantee of machining quality. Therefore, a new method for chatter types detection based on adaptive variational mode decomposition and difference of power spectral entropy (AVMD-ΔPSE) is proposed. By analyzing the chatter characteristics and dominant modes of robotic milling, the chatter of robotic milling is divided into two types:mode coupling chatter dominated by robot mode and regenerative chatter dominated by tool-spindle mode. In order to extract the sensitive chatter signals, the original signals are decomposed by adaptive variational mode decomposition method. According to the power spectral entropy and frequency elimination algorithm, the chatter types detection indicators about difference of power spectral entropy is designed, and the optimal classification thresholds of the identification indicators are determined by using gaussian mixture model adaptively combined with many experimental data. Chatter detection experiments show that milling machine chatter detection method is applied to robotic milling can only identify chatter cannot distinguish between different types of chatter, and AVMD-ΔPSE method can accurately recognize and distinguish between mode coupling chatter and regenerative chatter, which provides theoretical guidance for the suppression of chatter in robotic milling.

Key words: robotic milling chatter, chatter types detection, AVMD, difference of power spectral entropy