Interpretative Modeling of UHPC for Strength and Eco-Efficiency Assessment

The growing demand for sustainable construction has intensified the need for advanced materials and robust assessment frameworks that integrate mechanical performance with environmental efficiency. Ultrahigh performance concrete (UHPC) stands out for its exceptional strength and durability, but presents environmental challenges due to its high cement content. This study introduces an integrated framework to evaluate the mechanical and environmental performance of UHPC, utilizing predictive models and interpretative analyses to quantify compressive strength (fc), embodied CO2 index (CI), and embodied energy index (EI). Drawing from a database of 918 UHPC mixtures and validated with seven nonproprietary compositions, the framework leverages multiple linear regression (MLR) and artificial neural networks (ANNs). To interpret variable interactions and nonlinear behaviors, Pearson correlation and SHapley Additive exPlanations (SHAP) are employed, offering insights into mix adjustments. Findings suggest that cement dosages between 600 and 900 kg/m3 optimize the balance between strength and environmental impact. SHAP analysis underscores the nonlinear contributions of supplementary cementitious materials (SCMs), including silica fume, limestone powder, quartz powder, slag, fly ash, and nanosilica, to both strength and sustainability. Moreover, optimized fiber content and curing temperatures of 80°C-90°C are shown to enhance eco-efficiency. The proposed framework enhances predictive accuracy while supporting sustainable UHPC design through data-driven insights into environmental efficiency, aligning material selection with circular economy principles and carbon reduction goals.