Intelligent Decision-Making in Sustainable Manufacturing: Coupling Machine Learning with Multi-Objective Optimization for Carbon Reduction

Abstract

Smart manufacturing systems increasingly face the dual challenge of maximizing productivity while minimizing environmental impacts. However, existing approaches often treat emission forecasting and operational optimization as isolated tasks. This study proposes an integrated intelligent decision-making framework that couples machine learning with multi-objective optimization (MOO) to achieve eco-efficient production scheduling. In the predictive phase, Support Vector Regression (SVR) is hybridized with four advanced metaheuristic algorithms: Hunger Games Search (HGS), Slime Mould Algorithm (SMA), Chaos Game Optimizer (CGO), and Grey Wolf Optimizer (GWO) to forecast production volume and carbon emissions. When evaluated on a representative industrial case study, the SVR–CGO model demonstrated superior predictive accuracy, achieving R² values of 0.980 and 0.989 for production and emission targets, respectively. In the optimization phase, the best-performing SVR model serves as the fitness function for MOO, with the NSGA-II and SPEA2 algorithms addressing the inherent trade-offs between productivity and carbon reduction. Finally, a TOPSIS-based multi-criteria decision-making approach is applied to evaluate distinct operational strategies, including high-production, energy-saving, and green-energy modes. Results reveal that the proposed expert system can effectively identify Pareto-optimal configurations, enabling up to a 17% reduction in CO₂ emissions while sustaining over 95% of maximum production capacity. The primary novelty of this research lies in unifying carbon footprint modeling, metaheuristic-driven prediction, and evolutionary optimization into a cohesive decision-support tool that provides manufacturing stakeholders with actionable insights for sustainable operations.