Hydro-Mechanical Behavior of Sand Subgrades under Cyclic Loading across Moisture States: Insights from Large-Scale Physical Modeling

Flooding, elevated groundwater, and seasonal moisture changes can substantially reduce subgrade stiffness and accelerate rutting by altering suction, saturation, and pore-pressure response. Although moisture-sensitive models and mechanistic–empirical tools exist, large-scale experimental datasets that simultaneously quantify stress transmission, deformation, stiffness evolution, and pore-pressure/suction behavior under controlled hydro-mechanical states remain limited. This study presents large-scale physical modeling results for Ottawa F-75 sand tested under dry, fully saturated, compacted-moist (high suction), and water-table-controlled unsaturated conditions. A rigid circular plate applied 105 force-controlled cycles using a resilient-modulus-style half-sine pulse program consistent with common repeated-load characterization practice, enabling comparison of early-cycle conditioning and stabilization trends across moisture states. Instrumentation captured subsurface stress attenuation, surface deformation profiles, and pore-water pressure where applicable. Results show that increasing suction (partial saturation) reduces transmitted stress amplitude, accelerates stabilization of permanent deformation rate, and increases resilient stiffness, while fully saturated conditions produce the largest cumulative settlements and the strongest pore-pressure response. Quantitative stabilization metrics demonstrate markedly stronger stabilization in unsaturated states than in that of saturated soils. In fully saturated tests, the measured pore-pressure-to-total-stress ratio indicates transient liquefaction/cyclic mobility near the load center, whereas lowering the water table suppresses pore-pressure buildup and eliminates liquefaction tendency. The dataset provides large-scale benchmarks to support moisture-sensitive pavement evaluation and to calibrate suction-aware constitutive and numerical models.