Mechanical properties of 3D printing metal cells: A combined homogenization and machine learning study

Abstract

A key challenge in additive manufacturing is the computational design of porous materials to optimize the mechanical performance of alloy-based components. We present an algorithmic framework for generating unit cells with truss-, plate-, shell-, tube-, and TPMS-based geometries, along with a method for calculating their mechanical properties. These properties were determined through numerical homogenization using the finite element method and validated against experimental measurements for lattice structures. Leveraging this dataset, we trained a convolutional neural network to predict stress–strain curves with high accuracy, achieving a mean absolute percentage error of less than 13%. Our approach established a robust pipeline bridging computational design and experimental realization for 3D-printed porous metamaterials.