Abstract: Railway subgrade frost heave poses a significant threat to the long-term operational safety and stability of railways in seasonal frozen ground regions. To address the insufficient understanding of the evolution patterns of internal multi-physical fields in seasonally frozen railway subgrades, a monitoring section was established at a low embankment in the Jiling to Daqingyangkou section of the Lanxin Railway for continuous ground temperature and moisture content monitoring. Based on the real-time monitoring data, a direct hydro-thermal coupling numerical model of the subgrade cross-section was developed on the COMSOL Multiphysics platform. This model incorporated ice-water phase change, liquid water migration, and the blocking effect of pore ice, and its reliability was validated. Combining this with the "hydrodynamic frost heave model," the spatiotemporal evolution of frost heave behavior was then quantitatively analyzed. The results indicate that the frost heave amount in this section reaches its peak in late February, corresponding to a maximum freezing depth of approximately 1.37 m, with the shoulder area being more sensitive to freeze-thaw cycles. The synergistic application of a polyurethane insulation layer, a two-layer graded gravel cushion, and an Impermeable Geocomposite can form a multi-layer collaborative protection structure that provides both thermal resistance and moisture blockage. This configuration reduces the peak frost heave at the slope shoulder by 74.2% and decreases the maximum freezing depth to 0.46 m. This study provides theoretical support and a basis for engineering design concerning frost heave mitigation in railway subgrades in seasonally frozen regions.