Genetic design of new aluminum alloys to overcome strength-ductility trade-off dilemma

• Published in: Journal of alloys and compounds 2023-06, Vol.947, p.169546, Article 169546
• Main Authors: Lee, Keunwon, Song, Yongwook, Kim, Sehoon, Kim, Minsang, Seol, Jaebok, Cho, Kisub, Choi, Hyunjoo
• Format: Article
• Language: English
• Subjects: Aluminum alloys; Genetic algorithm; Inverse design; Machine learning ensemble model; Mechanical properties
• ISSN: 0925-8388
• DOI: 10.1016/j.jallcom.2023.169546

In this study, machine learning and inverse design based on a genetic algorithm was used to design three aluminum wrough alloy types to overcome the strength-ductility trade-off. The composition of the new alloys was advantageous in relation to that of commercial alloys, and this was experimentally validated using samples produced by a semi-mass-production-scale process. The relationship between microstructures and mechanical properties was exploited to characterize the alloys, and each alloy exhibited different precipitation types. The major precipitate of alloy 1 was the spheroidal α-AlMnSi phase, which contributed to the Orowan mechanism. In contrast, the major precipitate of alloys 2 and 3 was the fine needle-type θ-series phase, which contributed to the dislocation shearing mechanism. The new alloys showed outstanding tensile strength (431.69, 527.03, and 527.79 MPa) without a decrease in ductility. These findings suggest that machine learning and inverse design methods are suitable for discovering new aluminum alloy types. •Artificial intelligence was used to design three new aluminum wrought alloys with improved mechanical properties.•Studied microstructures & mechanical properties of aluminum alloys manufactured via semi-mass-production process.•AI-based design effective for developing aluminum alloys without strength-ductility trade-offs in conventional alloys.