The presence of frozen soil layers leads to stratification in soil stiffness, thereby influencing the dynamic response of pile foundations in seasonally frozen soil regions. This study investigated the dynamic response of pile-soil interaction (PSI) systems in such regions. A reduced-scale (1/10) model of a pile group with an elevated cap in railway bridges was subjected to shake-table testing. During these tests, measurements were taken of soil and pile accelerations, displacement time histories, and pile strain. The acceleration amplification factor (AMF) and response spectrum of the soil and pile foundation were analyzed based on these data. Additionally, the pile-soil interaction and the dynamic shear stress-strain relationship of the soil were investigated. The experiment indicated that the presence of a frozen soil layer alters the energy dissipation order of the pile-soil interaction system. This leads to a weakened dynamic response of the pile foundation. Furthermore, the seasonally frozen soil layer acts as a filter for high-frequency ground motion, thereby mitigating resonance between ground motion and the pile foundation, ensuring the protection of the pile foundation. However, the significant stiffness contrast induced by the seasonally frozen soil can pose a threat to structural safety under increasing peak ground acceleration (PGA). As PGA increases, there is a transition from linear to nonlinear interaction between the pile and soil, initially affecting the unfrozen soil layer, then the frozen-unfrozen transition layer, and ultimately impacting the seasonally frozen soil layer.

The dynamic interaction between a civil infrastructure and the soil beneath is crucial for seismic risk assessment. Due to the increased computational capacity, more frequently, this problem is starting to be addressed by 3D modelling based on the finite element method (FEM). However, because of the interaction between the stresses and strains in the orthogonal directions of the soil volume, i.e. the Poisson effect, it is not trivial to achieve specific spectral ordinates at the surface of the FEM model, after propagation from the bedrock. This study introduces a novel method aimed at obtaining surface-level ground motions with specific spectral intensities by using 3D FEM models. This method integrates spectral matching, filters, deconvolution using 1D models, and frequency modulation techniques, to address misalignments between the outcomes of 1D and 3D models, particularly focusing on high-frequency spectral amplification in soil response. It has been tested by analysing two seismic scenarios, which have been characterized from a probabilistic perspective. The proposed approach ensures the development of ground motion records accurately producing specific spectral intensities at the surface, enhancing seismic risk assessments and structural analysis. The study emphasizes the importance of accurate seismic hazard characterization, providing valuable insights for earthquake engineering practices.