<p>The freezing damage in ballasted railway tracks constitutes a widespread challenge in cold areas. There exists a critical need to develop suitable numerical simulation approaches to investigate the mechanical properties of track structure under low temperatures, particularly icing conditions. The discrete element method (DEM) has proven effective for characterizing the macro-meso mechanical behavior of granular ballast bed, while the bonded particle model (BPM) in DEM enables explicit modeling of the ice-bonded behavior between ballast particles. However, the relationship between the microscopic contact parameters in BPM and the macroscopic material properties has been confusing. To address this, an efficient parameter calibration strategy is imperative. Traditional trial-and-error calibration not only suffers from computational inefficiency but also introduces subjective bias. This study proposes a DEM calibration framework for frozen ballast aggregate based on the response surface methodology (RSM) with central composite design (CCD). Through uniaxial compression simulations implemented with a dilated polyhedron BPM, explicit second-order response surface functions were established between the bond micro-parameters and macro-mechanical properties (the compressive strength and elasticity modulus). Model validation against experimental data confirmed the predictive accuracy of the calibrated parameters. Finally, a DEM-FEM coupled model of the ballasted railway track was developed to evaluate structural responses in unfreezing and freezing conditions. The results show that the freezing behavior of granular ballast bed enhances the overall stiffness and load-bearing performance of the structure and helps to reduce the dynamic stress effect of the subgrade below.</p>