Optimizing Self-Compacting Concrete (SCC) Mix for Different Grades Using AI (ANN/RF models) for Low- Carbon Construction

Abstract

The construction industry is under increasing pressure to reduce carbon emissions while maintaining the performance requirements of modern infrastructure. Self-Compacting Concrete (SCC) has gained widespread acceptance due to its superior workability, self-compaction ability, and enhanced durability; however, its high cementitious material content often results in a significant carbon footprint. This study presents an Artificial Intelligence (AI)-based framework for optimizing SCC mix designs across different concrete grades (M20, M30, M40, M50, and M60) to achieve low-carbon and sustainable construction. A dataset comprising 120 SCC mix samples with varying proportions of Ordinary Portland Cement (OPC), Ground Granulated Blast Furnace Slag (GGBS), Fly Ash, Limestone Calcined Clay Cement (LC3), and water-binder ratios was developed and utilized for model training and validation. Artificial Neural Network (ANN) and Random Forest (RF) models were employed to predict key SCC properties, including compressive strength, slump flow, viscosity, and carbon emissions. The predictive models demonstrated high accuracy, with RF outperforming ANN in terms of prediction performance. Subsequently, a Genetic Algorithm (GA) was integrated to identify optimal mix proportions that minimize embodied carbon while satisfying strength and workability requirements. The optimized SCC mixtures achieved carbon emission reductions ranging from 21% to 29% across different concrete grades without compromising mechanical performance or SCC workability criteria. The results demonstrate that the proposed AI-driven optimization framework provides an effective and reliable approach for designing sustainable SCC mixes and offers significant potential for advancing low-carbon construction practices in the concrete industry.