A review on properties and multi-objective performance predictions of concrete based on machine learning models

The increasing complexity and diversification of concrete raw materials have posed significant challenges in accurately predicting concrete properties, particularly under coupled external environmental conditions. In such situations, traditional empirical and statistical models are generally inadequate, and experimental approaches are hindered by heavy workloads, long durations, and high costs. In response, machine learning (ML) has emerged as a powerful tool for predicting concrete performance and optimizing mix proportions based on target outcomes. This review provides a comprehensive analysis of the current state of research on the application of ML models in predicting the mechanical properties, shrinkage and creep behavior, durability, and multiobjective performance of concrete. It also highlights the excellent predictive capabilities of ML models over traditional approaches, while addressing the critical importance of input feature selection, model optimization, and the careful application of specific algorithms. Furthermore, the review identifies key challenges (integrating particle packing models, considering cementitious mineral composition, and exploring concrete microstructure) with ML and proposes developing automated data tools and a standardized ML model database for improved prediction accuracy. By summarizing existing research and providing practical suggestions for future investigations, this review aims to guide the ongoing development of intelligent concrete design, addressing both current limitations and potential advancements in the field.