Generating Periodic Lattice Truss Materials and Determining Mechanical Properties Based on Canonical Descriptors

Abstract The development of additive manufacturing and artificial intelligence spurs an explosion in the methods exploring the design space of periodic lattice truss materials (PLTMs) in recent years. However, a normative description with completeness and uniqueness has not yet been proposed, which confines the design to some discrete and small scrutinized space. Based on their intrinsic nature, we here theoretically develop a system of canonical descriptors for PLTMs, leading to concise descriptions and sufficient information, to establish good quantitative correlations between structures and mechanical behaviors. The system mainly consists of the geometry matrix for the node configuration, density, stretching and bending stiffness matrices for the strut properties, as well as a packing matrix for the periodic tessellation orientation. The numerical characteristics of the descriptors corresponding to the PLTMs are discussed, and the completeness and uniqueness are proved. Using the canonical descriptors and the machine learning (ML) method, we generate more than 20,000 PLTMs with the corresponding elastic constants. By the analysis of the database, we visualize the vast but discrete property space, in which anisotropic PLTMs are predominant while isotropic categories are rare. Our study provides an insight into the fundamental definition and understanding of PLTMs, contributing to new material discovery with expected diversiform behaviors, especially in the current age of artificial intelligence.