Prediction of the tensile properties of A356 casted alloy based on the pore structure using machine learning

Abstract

Computed tomography (CT) is increasingly used to investigate porosity, which significantly affects the mechanical properties of castings. The aim of this study was to explore the relationship between the tensile properties – yield strength (YS), ultimate tensile strength (UTS) and elongation – and the pore structure of A356 castings produced under industrial conditions with low-pressure die casting (LPDC) technology. The novelty of our method lies in determining relationships not only with bivariate analyses but also by applying regression modeling using machine learning (ML). CT images of the test specimens were generated, and the pores in both two and three dimensions were quantitatively characterized. After the tensile tests, the fracture surfaces were numerically characterized using scanning electron microscope (PFIB-SEM) images. Prior to regression modeling, an exploratory data analysis (EDA) was conducted. Based on the tests, it was concluded that the findings for tensile strength are partially consistent with the literature, while those for yield strength are entirely consistent. Furthermore, it was newly observed that fracture surface porosity (A_{A, Proj}^f-sf) is influenced by the projected area of the largest volume pore in the fracture segment (A_{Proj, Vmax}^f-sm), the maximum porosity of the fracture segment cross-sections (A_A^f-cs), and the overall porosity of the fracture segment (V_V^f-sm). Another new finding is that total elongation at rupture is significantly affected not only by the fracture surface porosity (A_{A, Proj}^f-sf) but also by the parameter (CSC) introduced in this study, which characterizes the pore location on the fracture cross-section. The regression modeling performed this way successfully complemented the results obtained with bivariate analyses. The presented method is suitable for characterizing the relationship between the pore structure and mechanical properties of castings produced in industry.