Microstructural feature-driven machine learning for predicting mechanical tensile strength of laser powder bed fusion (L-PBF) additively manufactured Ti6Al4V alloy

The rapid solidification inherent in laser powder bed fusion (L-PBF) additive manufacturing (AM) introduces segregation phenomena and formation of non-equilibrium phases in duplex titanium alloy components, thereby impeding their suitability for high-reliability engineering applications. Consequently, heat treatment becomes indispensable for optimizing both the microstructure and mechanical properties to meet application requirements. This study aims to investigate the influence of varied annealing temperatures on the evolution of L-PBF-built Ti6Al4V alloy microstructure, subsequently elucidating their impact on tensile properties by analyzing of L-PBF process parameters, building orientations, and annealing temperatures. The findings reveal that annealing at 850 °C for 2 h facilitates the transformation of brittle martensite into a ductile lamellar (α + β) microstructure, thereby conferring excellent tensile properties upon the L-PBF-built Ti6Al4V alloy. Furthermore, to accurately predict the tensile strengths of the Ti6Al4V, we take into account the L-PBF process parameters and the as-built microstructures in a comprehensive manner, extracting the pertinent microstructural features. A machine learning (ML)-based model is built to facilitate accurate predictions. Accurate and reliable predictions are demonstrated by this model when applied to Ti6Al4V. This data-driven approach establishes a novel avenue for AM material property prediction and process parameter optimization.