Research Information System
12th International European Conference on Interdisciplinary Scientific Research, Rome, Italy, 11 - 13 July 2025, pp.695-705, (Full Text)
This study introduces a light gradient boosting machine (LightGBM)-based data-driven modeling approach for predicting cutting parameters in band sawing operations involving high-hardness tool steels. The analysis is based on a dataset comprising 1134 samples, incorporating two industrially significant tool steels from different application classes: 1.2379 (AISI D2), a cold work tool steel, and 1.3343 (AISI M2), a high-speed tool steel. Each sample is defined by hardness values ranging from 15 to 44 HRC and diameters between 100 and 500 mm. Cutting parameter values were derived from a combination of manufacturer recommendations and experimentally validated data. Model performance was evaluated using four regression techniques, namely least absolute shrinkage and selection operator (LASSO), support vector regression (SVR), artificial neural network (ANN), and LightGBM. Among them, LightGBM outperformed the others in capturing the nonlinear relationship between material properties and machining outputs. In predicting cutting speed, LightGBM achieved a root mean square error (RMSE) of 0.222, a mean absolute error (MAE) of 0.132, a coefficient of determination (R²) of 0.999, and a mean absolute percentage error (MAPE) of 0.51%. For feed rate estimation, the same model yielded an RMSE of 0.187, MAE of 0.131, R² of 0.999, and MAPE of 1.42%. These results demonstrate the strong predictive power and generalizability of LightGBM in capturing material-dependent cutting dynamics, clearly outperforming both linear and kernel-based alternatives. Additionally, this study emphasizes the promising role of advanced machine learning techniques in optimizing process planning and facilitating informed decisions within intelligent manufacturing environments.