Data-driven diffusion generative design of energy-absorbing metamaterials using implicit surface representation

Abstract

Currently, the design of energy-absorbing materials and structures primarily relies on empirical or heuristic methods. Motivated by advances in generative artificial intelligence techniques, a data-driven diffusion generative approach using implicit surface representation for energy-absorbing metamaterial design is proposed. This approach utilizes the diffusion model to learn the conditional distribution of metamaterial based on specified mechanical properties, converting target properties into potential metamaterial topologies. Additionally, level set-based implicit surface representation ensures that the generated metamaterials have high-quality geometric shapes and clear boundary definitions, enhancing design flexibility while requiring fewer design variables. Numerical simulations and experimental results consistently verify that the proposed approach enables rapid and accurate design of metamaterials tailored to target mechanical performance. This method offers an innovative and efficient solution for the accelerated design of energy-absorbing metamaterials, providing a new perspective and approach to address complex engineering design challenges.