Accurate prediction of soil compaction parameters (Optimum Moisture Content and Maximum Dry Unit Weight) is essential for earthwork infrastructure. While Artificial Neural Network (ANN) models offer high predictive accuracy, their heavily overparameterized, "black-box" architecture hinders practical deployment in geotechnical engineering, where transparent and explicit formulations are strongly preferred. This study proposes a Kolmogorov-Arnold network (KAN) framework to resolve this structural inefficiency. By integrating dynamic B-splines and L1 sparsity, the KAN maps nonlinear, multi-phase soil behavior with greater mathematical parsimony. To ensure field applicability, Symbolic Regression (PySR) was subsequently employed to distill the KAN's optimized topology into explicit algebraic equations. Results show that the KAN model achieves comparable hold-out accuracy to the baseline ANN model while requiring 7.4-fold fewer trainable parameters (1,248 versus 9282) and maintaining phase-compliant volumetric boundaries (Sr ≤ 1.0). Benchmarking against Random Forest and XGBoost under an identical protocol showed that both mainstream ensembles matched or exceeded in-distribution accuracy, but generalized less reliably when applied to a geographically distinct Nepal database (n = 40); the KAN and PySR models instead retained their accuracy on this independent set, with the PySR equations achieving R2 = 0.78 for both targets, a20 = 100
The structural integrity of highway tunnels situated beneath railroad tracks is a critical engineering challenge. High-speed trains exert significant dynamic loads, leading to vibrations, stress concentrations, and dynamic amplification that can compromise safety. Despite its importance, a notable research gap remains in parametric investigations of the modal and dynamic characteristics of these specialised tunnel configurations. This study aims to assess the structural response of an existing railway-track/ground structure and the highway tunnel beneath it to imposed moving axle loads representative of high-speed trains. The goal is to measure the tunnel system's deformation, stress, velocity, and acceleration responses under various railway loading conditions and train speeds. A fully coupled vehicle/track/ground/tunnel interaction model, including explicit vehicle suspension and wheel/rail contact interaction, has not been developed. The three-dimensional transient finite element model of the railway track, surrounding ground, and highway tunnel was built in ANSYS. The prescribed axle forces are applied to the train, which travels along the track at a constant speed and maintains the same axle magnitudes and longitudinal spacing as the load configurations selected from Indian Railways. This results in a moving-load model coupled to the track–ground–tunnel structural system, rather than a complete vehicle–track–ground–tunnel interaction model. The results indicate that damping, surrounding rock grade, train speed, and loading condition are important parameters to consider for the dynamic response of the track–ground–tunnel system. All reported equivalent (von Mises) maximum stresses are considered elastic response demand indicators for the assembly and are not assigned to the tunnel lining itself without component-level verification. They are not considered stresses acceptable in concrete, since a linear-elastic (LE) constitutive law is used for the tunnel lining and no cracking, crushing, fracture energy, or damage evolution are considered. The present study thus focuses on deformations, velocities, accelerations, and elastic stress demands for the given moving axle loads, with structural safety against concrete cracking or crushing considered only through a dedicated non-linear concrete analysis based on principal tensile/compressive stresses and corresponding strength and damage parameters. The existing research is limited to analysing tunnel structures. Tunnel and railway engineers can use this study to further assess fatigue in highway tunnels under various dynamic loads from railway traffic. From a geotechnical engineering standpoint, the results highlight the importance of layered-ground stiffness and strength, stress confinement, damping, groundwater conditions, and ground–tunnel load transfer in controlling the dynamic response.
To address the insufficient systematic modeling capabilities for high-precision numerical characterization of train dynamic loads and multi-field coupling effects, this study developed a custom subroutine for train dynamic loads using Fortran linked with ABAQUS simulation software. This subroutine simulates the dynamic load conditions of a train under the influence of various factors. Based on unsaturated soil seepage theory and train vibration load theory, a finite element (FEM) numerical simulation analysis method for seepage-stress-dynamic load coupling was proposed. This model provides a quantitative analytical method and data support for elucidating the mechanical response patterns of embankment slopes under the coupled effects of rainwater infiltration and dynamic train loads. In addition, the results show that: During rainfall, after the soil becomes saturated, water diffuses laterally. Restricted by the drainage conditions on the slope surface, a pore-water pressure distribution characterized by high values in the middle and low values on both sides forms inside the slope, which further drives the groundwater table to shift toward the slope toe; the stress at the foot of the slope undergoes a three-phase evolution-exponential growth at the initial stage, linear increase in the middle stage, and decrease in the later stage; the train load further increases the extent of the plastic zone; the slope safety coefficient of the embankment was calculated by the strength reduction method, and it was found that the variation rule of the slope safety coefficient was: before rainfall (1.363) > after rainfall (1.025) > rainfall-dynamic load (1.01). This study provides scientific and technological support for disaster mitigation and prevention.
Deep excavation inevitably impacts adjacent structures. Studying the variation patterns under different parameters during this process allows for targeted measures to reduce the disturbance. With this objective in mind, and based on an underground station of the Beijing Subway, a MIDAS-GTS finite element model was first established according to an engineering case to analyze the deformation characteristics of the deep excavation process under the dual support system of “retaining piles + steel struts.” Subsequently, the influence of different parameters on adjacent pile foundations was analyzed using the controlled variable method. The results show that after excavation, the ground surface settlement exhibits a trough-shaped distribution, with a maximum value of 28.22 mm, exceeding the 15 mm limit specified in the code. The existing pile foundation exhibits a maximum settlement of 2.42 mm and a maximum horizontal displacement of 3.42 mm. The parametric analysis showed that the retaining pile diameter had the most significant influence on deformation control. When the retaining pile diameter was increased from 400 mm to 1200 mm, the maximum ground surface settlement, pile foundation settlement, and horizontal displacement decreased by 78.6
