Analysing surface quality evolution and cutting forces in micro-milling of Ti6Al4V using experimental and machine learning approaches

In modern manufacturing, machine learning (ML) plays a crucial role in predicting response measures, reducing machining trial costs and saving energy. This study investigates surface evolution and cutting forces (F_avg) during the micro-milling of selective laser-melted (SLM) Ti6Al4V, focusing on spindle speed (SS), depth of cut (D_OC), and feed rate (F_Z). Analysis of variance (ANOVA) identified F_Z as the most influential on surface roughness (Ra) of about 38.54% and D_OC on F_avg (42.16%). An artificial neural network (ANN) was trained to map the process-performance relationship, showing strong correlation with experimental results, thereby reducing material and resource costs. Additionally, a ML-based non-dominated sorting genetic algorithm (NSGA-II) was used for multi-objective optimization, achieving significant improvements in Ra and F_avg by 335.61% and 592.61%, respectively, with optimal parameters of SS = 74,994 RPM, D_OC = 19.94 μm, and F_Z = 1 μm/tooth. Detailed analysis of F_avg signals, tool wear, and surface evolution was conducted using scanning electron microscopy (SEM). The cumulative regression coefficient (R) for the ANN model is 0.99669, which demonstrates a strong correlation between the predicted values and the actual outcomes. The developed ML model will help the manufacturing industry by reducing the costs for materials, resources, and trial runs.