A mix design–supporting multi-objective optimization framework for ultra-high-performance concrete balancing mechanical strength, workability, cost efficiency, and carbon emission criteria

Ultra-high-performance concrete (UHPC) exhibits exceptional mechanical strength and durability; however, its application is often constrained by high binder demand, leading to increased material costs and carbon footprint. To address these competing objectives in a systematic and automated manner, this study proposes a novel and integrated mix design–supporting multi-objective optimization framework that, for the first time, systematically combines interpretable machine learning (SINDy), life cycle assessment (LCA), and the NSGA-II algorithm within a unified workflow. A comprehensive database comprising approximately 900 UHPC mix proportions compiled from over 80 published studies, exceeding the scale of most prior works, was used to develop and validate predictive models covering compressive strength, flexural strength, and workability. The Sparse Identification of Nonlinear Dynamics (SINDy) algorithm was employed to derive explicit and interpretable mathematical relationships between mixture constituents and mechanical performance indices, overcoming the limitations of conventional black-box models. In addition, a material cost function based on local Vietnamese price data and a carbon footprint model developed using LCA principles were formulated and integrated into the optimization framework. The Non-dominated Sorting Genetic Algorithm II (NSGA-II) was then applied to identify Pareto-optimal UHPC mixtures under practical performance constraints. The results demonstrate that the proposed framework enables a 30–40% reduction in carbon emissions and a 20–25% reduction in material cost while maintaining required mechanical performance and workability, thereby supporting the development of low-carbon and cost-efficient UHPC design.