Deep learning method for determining the surface elastic moduli of microstructured solids

Abstract

Surface microstructures greatly affect many physical and mechanical properties of solid materials. Surface constitutive relations are often required to analyze the mechanical responses of materials and structures at the micro scales, but it remains a challenge to determine the surface constitutive parameters of microstructured materials. In this paper, a machine learning-based approach is proposed to predict the surface elastic properties of materials with complex microstructures on their surfaces. Using a data set generated from the finite element method, we demonstrate that the trained deep neural network can efficiently provide an accurate mapping between the equivalent surface elastic properties (e.g., elastic moduli and Poisson’s ratios) and the mechanical and geometric features of surface microstructures. This study provides an efficient way to evaluate the surface properties of materials and promises for wide applications in the analysis of micro-/nanostructured materials and devices.