Architecture Prediction of 3D Composites Using Machine Learning and No‐Destructive Technique

Abstract

AbstractThe field of composites has seen a surge in the adoption of machine learning techniques due to their ability to achieve once unattainable goals. Presently, machine learning research in composites primarily centers around predicting composite properties or optimizing microstructures to attain specific properties. This paper presents a data‐driven approach to predict the complete architecture of composites. A multi‐output machine learning model, based on conventional XGBoost algorithms, is developed to comprehend the intricate correlation between composite architecture and elastic wave propagation in them. The machine learning model uses input elastic wave signals collected at one face of the composite cube, induced by an actuator on the opposite face of the cube, as features. The composition labels are 3D matrices that represent the architectures of the composite cubes. The results show that the architecture of composites can be predicted using a short period of elastic wave travel through the composites, with up to 96% accuracy. This method can be readily adapted and implemented for any industry application requiring the determination of the architecture of unknown composites without destruction.