Machine Learning-Driven Optimization of Computer Numerical Control Milling Parameters for Enhanced Surface Integrity and Burr Control in Additive Manufacturing

Abstract This study presents a comprehensive experimental and machine learning-based investigation into the effects of critical machining parameters—cutting speed (S), feed rate (F), and cutting depth (dp) on the burr formation and surface roughness (Ra) during computer numerical control milling of additively manufactured Ti-6Al-4V alloy. Statistical analyses, including factorial and one-way analysis of variance, identified cutting depth as the most influential parameter affecting burr formation (F = 171.379, p < 0.001), while cutting speed had the greatest effect on surface roughness (F = 39.324, p < 0.001). Burr formation ranged from 0.09 mm to 1.46 mm, and surface roughness varied between 0.07 µm and 0.22 µm, highlighting the sensitivity of machining performance to process parameters. To model and predict machining outcomes, six machine learning algorithms—decision tree, random forest, K-nearest neighbors, gradient boosting, LightGBM, and CatBoost were evaluated. Gradient boosting achieved the highest predictive accuracy for burr formation (test R2 > 0.98), while LightGBM and gradient boosting also performed best for Ra prediction (test R2 > 0.81). Feature importance and Shapley additive explanations analysis using XGBoost revealed that cutting depth contributed over 90% to burr formation prediction, reaffirming its dominant role. The integration of statistical and machine learning approaches enabled a dual-response optimization strategy, minimizing burrs while enhancing surface quality, providing a data-driven framework for precision process control. These findings offer actionable insights for high-performance machining of Ti-6Al-4V and underscore the value of intelligent optimization techniques in advanced manufacturing applications.