Sequential spindle current-based tool condition monitoring with support vector classifier for milling process

Abstract

Inexpensive tool condition (TC) monitoring plays a significant role for less human duty machining process in large-scale manufactory. A time-sequential spindle current-based tool breakage diagnosis technique with least squares support vector machine (LS-SVM) classifier is investigated to provide an inexpensive on-line TC monitoring system for milling process. The recognition technique consists of a spindle motor current feedback sensor, a signal processor, and an intelligent classifier. The processor generates machining condition features with Sym6 wavelet transformation to decompose the feedback signals in time domains to generate sequence samples. The features involving both normal and broken tool conditions during machining are fed into the classifier to conduct kernel-based LS-SVM training. With the transformation and training, an object oriented representation function as the LS-SVM classifier is set up and then utilized to diagnose tool fractures in the real time under varying cutting conditions. Experiments were conducted on a milling platform with the built monitoring system consisting of a current acquisition system and its processing software. Experimental results show high accuracy rate and high calculation performance in on-line monitoring of cutting tool conditions for milling process.