AI-aided design and multi-scale optimization of mechanical metastructures with controllable anisotropy

Abstract

Anisotropic mechanical metamaterials with controllable properties are crucial for additive manufacturing design. However, manually regulating microstructural anisotropy remains challenging. This study introduces a method for artificial intelligence-aided design (AIAD) and mechanical metastructures optimization (MMO) to achieve extensive multi-scale structural enhancements. The approach involves compiling a comprehensive database of lattice materials with anisotropic characteristics. This is achieved by manipulating the central node position and rod diameter of a cubic-BCC microstructure. Homogenization theory then determines the elastic tensor of each microstructure. A 3D convolutional neural network (3D-CNN) maps the relationship between geometric properties and mechanical performance. An inverse design model based on a backpropagation neural network (NN) and parametric design acquires microstructures with desired elastic tensor attributes. Finally, a novel optimization approach for large-scale multiscale structures applies this method to control structural anisotropy. The resulting material distribution resembles a truss, significantly improving structural performance.