Predicting tool wear with multi-sensor data using deep belief networks

Abstract

Tool wear is a crucial factor influencing the quality of workpieces in the machining industry. The efficient and accurate prediction of tool wear can enable the tool to be changed in a timely manner to avoid unnecessary costs. Various parameters, such as cutting force, vibration, and acoustic emission (AE), impact tool wear. Signals are collected by different sensors and then constitute the raw data. There are two main types of methods used to make predictions, namely model-based and data-driven methods. Data-driven methods are typically preferred when a mathematical model is not available. In such a situation, artificial intelligent methods, such as support vector regression (SVR) and artificial neural networks (ANNs), are applied. Recently, deep learning algorithms have been widely used because of their accuracy, computing speed, and excellent performance in solving nonlinear problems. In this study, a deep learning network called deep belief network (DBN) is applied to predict the flank wear of a cutting tool. To confirm the superiority of the DBN in predicting tool wear, the performance of the DBN is compared with the performances obtained using ANNs and SVR in terms of the mean-squared error (MSE) and the coefficient of determination (R2), considering data from more than 900 experiments.