Nonlinear free vibration of sandwich beam with data-driven inverse-designed auxetic core based on deep learning

Abstract

This paper presents a novel data-driven inverse design approach for auxetic sandwich beams using a deep generative model (DGM). By integrating a conditional estimator with enhanced loss backpropagation, the DGM can produce 3D truss auxetic unit cells exhibiting the desired negative Poisson's ratio, which serve as the microstructures for the auxetic core of sandwich beams. Employing finite element analysis and advanced 3D metal printing techniques, both numerical and experimental investigations on the linear and nonlinear free vibration characteristics of the 3D lattice specimens are conducted. Remarkably, the free vibration analysis results demonstrate that auxetic sandwich beams designed with DGM achieve significantly higher natural frequencies than those optimized using common topological approaches. We conclude that the proposed DGM-based inverse design methodology holds substantial promise within the field of sandwich structure design. The parametric studies are carried out and the numerical results reveal that factors such as the functionally graded configuration of the core, the facesheet-to-core thickness ratio, boundary conditions, and thermal environments critically influence both linear and nonlinear vibration characteristics of the DGM-engineered sandwich beams.