Study on Identification Method of Tool Wear Based on Singular Value Decomposition and Least Squares Support Vector Machine

摘要

In view of the non-stationary characteristics of acoustic emission signal of tool wear, and the slow convergence rate of learning algorithm and easily dropping into the local minimum value for BP neural networks, a novel method of tool wear state identification based on singular value decomposition and least squares support vector machine was proposed. Firstly, the empirical mode decomposition method was used to decompose the collected acoustic emission signals into a number of intrinsic mode function which was used to construct the initial feature vector matrix. Then by applying the singular value decomposition method to the initial feature vector matrix, the singular values were obtained. Finally, the singular spectrum was selected to constitute the feature vector. The feature vector was divided into two groups, one group was used to train the least squares support vector machine and the other was used to identify the tool wear state. The identification result proves that this method is superior to BP neural network, and it has a higher identification rate. It is proved that this method is efficient and feasible.