Multi-Objective Optimization of Self-Compacting Concrete Using Machine Learning and Life Cycle Assessment

This work introduces a data-driven framework for designing and optimizing sustainable fly ash (FA)-based self-compacting concrete (SCC). It addresses the need for jointly evaluating mechanical performance and environmental impact of concrete mixtures. While current methods usually optimize either compressive strength (CS) or carbon footprint separately, they often overlook the trade-offs between strength, emissions, and cost. To address this, a unified machine learning (ML) model is developed to predict CS, global warming potential (GWP), and environmental cost indicator (ECI) simultaneously from mix design parameters. The model shows high predictive accuracy on unseen data for CS, GWP, and ECI. Using this model, an Environmental–Mechanical–Economic ( EME ) index is defined as an optimization tool to achieve high CS while minimizing GWP and ECI, depending on application-specific requirements. A ML-guided genetic algorithm (GA) is then used to generate optimized concrete mixes within realistic constraints and application-specific strength requirements that outperform the ones in the experimental dataset by achieving higher strength while lowering GWP and ECI. For example, in low-strength applications (concrete fill), optimized mixes increased CS from 12.2 MPa to 13.6 MPa, while reducing GWP by 13% and ECI by 2%. Overall, the proposed work provides a predictive, multi-objective approach to concrete mix design, supporting the selection of high-performance mixes with reduced environmental impact. • A unified ML model predicts strength, carbon footprint, and cost of fly-ash-based SCC • An Environmental–Mechanical–Economic Index supports application-specific mix selection • ML-guided optimization generates mixes with improved strength–environment trade-offs • Higher strength achieved alongside reduced cost and environmental impact