Pre-processing for vibration signals features extraction and selection in real time investigating of CNC tool wear

Abstract

This research attempts to investigate CNC milling tool wear based on machining vibration signals. Vibration data acquisition uses devices from NI DAQ USB-6008 and the analog accelerometer sensors, MMA 7361. Vibration feature extraction is done based on statistical analysis of time domain vibration signal and order domain vibration analysis. Feature selection using filter approachment with the Linear Discriminant Classifier, successfully selected 10 main features. Whereas the multi-layer perceptron is used as the final classifier. When implemented on the CNC milling process, this system successfully detects tool wear with an accuracy of 96.4%. Error detection that still occurred consisted of 4.4% missed alarm and 2.8% False alarm.