Predicting ground vibrations induced by high-speed trains on the inhomogeneous soil

Abstract

Ground vibrations generated by high-speed train operations have raised significant environmental concerns. Conventional methods for predicting train-induced ground vibrations typically assume that soil layers are homogeneous. However, soil layers undergo various geological processes that result in significant variability in soil properties over short spatial extents. This paper therefore proposes an efficient method for predicting ground vibrations from high-speed trains while accounting for spatial variability in soil properties. A semi-analytical deterministic model is introduced to simulate the coupled train-track-ground system. Dynamic responses of stratified soils are calculated using the thin layer method (TLM) – perfectly matched layer (PML). The TLM is employed to simulate finite-depth soil layers, while the PML is applied to account for wave attenuation in the underlying half-space. Train-induced ground vibrations in the frequency-spatial domain are calculated by analytically solving the inverse Fourier transforms with respect to the wavenumbers, leading to a high computational efficiency. The model’s accuracy is validated through comparison with existing methods. Vertical variability in soil properties is simulated using the TLM in conjunction with Karhunen-Loève (K-L) expansion. Numerical results indicate that local soil inhomogeneity can significantly impact the prediction of train-induced ground vibrations, resulting in changes in vibration levels of approximately 30 dB. As soil depth and degree of variation increase, the uncertainty in vibration predictions also increases. Therefore, it is crucial to consider soil inhomogeneity when predicting ground vibrations induced by high-speed trains.