Comparative evaluation of machine learning models for thrust force estimation in aluminum 5083 H111

Abstract This study investigates the comparative performance of four different supervised machine learning algorithms for the prediction of maximum thrust force F z max during the drilling process of aluminum 5083 H111 alloy. Accurate prediction of the forces generated during drilling operations is of great importance for extending tool life, increasing hole quality and improving production efficiency. In this direction, a total of 27 experimental conditions were created including three different cutting speeds (80, 100, 120 m min −1 ), three different feed rates (0.06, 0.09, 0.12 mm tooth −1 ) and three different cooling strategies (dry, air, liquid) and the maximum thrust force was measured for each condition. Artificial neural networks (ANN), decision trees (DT), random forest (RF) and extreme gradient boosting (XGBoost) algorithms were trained with the obtained data set and evaluated by MAE, RMSE and R 2 performance metrics. The findings revealed that the XGBoost algorithm achieved the lowest error rates (MAE: 56.742 N, RMSE: 67.179 N) and the highest explanatory power ( R 2 : 0.929) compared to other models. These results show that ensemble-based methods can be used as reliable and effective estimation tools in production processes.