Inverse machine learning framework for optimizing lightweight metamaterials

Abstract

Abstract Structure scouting and design optimization for superior mechanical performance through inverse machine learning is an emerging area of interest. Inverse machine learning can be a substantial approach in structural design to explore complex and massive numbers of geometrical patterns within short periods of time. Here, an inverse design framework using generative adversarial networks (GANs) is proposed to explore and optimize structural designs such as lightweight lattice unit cells. Lightweight lattice structures are widely accepted to have excellent mechanical properties and have found applications in various engineering structures. Using the proposed framework, different lattice unit cells that are 40–120% better in load carrying capacity than octet unit cell are discovered. These new lattice unit cells are analyzed numerically and validated experimentally by testing 3D printed lattice unit cells and lattice cored sandwiches. The proposed inverse design framework can be applied to the design and optimization of other types of load bearing structures.