Machine learning-enhanced analysis of triply periodic minimal surface structures integrating topological features and hypergravity effects

Existing constitutive models for triply periodic minimal surface (TPMS) structures often fall short of accurately capturing the influences of critical parameters such as topology, cell size, and mid-surface thickness on their mechanical behavior. These models also frequently overlook the effects of hypergravity, which are crucial for aerospace applications. In this study, using finite element analysis, we develop a comprehensive dataset that encompasses a wide range of TPMS configurations and parameters, and we validate it with 56 experimental data points, achieving a maximum deviation of only 5%. We employ six optimized machine learning models to predict the elastic stiffness matrices, with the extreme gradient boosting and gradient boosting regression models demonstrating high precision. Feature importance analysis identifies the mid-surface thickness and cell-size ratio as the most influential factors. Notably, this work is the first to unify the mechanical characterization of TPMS structures across normal and hypergravity conditions using machine learning techniques.