Additive manufacturing metallurgy guided machine learning design of versatile alloys

Additive manufacturing (AM) is distinguished by its near-net-shape fabrication capability, enabling single-step production of geometrically complex components. However, unlike conventional manufacturing processes, AM-fabricated parts generally lack post-process thermo-mechanical treatments. As a result, the performance of AM-built materials is predominantly governed by their composition and the thermal history inherent to AM. This underscores the necessity for developing materials dedicated to AM. To address this challenge, this study introduces an AM metallurgy-guided machine learning (ML) alloy design framework aimed at developing high-performance AM-specific alloys. The framework combines high-throughput thermodynamic simulations with ML surrogate models to predict key AM-oriented properties, including solidification freezing range, growth restriction factor, hot cracking susceptibility, and carbide precipitation speed. These AM-oriented properties are optimised through multi-objective optimisation and decision-making to design alloys with optimal AM performance. To validate this framework, pre-alloyed powders of a designed novel alloy were prepared and printed using various laser-directed energy deposition strategies. Comprehensive characterisations confirmed that the resulting microstructures and properties aligned well with the AM-oriented design objectives. Remarkably, the novel alloy exhibited superior yet highly tunable mechanical properties, with yield strength ranging from 1062 to 1769 MPa and uniform elongation varying between 2.1 % and 11.7 %, depending on the printing strategy. The superior yet tunable mechanical properties are attributed to the temperature-dependent phase transformations and rapid carbide precipitation kinetics of the novel alloy. Overall, this study establishes a robust data-driven framework for AM-specific alloy design, providing a powerful tool to reliably accelerate the development of high-performance and versatile alloys for AM. © 2025 Elsevier Ltd.