Time-dependent rheology of seawater-mixed cement paste: Role of binder characteristics revealed by experiments and machine learning

This study investigates how the characteristics of fly ash (FA), ground granulated blast-furnace slag (GGBS), and silica fume (SF) affect the time-dependent rheology of seawater-mixed cement pastes. Single-, binary-, and ternary-blended binders were prepared at a fixed water-to-binder ratio of 0.40. Fluidity, dynamic yield stress, and plastic viscosity were measured at 5, 35, and 65 min, together with water film thickness (WFT), wet packing density, and early hydration heat. FA and GGBS generally reduced dynamic yield stress, whereas SF increased both dynamic yield stress and plastic viscosity. Increasing the SF content to 6% reduced the 1 h fluidity loss from approximately 23% to 5%, but increased the growth rate of dynamic yield stress to approximately 83%, indicating a trade-off between flow retention and the time-dependent increase in resistance to flow initiation. Compared with values estimated by linear superposition, blended systems generally exhibited higher fluidity and lower plastic viscosity, although the response of dynamic yield stress depended on the binder combination. The relationships between WFT and rheological properties varied with binder type and testing age, indicating that WFT alone could not describe the rheological evolution of all investigated systems. Machine-learning models using physical, chemical, hydration-related, and time-dependent descriptors were further employed to quantify the relationships between binder characteristics and rheological responses within the investigated dataset. SHapley Additive exPlanations (SHAP) analysis was then used to examine dataset-specific feature–response associations. WFT and particle-size-related descriptors showed strong associations with fluidity and plastic viscosity, whereas Al₂O₃ content and the CaO/SiO₂ ratio exhibited relatively high importance for dynamic yield stress. These findings are restricted to the investigated seawater-mixed paste systems at a water-to-binder ratio of 0.40, and further validation is required before extending them to other binder systems, mixing-water conditions, or material scales.