Experimental investigation and laser control in Ti10Mo6Cu powder bed fusion: optimizing process parameters with machine learning

Abstract

Abstract Laser Powder Bed Fusion (LPBF) is a key additive manufacturing technique, yet achieving optimal track formation remains a challenge due to the complex interplay of laser parameters. This study presents a hybrid experimental and machine learning (ML) approach to enhance laser control in the LPBF process for Ti10Mo6Cu alloys. Experimental investigations were conducted to analyze track morphology, surface roughness, and microhardness under varying laser power and scanning speed conditions. A Gradient Boosting Decision Tree (GBDT) model was developed to predict track characteristics which include width, height, and depth, based on experimental data. The model demonstrated exceptional accuracy, achieving an R² of 0.9812, RMSE of 0.8547, and MAE of 0.6231 for the forecasted average width from the cross-sectional and top views, while other predictions also showed high correlations—such as R² values of 0.9786 for average track height and 0.9652 for average depth predictions. A novel ML-assisted laser control framework was proposed by the means of integrating GBDT predictions into a feedback-based optimization system to dynamically adjust process parameters. The findings indicate that ML-driven control strategies can significantly enhance track uniformity, minimize defects, and improve the repeatability of LPBF processes. While the study successfully optimized process outcomes, future work should focus on real-time adaptive control mechanisms and the expansion of training datasets to further refine predictive capabilities. This research contributes to bridging the gap between experimental optimization and intelligent process automation in advanced laser-based manufacturing.

Keywords

• Machine learning
• Laser powder bed fusion
• Control
• Gradient boosting decision tree