Dynamic characteristics of a clayey soil under different moisture-temperature cycling conditions

This study investigates the pore structure, water-retention capacity, and dynamic characteristics (including the resilient modulus M_R, accumulated plastic strain ε_p, and damping ratio λ) of a clayey pavement subgrade soil that has been subjected to different moisture-temperature cycling conditions. Three distinct moisture-temperature cycling conditions are designed by combining freeze-thaw and wetting-drying processes in different sequences to simulate complex environmental actions. Experimental results reveal that (i) when the impacts of the moisture and temperature have reached equilibrium, the soil exhibits comparable pore-structural, water-retention, and dynamic characteristics across different moisture-temperature cycling conditions; (ii) moisture-temperature cycling induces significant pore structure evolutions, which are characterized by micropore contraction and macropore development. These morphological changes directly affect the soil-water characteristic curve (SWCC), manifesting as reductions in the air-entry value and decreases in the slope of the SWCC in the transition zone slope; (iii) freeze-thaw cycles increase the moisture sensitivity of the ε_p but impose minor impacts on that of the M_R and λ. Under unsaturated conditions, the ε_p and M_R constitute consistent relationships with suction regardless of freeze-thaw histories; (iv) there are linear relationships between the ε_p, M_R, and λ during cyclic loading, which show strong dependence on the humidity and freeze-thaw conditions. For specimens under different external stress, humidity, and freeze-thaw conditions, the relationships between their ε_p and M_R at the end of long-term cyclic loadings show consistent nonlinear relationships, which can be well described by a simple model. Ten machine-learning approaches were employed to predict the ε_p from the soil's stress state and moisture-temperature conditions. The Bayesian Neural Networks were found to be most capable in terms of computational precision and efficiency.