Investigation on Dynamical Prediction of Tool Wear Based on Machine Vision and Support Vector Machine

In this chapter, a new dynamical prediction method of tool wear is proposed. The method employs machine vision and support vector machine for assessing the degree of wear. First, machine vision is used for online measuring the tool wear. The tool wear models are established with considering the joint effect of machining conditions, namely model by using least square support vector machines (LS-SVM). Then, in order to predict tool wear more accurately, a dynamical prediction method is designed to adjust the established tool wear model by using the online measured tool wear. Experimental works with the purposes to train LS-SVM-based tool wear model and to verify dynamical prediction results are performed on MIKRON UCP710 five-axis milling centre with using stainless steel 0Cr17Ni12Mo2 as a test material. The tool wears measured in experiments are used to train the established LS-SVM-based tool wear model and the inter-connection relationship between input and output parameters is determined after training. The on-line tool wears obtained through machine vision are used to adjust the LS-SVM-based tool wear model on real time basis. The experimental comparisons show that the proposed on-line dynamical prediction method of tool wear is satisfied and provides a basis for industrial application of on-line tool wear monitoring.