Machine learning-enabled pathways for low-carbon concrete

Machine learning (ML) can support the development of lower-carbon concrete through improved cement production, mix design, supplementary cementitious materials and alternative binders, and recycled aggregate concrete. This review synthesizes these applications and shows that ML can strengthen process control and multi-objective optimization. Current limitations arise mainly from data heterogeneity, limited generalizability, and weak interpretability. Closer integration with physical understanding is needed for reliable low-carbon decision-making.