Effect of feature and sampling ratio on tool wear classification during boring operation using tree-based algorithms

Abstract

Tool condition monitoring (TCM) system is used to predict the tool wear during the machining process. The predominant wear is the flank wear which influences the surface roughness of the workpiece. The quantum of flank wear is to be ascertained so that a decision could be made whether the time has come for the insert to be replaced. Although the wear is continuous, it may be divided into three stages and may be classified as to which stage the tool wear falls into. Wear prediction may be carried out by extracting information from the vibration signals acquired during machining and interpreting them using machine learning. This paper focuses on monitoring the uncoated carbide tooltip during boring operation using tree based classifier algorithms such as random forest, J48, logistic model tree and gradient booted tree, in order to study the effect of feature and sampling ratio on tool wear classification when tree-based algorithms are used. Also, the statistical features and histogram features were compared for various cutting tool conditions to explore a better classifier-feature combination.