Tool Wear Detection Based on Wavelet Packet and BP Neural Network

Abstract

Based on wavelet packet decomposition and the BP neural network of pattern recognition theory, this article puts forward the theory that can identify the different tool wear conditions during the cutting process, and thus we can use this theory to forecast the tool breakage accurately. The main thinking of this article is that decomposing tool acoustic emission signal by using wavelet packet to get spectrum coefficient as eigenvector, and then putting it into the BP neural network to be trained in order to accomplish the final pattern recognition of tool wear conditions by making use of BP algorithm. By testing the samples of well-trained network, it is proved that the BP neural network constructed has good generalization ability which can identify tool conditions accurately.