• TiN-coated ceramic insert improves wear resistance. • Boosting models achieve the highest accuracy for force and roughness prediction. • Multi-objective GA enables balanced optimization of Fz, Ra, and VB. • GUI implementation supports practical decision-making in conventional turning. • System bridges experimental research with shop-floor machining practice. This study presents an integrated experimental and data-driven framework for analyzing, predicting, and optimizing the turning of EN-GJL-250 grey cast iron using silicon nitride (Si₃N₄) ceramic inserts. Two insert grades are used: an uncoated ceramic insert (CC6090) and a TiN-coated ceramic insert (CC1690) under 5 controllable input parameters were considered in the experiment: insert type, depth of cut, feed per revolution, cutting speed, and machining length in order to assess the tangential cutting force (Fz), surface roughness (Ra), and flank wear (VB). The ANOVA and response plots are used to evaluate the effects of cutting parameters on machining performance. Their results indicate that feed rate is the dominant factor affecting surface roughness, with a contribution of 55.98%, while machining length plays a major role in tool wear progression. To enable data-driven prediction, five machine learning models (SVR, RF, GBM, XGBoost, and ANN) were compared to predict Ra, Fz, and VB. Where The XGBoost achieved the highest accuracy for cutting force prediction (with R² ≈ 0.943), while GBM/XGBoost provided a good prediction of surface roughness (R² ≈ 0.914-0.916), and GBM/ANN yielded the highest flank wear prediction accuracy (with R² ≈ 0.968). The trained prediction models are then coupled with a multi-objective Genetic Algorithm to identify Pareto-optimal cutting conditions to obtain the desired surface quality, tool wear, and cutting force. Finally, the complete prediction and optimization workflow is implemented in an interactive graphical user interface, in order to provide a practical decision-support tool for conventional turning applications.
Nouioua et al. (Sun,) studied this question.