Tool Wear State Diagnosis Based on Wavelet Analysis-BP Neural Network

Abstract

Cutting force collected by experiment is transformed by continue wavelet in order to overcome the disadvantage that signal processing analyzes single variable. The eigenvector which can reflect tool wear state is extracted from scale-energy matrix based on analysis, and BP neural network is established to predict tool wear. Trained network is used for prediction by unknown sample. Results show that this method can identify and diagnose accurately tool wear state.

Online tool wear prediction in wet machining using modified back propagation neural network

“…However, the radial basis function, though accurate in the prediction for a given range of data, has poor generalization abilities [1921] and hence is not preferable. Fan et al [22] transformed the cutting force data by a continuous wavelet in order to overcome the disadvantage that signal processing analyses a single variable. The eigenvector that can reflect the tool wear state is extracted from a scale-energy matrix based on analysis, and a BP neural network is used to predict tool wear.…”