Chatter Recognition in Band Sawing Based on Feature Extraction and Discriminant Analysis

Abstract

Tool damage due to chatter poses harmful economic impact in modern machining production therefore it is important to avoid or suppress chatter during the production process. In order to establish automated chatter-free cutting conditions, the methods for online recognition of chatter and chatter-free cutting should be developed. It this paper a band sawing cut-off process is considered where a combination of selected workpiece properties and cutting parameters result in chatter. A method for online chatter detection based on processing of acoustic signals is proposed. The method consists of pre-processing sound signals with Short-Time Fourier Transform (STFT), extracting frequency invariant features, and finally applying Quadratic Discriminant Analysis (QDA) for classification. The proposed method, tested with two-fold cross validation on experimental data, yields high recognition rate (over 96%).