Data-driven approaches in concrete science: applications, challenges and future prospects

This review paper provides a comprehensive exploration of integrating data-driven approaches in the domain of concrete science. The paper commences with an introduction elucidating the background and context of data-driven concrete science, outlining objectives and scope, and underscoring the importance of data-driven methodologies. Subsequently, it delves into the traditional analytical approaches and the potential for data-driven methods. The paper elucidates data collection and pre-processing techniques tailored to the domain, encompassing concrete-related data types, collection methodologies, and data pre-processing strategies. Moreover, it extensively covers data-driven modelling and prediction in concrete science, presenting an overview of data-driven models, machine learning techniques deep learning approaches and integration of big data analytics. The review consolidates insights into diverse applications, including concrete strength prediction, durability analysis and concrete microstructure characterisation, employing data-driven approaches. Furthermore, it highlights challenges and opportunities in this burgeoning field, encompassing data quality and availability, interpretability and explainability of models, and ethical consideration. The paper concludes with recommendations for researchers and practitioners aiming to harness the full potential of data-driven methodologies.