Probabilistic seismic performance analysis of railway embankments on spatially variable loose deposit slopes

A systematic assessment of the influence of soil spatial variability parameters on railway embankment seismic performance under near-fault pulse-like ground motions is essential. This study employs a three-dimensional random finite difference method to investigate the effects of spatially variable loose deposits on seismic failure mechanisms and the probabilistic characteristics of peak permanent settlement (PPS) at railway embankment crests. The results demonstrate that loose deposit spatial variability directly governs potential slip surface evolution through strength heterogeneity, inducing two distinct seismic failure mechanisms: shallow and deep sliding along continuous weak zones. Probabilistic seismic analysis reveals that spatially variable loose deposits significantly increase both the geometric mean and standard deviation of PPS compared to deterministic results. Specifically, the probabilistic characteristics of PPS are more sensitive to the coefficient of variation than to the correlation distance, both of which are key parameters for random fields. Finally, this study develops the amplification factor method (AFM) and parameter reduction method (PRM), providing statistically based dynamic response amplification factors and internal friction angle reduction factors at the 95% confidence interval to simplify the incorporation of spatial variability effects. The findings advance reliable seismic design methodologies for railway embankments on spatially variable loose deposits by quantifying the impact of spatial uncertainty and providing practical simplification frameworks.