Generative Adversarial Networks for the Inverse Design of 2D Spinodoid Metamaterials

The geometrical arrangement of metamaterials controls their mechanical properties, such as Young's modulus and shear modulus. However, optimizing the geometrical arrangement for user-defined performance criteria leads to an inverse problem that is intractable when considering numerous combinations of properties and underlying geometries. Machine learning techniques have been proven to be effective and practical to accomplish such non-intuitive design tasks. This paper proposes an inverse design framework using conditional generative adversarial networks (CGANs) to explore and optimize 2D metamaterial designs consisting of spinodal topologies, called spinodoids. CGANs are capable of solving the "many-to-many" inverse problem, which requires generating a group of geometric patterns of representative volume elements with target combinations of mechanical properties. The performance of the networks was validated by numerical simulations with the finite element method. The proposed inverse design framework vastly improves the efficiency of design exploration and optimization of spinodoid metamaterials.