Tool-wear monitoring during micro-end milling using wavelet packet transform and Fisher’s linear discriminant

Tool wear is one of the most important parameters in micro-end milling, and can be used to monitor the condition of the machine and the tool. A micro-end mill has different characteristics from a macro-scale end mill; in particular, shank run-out (which is negligible in the macro-scale tool due to the low aspect ratio) is significant in micro-end milling, inducing excessive tool wear and reduced tool life and leading to sudden, premature failure. In this paper, a novel tool-wear monitoring method is described for determining the state of a micro-end mill using wavelet packet transforms and Fisher’s linear discriminant. Force and torque signals were measured using a dynamometer and were used to reflect geometric changes in the micro-end mill due to wear. Because of the small signal-to-noise ratio, sensor signals measured during the milling process were periodically averaged, and the resulting single-period signals provided improved efficiency of feature extraction using wavelet packet transforms. The extracted features were classified in the wavelet domain and used to determine the tool state employing a hidden Markov model. The recognition results were compared with those of an energy-based monitoring technique, and we found that our method could determine the tool state more accurately for both normal wear and premature failure of micro-end mills.