Milling cutter wear feature extraction based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and Lempel-Ziv complexity

Abstract To accurately and quantitatively extract the wear characteristics of milling cutters during the milling process, the wear signal of the milling cutter is collected using a laser vibrometer. The characteristics of the tool wear signal during the milling process are analyzed in the time, frequency, and time-frequency domains. The Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) method is employed to extract the characteristic quantities from different wear signals of the milling cutter. Subsequently, the Lempel-Ziv complexity algorithm is applied to analyze the wear characteristics under various wear states, allowing for the quantitative evaluation of these states. Finally, the statistical principle is applied to establish the wear discrimination intervals for the milling cutter under different wear states: the normal working complexity interval is [0.139, 0.174], and the wear failure complexity interval is [0.202, 0.228]. The proposed CEEMDAN + Lempel-Ziv method achieves an accuracy of 93.33% in accurately identifying the normal operation of the milling cutter and an accuracy of 95% in identifying wear failure of the milling cutter.