The rapid discovery of multi-component alloys with optimized properties is challenging, requiring both computational efficiency and precision. Recent Bayesian optimization (BO) advancements [1] in continuous design spaces enhance alloy composition optimization, identifying novel alloys with targeted properties. Multi-objective BO acquisition functions like TSEMO, parEGO, and qNEHVI are benchmarked for efficiency and robustness. To accelerate development, an ICME framework integrates CALPHAD-based simulations for phase stability and thermodynamic predictions, autonomously selecting acquisition functions based on material properties and objectives. Design of Experiments (DOE) strategies minimize validation iterations, reducing timelines, while pool-based active learning enables concurrent computational and experimental work in high-throughput environments. Multi-scale modelling correlates microstructure with performance, with uncertainty quantification ensuring robustness. Data-driven machine learning enables real-time predictions and feedback loops, refining optimization iteratively. This framework, combining advanced optimization, simulations, and experimental validation, provides a scalable solution for rapid discovery and qualification of novel alloys.

[1]  O. Mamun, M. Bause, and B. S. M. Ebna Hai; Accelerated development of multi-component alloys in discrete design space using Bayesian multi-objective optimisation, Machine Learning: Science and Technology, 2025, volume 6, number 1, 015001, DOI: 10.1088/2632-2153/ada47d