Multi-axed phase-transforming cellular material: A data-driven design and validation using finite-element method and machine learning

We developed the novel Atom-Mimetic Cube-Diagonally Multi-Axed Phase-Transforming Cellular Material (AMCDMA-PXCM), hereafter AM-PXCM for short, for a multi-axial bistable metamaterial designed with inspiration from a face-centered cubic (FCC) crystal structure: the designed AM-PXCM consists of spheres at atomic positions of structure and dogleg-shaped beams connecting nearest neighbor spheres. Stress-strain relationship of AM-PXCM was investigated by Finite Element Method (FEM) simulation. Analyzing the results by Logistic classification revealed that the mechanical properties significantly depend on the designing parameters and the distance between the beam and the tetrahedron ( k ) dominantly determines the bistability of the FCC-based AM-PXCM. In addition, combined with the machine learning method (i.e., inverse design), we succeeded to predict the designing parameters to have the desired mechanical properties for a bistable metamaterial. The designed AM-PXCMs were realized using a 3D printer and validified to show the predicted mechanical properties. This established method for developing AM-PXCM is suggested to be also applied to a development of an AM-PXCM with the symmetry of other crystal structures. • Novel multi-axial bistable metamaterial was designed with inspiration from a face-centered cubic crystal structure. • Finite Element Method simulations revealed that the bistability largely depends on the designing parameters. • The designing parameters with desired mechanical properties are succeeded in inversely predicted by the machine learning. • The prediction accuracy was verified through the actual testing of the designed metamaterials fabricated by 3D printing.