Mechanical Performance of 3D Printed Cellular Structures through Machine Learning Approach

Abstract

Cellular materials, essential in both engineering and natural systems, demonstrate dynamic mechanical behaviours. Material response to compressive forces is characterised by deformation and energy dissipation during impacts. The internal structural arrangement plays a fundamental role in establishing the material's mechanical performance. However, nonuniform arrangements complicate material selection and design processes. A Machine Learning (ML) approach is presented to optimise the geometric configuration of cellular structures for enhanced compressive mechanical performance. Various internal geometric patterns were simulated using finite element methods to construct an ML training dataset. Correlations between geometric patterns and mechanical performance metrics were established using a neural network. The findings indicate that the ML-based approach accurately predicts the mechanical responses, providing valuable insights for designing cellular structures with specific properties.