Sustainability Trade-Offs in Machine Learning–Optimized Concrete Mixes
摘要
This study evaluates the environmental performance of optimized and non- optimized concrete mixes. The optimized mix was reconstructed using constraints and objective functions of a published machine learning optimization model. A cradle-to- gate life cycle assessment, conducted using the ReCiPe Endpoint method (H), was employed to quantify trade-offs between concrete performance and environmental impacts. Results show that while the optimized mix achieves a comparable compressive strength of about 50 MPa, it exhibits a higher binder intensity (8.50 vs. 7.85 kg/m3·MPa). Despite this, it achieves a 10% reduction in embodied energy due to increased use of supplementary cementitious materials. The optimized mix also demonstrates lower impacts in global warming and resource scarcity categories. However, it incurs higher burdens in several other areas, including toxicity, ecotoxicity, and water consumption. These findings highlight the trade-offs inherent in performance-driven optimization and emphasize the need to integrate environmental impact metrics into concrete mix design to support genuinely sustainable construction practices.