A tree-based machine learning surrogate model for predicting off-axis tensile mechanical properties of 2.5D woven composites at high temperatures

Abstract

Textile composite structures in specific engineering applications can face safety concerns arising from exposure to high temperatures and off-axis loadings. High-fidelity finite element (FE) simulations and analytical models are both labor-intensive and time-consuming when predicting the mechanical behavior of textile composites under such loadings. To address this challenge, we develop a tree-based machine learning (ML) surrogate model for predicting the off-axis mechanical properties of warp-reinforced 2.5D woven composites in high temperature environments. To this setting, the tensile modulus and strength can be directly obtained based on the given temperature and off-axis angle, and the predicted results are in good agreement with FE simulations solutions. This study is expected to offer novel insights for the development of early warning systems that monitor abnormal temperatures and off-axis loadings in textile composite structures.