Optimizing Concrete Mix Design for Cost and Carbon Reduction Using Machine Learning

Cement is the main component of concrete and one of the most significant contributors to carbon emissions. Reducing cement use can significantly reduce global carbon emissions. This study aims to create an optimal concrete mixture of cost and minimal carbon emissions, but the compressive strength meets the requirements. XGBoost Machine Learning Algorithm is used to make predictions, and PSO is used to obtain the optimal mixture. The novelty of this study is the presence of concrete age variables, determination of PSO parameter weights using stakeholder preference analysis of construction in Indonesia with the AHP method, and validation of the PSO-recommended mixture using laboratory tests, which is still rarely done. The research findings indicate that the ML model provides satisfactory prediction values with an R2 value of 0.9043, root mean square error of 48.5147 and mean absolute percentage error of 0.0484. PSO results show that cement reduction in concrete can be achieved with optimal use of admixture while reducing 1-3% costs and 7-10% carbon emissions. The research findings provide critical insights into the importance of using innovative techniques to optimize sustainable concrete mixes, accelerating the market implementation of products with cost benefits.