Pattern recognition of tool wear in high-speed milling based upon nonlinear analysis

Tool wear may result in negative consequences on the quality of machined surface, machine life and manufacturing productivity. To address this problem, an online tool condition monitoring (TCM) system plays a significant role in metal cutting field. To reveal the relationship between the mill signal features and the wear degradation behavior of milling tool and improve the accuracy of tool wear recognition, this paper presents a methodology to identify tool wear states online based on AR Bispectrum, Chaotic Characteristics and Hidden Markov Model (HMM), using cast iron material and steel J45 material high-speed milling as a case study. The acoustic emission and vibration acceleration signals were collected continuously during the tool's whole life-cycle time. The extracted features associate with AR bi-spectrum coefficients, fractal box dimensions (FBD) and Largest Lyapunov exponent (LLE) are given as an input to the HMM classification model to identify the wear state of milling tool. Experimental results are shown that the nonlinear features are superior to conventional statistical features in term of recognition accuracy.