Intelligent design of mixture proportions of manufactured sand concrete from environmental, economical and mechanical perspectives

Manufactured sand (MS) is increasingly utilized in construction for its strong mechanical properties and low environmental impact. However, optimizing the mixtures of manufactured sand concrete (MSC) is complex due to the variability in MS properties and the need to balance multiple objectives, such as uniaxial compressive strength (UCS), CO2 emissions, and cost. This study proposes a multi-objective optimization (MOO) method based on machine learning (ML) and the non-dominated sorting genetic algorithm II (NSGA-II) to optimize MSC mixtures. The results indicate that the extremely randomized trees (ERT) model exhibits the best predictive performance for UCS, with an R value of 0.988 on the test set. The SHapley Additive exPlanations (SHAP) analysis identifies that the UCS is most sensitive to water-binder ratio (W/B), curing age and the maximum diameter of coarse aggregates. The developed MOO model effectively identifies the Pareto front, balancing cost, UCS, and CO2 emissions for MSC mixtures. By offering a systematic approach to optimizing MSC design, this framework enables cost-effective and sustainable concrete production, supporting the development of environmentally friendly construction practices.