Exploring the Elastic Properties of Interfacial Transition Zone in Concrete Materials Using an Ensemble Learning Approach

ABSTRACT Concrete materials consist of multiple phases with distinct mechanical properties, making it essential to accurately identify the mechanical behavior of both constituent phases and their interfaces for effective multiscale modeling. This study estimates the elastic properties of the interfacial transition zone using a machine learning (ML) approach. A dataset is generated from numerical simulations based on a Fast Fourier Transform method, validated against experimental data. Seven ML models are developed and trained, including four independent artificial neural networks and three ensemble models. The best‐performing ensemble model is identified and described in detail. Further analysis, including over‐fitting analysis, parameter investigation, and sensitivity study, confirms the model's validity and practical applicability.