Machine learning-aided design of metallic mechanical metamaterials: fabrication-aware modelling, property prediction, and inverse design

Machine learning (ML) has emerged as a powerful framework for metallic mechanical metamaterials by helping disentangle the complex interplay among digital architectures, additive manufacturing processes, and as-built defects. By circumventing costly trial-and-error physical prototyping, ML accelerates the design and qualification pipelines, fully unlocking the intrinsic metallic advantages and topology-governed mechanical behaviours of these advanced structures. This review first outlines the major classes of metallic mechanical metamaterials alongside their representative additive manufacturing technologies. Subsequently, we explore how ML bridges the gap between design intelligence and manufacturing reality by actively accounting for nominal topologies, process sensitivities, and fabrication-induced deviations. Specifically, we detail how data-driven approaches facilitate the forward prediction of mechanical responses and the inverse design of complex architectures. Finally, current challenges and future directions are discussed, with emphasis on data quality, validation level, structural representation, and the development of a closed-loop digital-to-physical workflow for metallic mechanical metamaterials.