Developing concrete mix designs that enhance structural performance while reducing environmental impact is vital for achieving sustainable construction. This study utilizes multiobjective optimization and machine learning (ML) that aim to reduce the carbon footprint of concrete production by utilizing industrial by-products (e.g., slag and fly ash) as replacements for cement without compromising concrete strength. A data set of 1,105 observations detailing mix compositions, ages, and compressive strengths was utilized to train and test various ML models, including neural networks (NN). The NN model, optimized using grid search, random search, and Bayesian optimization, demonstrated superior performance with an R2 value of 0.89 for compressive strength prediction. The model was further validated using material compositions and compressive strength tests not included in the training or verification data sets. A key innovation of this research lies in applying dimensionality reduction and multiobjective optimization techniques to navigate the trade-off between compressive strength and environmental impact. The Pareto front was generated, highlighting optimal concrete mix designs that achieve compressive strength requirements while reducing carbon emissions by 30%–70%. This design space, derived through ML, enables engineers to make informed decisions about designing sustainable, high-performing concrete mixtures. The findings underscore the drastic potential of ML in advancing sustainable construction practices, achieving structural quality, and mitigating the environmental footprint of building materials.
Hammoud et al. (Thu,) studied this question.