Mechanical metamaterials are architected materials that exhibit unusual mechanical properties arising from their internal geometry. This paper reviews recent developments in the application of machine learning for the design and analysis of these structures. It categorizes common architectures, including strut-based lattices and triply periodic minimal surfaces, and details the end-to-end design workflow, from dataset preparation and preprocessing to the iterative, simulation-based validation approach. The review compares a range of model architectures. These include foundational models like deep neural networks, fully connected and convolutional neural networks, graph neural networks, and generative models such as GANs and diffusion models. Applications in mechanical property prediction and inverse design are highlighted with examples using finite element simulations and generative design models. A structured design workflow and a comparative summary of recent studies are presented to guide future research and application. This review aims to support the development of ML frameworks for next-generation metamaterial design.
Turysbekov et al. (Tue,) studied this question.