An intelligent chatter detection method for high-speed milling under variable tool-workpiece systems and cutting parameters

Chatter is prone to occur in high-speed milling, which can reduce machining quality, shorten tool life, and generate excessive noise to pollute the environment. Therefore, timely detection of chatter is crucial for sustainable production. In this paper, an intelligent chatter detection method is proposed under variable tool-workpiece systems and cutting parameters. To eliminate the interference of cutting parameters, the chatter signals are first automatically extracted based on the successive variational mode decomposition (SVMD) method and two metrics with physical meaning. Then, dimensionality reduction processing is performed according to the energy transfer characteristic, which can simultaneously eliminate the effect of tool-workpiece system variations. Finally, a novel deep attention information fusion network (DAIFN), consisting of SE module, MAConv module, CapsNet module, and classification module, is constructed to extract chatter features and achieve chatter detection from multi-channel time series signal data. Experimental results show that the average accuracy of milling chatter detection can reach 99.16%. The accuracy and generalization of the proposed method are verified in different scenarios by comparison with other methods.