Prediction and optimization model of sustainable concrete properties using machine learning, deep learning and swarm intelligence: A review

Among the many sustainability challenges in the construction industry, those related to the application of concrete and its components are the most critical. Particularly, the production of cement significant threats to pollution, waste, biodiversity, and human health. Sustainable concrete formed by construction & demolition waste (CDW) instead of natural materials has good low-carbon potential, which has been considered as an effective solution to solve the above problems. With the incorporation of new solid waste (supplementary cementitious materials, recycled aggregates and geopolymers, etc.), the design complexity of such concrete becomes increasingly challenging, but the traditional model based on linear regression and assumptions are insufficient to evaluate the performance of the multi-level material system. However, the complex interaction mechanism of these multiple factors is particularly suitable for artificial intelligence (AI) due to its strong nonlinear processing capability. This paper systematically reviewed the research on using AI technology for evaluating sustainable concrete, focusing on mixture ratio, static performance and durability. The results shows that the establishment of a database covering factors such as material composition and curing conditions is crucial to ensure the universality of the prediction model. Machine learning (ML) models considering the influence of multicollinearity can realize the multi objective optimization of concrete mixture and predict the performance of sustainable concrete. On this basis, feature importance analysis can reveal the influence of different input variables on evaluation indicators and overcome the black box problem of the model. Furthermore, this paper discusses the limitations of existing research in detail and proposed feasible schemes in terms of algorithmic mechanisms and performance assessment of sustainable concrete.