AI-assisted design of 3D NPR lattice materials with programmable mechanical properties via irregular unit cells

Abstract

Lattice materials with negative Poisson’s ratios (NPR) exhibit exceptional mechanical properties, but their design has largely been limited to periodic cell structures, constraining their anisotropic potential. Irregular lattice cell architecture offers superior tunability, yet the complex relationship between their non-cyclic geometries and metamaterial properties has posed significant design challenges. Here, we introduce an AI-driven framework combining deep neural networks and genetic algorithms to parametrically optimize the anisotropic NPR and energy absorption of irregular 3D lattice cells. Through microscale and macroscale 3D printing, coupled with in-situ and quasi-static compression tests, we experimentally validate the programmable NPR effects across varied materials and scales. Micro-DIC analysis reveals the strain localization patterns governing microscale deformation and pinpoints the critical buckling instabilities in compressed architectures. Our approach enables the inverse design of 3D lattice metamaterials composed of irregular unit cells with tailored mechanical properties, unlocking new possibilities for applications in lightweight structures, energy absorption, and beyond.