CALPHAD-based cross-system knowledge transfer for rapid discovery of high-performance Al–Mg–Zn alloys

Machine learning is widely used to accelerate materials design, but its performance remains limited in small-data scenarios, reducing effectiveness in guiding novel materials development. Here, we developed a novel representation transfer framework based on the Calculation of Phase Diagrams (CALPHAD) method to address the challenge. By integrating CALPHAD with feature engineering, the thermodynamics-informed descriptors related to strength and ductility were constructed and selected. These descriptors capture the coupled effects of composition and temperature on material properties, offering strong physical interpretability and enabling a knowledge transfer from well-studied 2xxx, 6xxx and 7xxx series aluminum alloys to the underexplored Al–Mg–Zn alloys. Subsequently, by coupling high-throughput CALPHAD calculations with the NSGA-II algorithm, the strength–ductility Pareto front is efficiently identified to guide alloy design. Experimental validation confirmed that the two designed alloys achieved ultimate tensile strengths of 472 ± 7 MPa and 569 ± 12 MPa, with elongations of 23.5 ± 0.5% and 14.9 ± 0.3%, respectively, demonstrating improved strength–ductility synergy in the Al–Mg–Zn system. This framework enhances model generalization and interpretability in small-data scenarios, offering a versatile strategy for rapid discovery of high-performance materials across diverse systems.