Deep learning-based cutting force prediction for machining process using monitoring data

Abstract

Machining is a critical process in manufacturing industries. With the increase in the complexity and precision of machining, computer systems, such as computerized numerical control, machining monitoring systems (MMSs), and virtual machining (VM), have been incorporated in modern machining processes. In this study, a deep learning-based cutting force prediction method was proposed. MMS and VM data were collected from real-world machining processes. Next, the prediction of the cutting force using five deep learning-based methods, including the long short-term memory (LSTM) and temporal convolutional networks, were analyzed and compared with values measured with a tool dynamometer. The experimental results revealed that the proposed LSTM model, including bidirectional and residual structures, outperformed other benchmark models in terms of predicting the cutting force. Furthermore, the proposed method trained only with MMS data exhibited excellent performance with a root-mean-square error of 12.55 and R2 of 0.99 on average. Thus, the cutting force required at each point can be predicted accurately, and the method can become a reference for further studies. © 2023, The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature.