Data underlying the publication: Inverse-designed growth-based cellular metamaterials

Our project aims to explore the design space of growth-based cellular metamaterials using a deep learning framework. These two-dimensional materials derive their properties from their microstructure rather than just their constituent material. We employ large datasets to develop forward and inverse models for designing metamaterials with tailored anisotropic stiffness. The forward model predicts mechanical properties based on design parameters, while the inverse model allows for the accurate prediction of designs based on anisotropic stiffness queries. Our framework's generalization capabilities are demonstrated by successfully designing for stiffness properties outside the design space domain. Here, we share the dataset we used to train our framework. More information on how to generate more data can be found in the README of this repository.

本项目旨在依托深度学习框架（deep learning framework），探究基于生长机制的胞状超材料（cellular metamaterials）的设计空间。这类二维材料的性能并非仅由其组成基体材料决定，而是源于其微观结构。本研究借助大规模数据集，开发了面向定制化各向异性刚度（anisotropic stiffness）超材料设计的正向模型与逆向模型：正向模型可基于设计参数预测材料的力学性能，而逆向模型则可根据给定的各向异性刚度需求，精准输出对应的设计方案。本框架的泛化性能已得到验证——我们成功为设计空间域外的刚度特性完成了超材料设计。在此我们共享了用于训练本框架的数据集，更多关于生成额外数据的细节，请参见本仓库的README文件。