2025 Volume 19 Issue 1 Pages JAMDSM0001
Accurate prediction of tool life is crucial for reducing production costs and enhancing quality in the machining process. However, such predictions often rely on empirical knowledge, which may limit inexperienced engineers to reliably obtain accurate predictions. This study explores a method to predict the tool life of a cutting machine using servo motor current data collected during the initial stages of tool wear, which is a cost-effective approach. The LightGBM model was identified as suitable for predicting tool life from current data, given the challenges associated with predicting from the average variation of current values. By identifying and utilizing the top 50 features from the current data for prediction, the accuracy of tool life prediction in the early wear stage improved. As this prediction method was developed based on current data obtained during the very early wear stage in experiments with square end-mills, it was tested on extrapolated data using different end-mill diameters. The findings revealed average accuracy rates of 71.2% and 69.4% when using maximum machining time and maximum removal volume as thresholds, respectively.