High Layer Thickness Additive Manufacturing of Ti–6Al–4V: Experiments and Ensemble Machine Learning

To increase production efficiency is one of the most critical development directions for the powder bed fusion–laser beam additive manufacturing. To address this, enlarged layer thickness (up to 150 μm) was examined in this study, with optimization carried out using ensemble machine learning and using Ti–6Al–4V alloy as a typical example alloy. Five experimental parameters, namely powder particle size, laser power, scan speed, hatch spacing, and layer thickness, were selected as features to study their influence on three material performances: relative density, hardness, and compressive strength. A total of 460 printing experiments were conducted, whose databases were collected and analyzed. Meantime, RF, XGBoost, LightGBM, WOA-XGBoost, and DBO-XGBoost were used to establish regression models. The results show that the WOA-XGBoost hybrid model has achieved good coefficient of determination ( R 2 ) values and it has the best prediction performance based on an optimal comprehensive performance model (Des), whose prediction was experimentally verified. The study eventually shows that high strength and high relative density can be achieved using layer thickness of ∼60 μm, while further enlarging layer thickness tends to reduce mechanical performances.