Nonlinear bending of sandwich plates with deep learning inverse-designed 3D auxetic lattice core

Based on generative deep learning (DL), this paper innovatively proposes a sandwich panel with inverse-designed 3D auxetic cores. Furthermore, we demonstrate that the bending performance of such data-driven sandwich structure is superior to existing design which consists of 3D lattice core evolved from the traditional 2D re-entrant honeycomb design. The deflection related to proposed DL-based sandwich plate is nearly one-third of that in existing studies. Through metal additive manufacturing techniques and finite element (FE) modeling, flexural behaviors of the inverse-designed 3D truss unit cell with different geometric factor is further examined. With different functionally graded core configurations and uniform distributed as comparison group, parametric studies are conducted to investigate the effect of various dimensional parameters, thermal environments and boundary conditions on the nonlinear bending behaviors and effective Poisson's ratio of sandwich plates subjected to a uniform pressure. This work provides a reference for the study of mechanical properties of novel auxetic sandwich structures and has potential guiding implications for accelerating the design process and performance optimization of sandwich plates.