Multi-Objective Optimization of Low-Carbon Concrete Mixtures Using Machine Learning and Genetic Algorithms: An Analytical Review

Multi-Objective Optimization of Low-Carbon Concrete Mixtures Using Machine Learning and Genetic Algorithms: An Analytical Review

baharsalari1

1) Bachelor of Science (B.Sc.) in Civil Engineering

IOI Code : ABEB-ACZFKE
Publication : 4th International Conference on Recent Advances in Engineering, Innovation and Technology - Belgium(eitconf.com)
Abstract :
    Concrete is the most widely used construction material in the world, and cement production accounts for roughly 8 to 10 percent of global CO₂ emissions. Reducing the embodied carbon of concrete without sacrificing strength and durability is a nonlinear, multi-objective problem for which trial-and-error methods are inefficient. Using an analytical review approach, this paper compares seven studies published in 2026 that employ machine learning (ML), genetic algorithms (GA), and life cycle assessment (LCA) for low-carbon concrete mix design. The studies reviewed cover fly-ash-based self-compacting concrete, recycled aggregate concrete, ultra-high-performance concrete (UHPC), 3D-printed recycled concrete, and one systematic review. The findings show that the dominant pattern is an ML surrogate model + multi-objective evolutionary optimization (particularly NSGA-II) + multi-criteria decision-making ; cement is the most important driver of embodied carbon in all studies; and strength-prediction accuracy ranges from R² = 0.33 in heterogeneous multi-source datasets to 0.997 in homogeneous datasets. Reported carbon reductions (ranging from 13 to 40 percent) depend on the comparison baseline and the application type. Cross-study (external) validation, uncertainty quantification, and restricting solutions to the data-valid region were identified as key requirements for practical deployment. Finally, a step-by-step workflow for low-carbon concrete mix design is proposed. **Keywords:** low-carbon concrete; multi-objective optimization; machine learning; genetic algorithm; NSGA-II; life cycle assessment; embodied carbon.

Keywords : 1 low carbon concrete 2 multi objective optimization 3 machine learning 4 genetic algorithm 5 life cycle assessment 6 embodied carbon