Milling Chatter Detection System Based on Multi-Sensor Signal Fusion

Abstract

The occurrence of chatter in the machining process accelerates tool wear and breakage, reduces the service life of tools and machine tools and the surface accuracy of workpieces. Therefore, how to quickly and accurately detect chatter and adjust process parameters to prevent chatter has become a research hotspot today. This article proposes a new method to detect the machining state in the milling process by using the sensor signal sensitive feature data set and the Mahalanobis-Taguchi System. First, a method for extracting the comprehensive characteristics of nonlinear time-domain signals in the cutting process is proposed. This method can realize the dimensionality reduction of the time-domain data features. Then, time-frequency domain methods such as variable mode decomposition energy entropy, multi-scale power spectrum entropy and multi-scale displacement entropy are used to process the processed signals collected by multiple sensors to obtain the characteristic data set related to flutter. Take the feature data set as the measurement scale to participate in the construction of the Mahalanobis-Taguchi System classifier. Using the experimental data of milling titanium alloy thin-walled parts with end mills, the Mahalanobis space of cutting chatter was constructed, and the chatter detection system was verified. It is found that this method can accurately detect the cutting processing state. The research results can lay the foundation for advancing the industrialization of intelligent cutting processing.