Health monitoring of milling tool using vibration signal analysis

Abstract

Milling process is one of the most widely used in the manufacturing industries. It is one of the effectively and efficiently used process to obtain the good surface-finished product. For this, milling tool used in the milling machine is rotated at variable speed. Sometimes, the rotational speed of the tool is very high. As the tool rotates at higher speed, an enormous amount of nondesirable vibration is generated. These unwanted vibration results in the impoverished surface quality of the workpiece and other defects. In this presented work, the emphasis has been given on the condition monitoring of the milling tool as the vibration sometimes cause the wear in rake or flank of the tool. Initially, vibration signals have been acquired with healthy as well as tool with rake and flank defects. The statistical features which is extracted from the time domain signals have been found to be a good indicator of the tool health. They are capable enough to determine the real-time health of the tool. Surface topography performed over the milled surface determines the variation of roughness parameters which confirms the information provided by condition indicators. In the succeeding section, Tunable Q-wavelet transform (TQWT) has been implanted in order to decompose the signals into different sub-bands with their respective energy level. The decomposition provides the better extraction of features form the vibration signals. For the automated fault detection in the milling tool, three different machine learning techniques i.e. artificial neural network (ANN), support vector machine (SVM) and decision tree (DT) have been used. ANN and DT are found to be capable with 100% detection capability whereas SVM could classify 98.10% of the samples.