The prediction research of tool VB value based on Principal Component Analysis and SVR

According to the amount of tool wear prediction problems, online prediction of tool wear model is established based on the theory of support vector regression (SVR) regression. The acoustic emission signals and current signals are, respectively EEMD decomposed and wavelet packet decomposed to get the energy values, which are combined with the spindle speed, feeding, and back engagement to form the original feature vectors. By principal component analysis for data processing, the principal elements as the Support Vector Regression (SVR) optimized by genetic algorithms are inputted. The results show that this model has high precision, fast operation. In the tool processing, the acoustic signals and current signals gained by sensor contain abundant processing information, which can reflect the variations of tool wear. As a very promising prediction technology, support vector regression is a model identification method based on statistical learning theory, which shows many advantages in solving the following problems, like small samples, non-linear and high dimensional recognition. The paper adopted support vector machine regression algorithm optimized by genetic algorithm to form a model and predict the VB values of tools ([1-2]).