This research systematically investigated the transfer of railway-induced vibrations through soil by implementing the moving load method within finite element modelling (FEM), evaluating the efficacy of linear and non-linear material models, and assessing the impact of a varying water table. Utilising PLAXIS 3D, the advanced moving load method proved more effective than triangular pulse methods, as it realistically simulated continuous train wheel contact.
Linear elastic model (M1), despite providing patterns similar to site measurements, consistently overestimated acceleration magnitudes, particularly at greater distances from the rail. This indicated that linear models, even with Rayleigh damping, were inadequate for fully replicating complex real-life conditions due to inherent approximations and site uncertainties. Conversely, the implementation of non-linear material models, specifically the Hardening Soil model with small-strain stiffness (HSsmall), demonstrated improved agreement with field data. This finding underscores the importance of non-linear models for accurately representing soil behaviour and its intrinsic damping characteristics. Investigation into water table fluctuations using the linear-nonlinear model, M2 revealed no significant time-domain variations in acceleration magnitudes; however, a notable reduction in the intensity of higher frequencies was observed as the water table rose from 1.5 m below the ground level to 0.5 m below the ground.
The analysis of a fully non-linear model, M3 gave results closer to the site data but had variation due to the discrepancies in the input data with the site. It primarily modelled most components, except for the rails, railpads, fasteners, and sleepers, using the Hardening Soil model with small-strain stiffness (HSsmall). Additionally, in fully non-linear models (M3), external dynamic load adjustments (e.g., 5% of axle load) were found to be less critical, suggesting these models inherently capture dynamic effects. Furthermore, M3 predictions indicate that vibration amplitude increases as the water table rises, with very shallow water tables leading to much greater strains and plastic effects than linear models. Literature suggests this phenomenon involves complex interactions, often linked to changes in soil stiffness and potential for amplified vibrations near resonance.
The study acknowledges significant assumptions and simplifications within the models, such
as the exclusion of surrounding buildings and the assumption of isotropic soil, which contributed to discrepancies with real-world data. Despite the higher computational demands associated with non-linear models, their incorporation is recommended for achieving results
that more accurately reflect actual soil behaviour.