Fault diagnosis of single point cutting tool through discrete wavelet features of vibration signals using decision tree technique and multilayer perceptron

Abstract

Tool condition monitoring in machining plays a crucial role in modern manufacturing systems, finding state of the tool wear in early with the help of condition monitoring system will reduce downtime and excessive power drawing while machining. Vibration analysis of mechanical systems can be used to identify the tool condition to distinguish good or worn tool. In this study, vibration signals were acquired during turning operation, fault diagnosis using machine learning techniques has been carried out with new and different type of worn-out tool inserts. Discrete Wavelet Features (DWT) were extracted from acquired vibration signal for various cutting tool conditions using MATLAB. Most significant features were selected out of extracted discrete wavelet features using decision tree technique (J48 algorithm). Multilayer perceptron has been used as a classifier, selected features were given as input for the classifier. The classification accuracy with multilayer perceptron was found to be 96%. KRISHTEL eMAGING SOLUTIONS PVT. LTD.