Use of wavelet packet and principal composition analysis method for feature extraction of chatter in milling

Based on wavelet packets transform, a new method was proposed for feature extraction of chatter in milling process by using vibration signal. These packets contain major chatter information of the original signal. Using the principal composition analysis method, a feature was reconstructed from these packets, and a principal composition score was proposed to assess the reconstructed feature. The realization of the procedure of automatic feature selection for a given process was studied. A uniform formula of the feature collectivity was given. A diagnosis model of chatter based on Euclidean distance is established. A practical milling process was employed to show that the proposed method is very effective.