Abstract:

Aluminum alloys with ultra-strength and high-toughness are fundamental structural materials applied in the aerospace industry. Due to the intrinsic restriction between strength and toughness, optimizing a desirable combination of these conflicting properties is always challenging in material development. In this study, 171 sets of data were curated based on the characteristics of high-strength and high-toughness aluminum alloys in the literature. Then, a machine learning design system (MLDS) with a property-oriented design strategy was established to rapidly discover novel aluminum alloys with ductility and toughness indexes (with elongation δ=8%-10% and fracture toughness K_IC=33-35 MPa·m ^1/2) comparable to those of current state-of-the-art AA7136 aluminum alloys when the ultimate tensile strength (UTS) exceeded approximately 100 MPa, with values reaching 700-750 MPa. With the MLDS for experimental verification, three typical candidate alloys show satisfactory performance with UTS of 707-736 MPa, δ of 7.8%-9.5%, and K_IC of 32.2-33.9 MPa·m ^1/2. The high contents of Mg and Zn alloying elements in the novel alloys form abundant η′ phases, which produce a significant hardening effect, while the reasonable matching of Cr, Mn, Ti and Zr dispersoids refines the grain size. The decreased Cu content compared with that in the AA7136 alloy inhibits the formation of the σ phase and S phase, so that the alloys show high toughness.

Key words: Aluminum alloy, Machine learning, Composition design, Hardening mechanism, Toughness