Prediction of properties of recycled aggregate concrete using machine learning models: A critical review

Abstract

Recycled aggregate concrete (RAC) not only alleviates the shortage of natural aggregates but also promotes the recycling of construction and demolition waste, contributing to sustainable development. Traditional empirical and statistical formulas have limited applicability and perform poorly when dealing with complex relationships. Therefore, machine learning has been widely used to predict the performance of RAC. In this paper, an overview is provided of the current research examining the impact of various parameters on the workability, mechanical properties and durability of RAC. The discussion and evaluation are focused on the applicability, accuracy, hyper-parameter selection, and input variables of machine learning algorithms that are employed to predict the performance of RAC. The advantages and disadvantages of different algorithms are compared, and the practical suggestions are given based on their performance. Additionally, the limitations of current research are discussed, and the perspectives for the prediction of RAC's performance based on the machine learning models are proposed.