Tool wear monitoring based on kernel principal component analysis and v-support vector regression

Machined surface quality and dimensional accuracy are significantly affected by tool wear in machining process, and severe tool wear may even lead to failing of the workpieces being processed. Tool wear monitoring is highly desirable to realize automated or unmanned machining process, which can get rid of the dependence on skilled workers. This paper mainly studies on the methods and techniques of on-line tool wear monitoring through static and dynamic cutting force signals. Sensitive signals related to tool wear are preliminarily selected by using correlation coefficient method. Kernel principal component analysis (KPCA) technique is adopted to fuse these sensitive features for improving training speed and prediction accuracy. Then, the tool wear predictive model based on v-support vector regression (v-SVR) is constructed through learning correlation between the fused features and actual tool wear. The obtained result shows that the prediction accuracy of the proposed tool wear model is proved effective beyond expectation. Besides, the proposed model still has better generalization ability even in small sample size.