Comparison of Classifiers for Chatter Detection

Abstract

Unstable machine cutting causes chatter and reduces quality of the production. Therefore it must be detected. Several techniques have been presented for this reason. The aim of this study is to determine the data, features and classifiers which fit on chatter detection. In order to detect chatter; acoustic emission and vibration data are collected, several features are generated which belong to time and frequency domains. Then the best features are chosen via k- means clustering, support vector machines, feed forward back propagation neural networks and perceptron classifiers. The performance of the system is analyzed. As results of the study, the best data, features and classifiers are chosen for the chatter detection.