Tensile properties prediction by multiple linear regression analysis for selective laser melted and post heat-treated Ti-6Al-4V with microstructural quantification

Selective laser melting (SLM), which is a powder bed fusion additive manufacturing method, is advantageous for fabricating of near-net-shape metal products. The microstructure of Ti-6Al-4V alloy fabricated by SLM and followed by post heat treatment intricately contributes to the mechanical properties. In this study, the microstructural features were quantitatively extracted from the scanning electron microscopy images by applying machine learning with a random forest algorithm and various image analysis techniques. These microstructural features together with defect-characterizing ones from X-ray computed tomography were used to develop a prediction model by multiple linear regression analysis for the tensile properties of heat-treated SLMed Ti-6Al-4V alloys. The analysis revealed that both yield strength (σ_YS) and ultimate tensile strength (σ_UTS) have a linear correlation with the reciprocal square root of the α grain size, and they were also attributed to other microstructural features depending on solution treatment parameters. The multiple linear regression models showed an error of less than 2.0% for σ_YS and σ_UTS prediction accuracies. On the other hand, the prediction of the fracture elongation and elasticity had significant scatter, implying that there are features, such as dislocation characteristics, that are missing for proper prediction.