Intelligent tool wear monitoring using XGBoost, SVR, and DNN models in NMQL environment

This study explores the implementation of artificial intelligence (AI)-based predictive frameworks for the precise evaluation of tool wear in the sustainable machining of Hastelloy X using PVD TiAlN-coated carbide inserts. To improve lubrication effectiveness and restrain excessive tool degradation under severe thermo-mechanical cutting conditions, a minimum quantity lubrication (MQL) strategy aided with a carbon nanotube (CNT)-based nanofluid was adopted. Tool wear evolution was modelled using advanced machine learning approaches, including Extreme Gradient Boosting (XGBoost), Deep Neural Networks (DNN), and Support Vector Regression (SVR), with key machining parameters serving as the primary input variables. Experimental investigations demonstrated that CNT-based MQL substantially reduced tool wear, with an optimal nanoparticle concentration of 0.6%, attributed to improved heat dissipation and superior tribological behaviour at the machining zone. Among the implemented models, XGBoost exhibited the highest predictive accuracy, attaining an R² of 0.9924 along with minimal error indices, including MAE of 0.0017, RMSE of 0.002, and MAPE of 0.6%. In contrast, DNN and SVR showed comparatively poor predictive capability for the evaluated dataset split, reflected by low or negative R² values, highlighting the importance of model selection and data sensitivity in tool wear prediction tasks. Sensitivity analysis based on Spearman correlation revealed that cutting speed exerted the most dominant impact on tool wear (correlation coefficient = 0.94), followed by feed rate and depth of cut. Overall, the outcomes indicate that CNT-based nano-MQL combined with appropriately selected AI models-particularly XGBoost-provides a robust pathway for extending tool life, enhancing machinability, and enabling intelligent tool condition monitoring aligned with Industry 4.0 and sustainable manufacturing paradigms.