Machine Learning-Driven Prediction of Surface Roughness in High-Speed Milling of Titanium Alloys

Abstract

Titanium alloys are widely used in aerospace and biomedical industries but present challenges in machining due to high cutting temperatures and rapid tool wear. This study integrates experimental trials with machine learning approaches to predict surface roughness in high-speed milling of Ti-6Al-4V alloy. Machining parameters including cutting speed, feed rate, and depth of cut were varied, and corresponding surface roughness values were measured using a profilometer. Data-driven models using support vector regression (SVR), random forest (RF), and artificial neural networks (ANNs) were developed. Among these, ANN achieved the highest prediction accuracy with an R² value of 0.96, outperforming conventional regression models. Feature importance analysis indicated feed rate as the most influential parameter, followed by cutting speed. Optimization studies suggested that surface roughness below 0.8 μm could be consistently achieved at moderate feed rates and high cutting speeds. The hybrid approach of combining machining experiments with machine learning offers a robust framework for process optimization, reducing trial-and-error methods and enhancing productivity in titanium machining.