A two-dimensional finite element model was developed to evaluate the deformation and stability of geosynthetic-reinforced Mechanically Stabilized Earth (MSE) walls under different flooding conditions. Seepage analysis was performed to obtain pore-water pressure distributions, while stress–deformation and strength reduction analyses were conducted to evaluate wall response and stability. Flood height and differential hydraulic head were varied to quantify their effects on lateral deformation and factor of safety. A parametric analysis was performed by varying backfill friction angle, unit weight, elastic modulus, reinforcement stiffness, reinforcement length ratio, and vertical reinforcement spacing. Balanced flooding produced limited changes in wall response. Flooding with a differential hydraulic head resulted in a marked increase in lateral deformation and a reduction in the factor of safety. Among the investigated flooding conditions, the 0.75 H flooding condition with a differential hydraulic head represented the most critical case, exhibiting the greatest lateral deformation and the lowest factor of safety. Reinforcement length, stiffness, and spacing strongly affected deformation, whereas friction angle and reinforcement length primarily affected stability. Differential hydraulic head governed the adverse response under flooding. Increased reinforcement length and stiffness, reduced vertical spacing, and improved backfill strength enhanced wall serviceability and stability.
This study investigated the dynamic response of tunnel support structures subjected to cyclic blasting in the Qingdao Metro Line 6 tunnel. Field monitoring and three-dimensional dynamic numerical simulations were combined to evaluate the spatial distribution and cumulative evolution of vibration velocity, axial stress, and anchor-bolt prestress loss during successive blasting cycles. The results showed that the lattice-girder crown exhibited substantially higher vibration velocity than the haunch and invert, and the vibration response generally decreased as the blasting face advanced. In the near-blast zone, the lattice girder remained predominantly in compression. Its transient stress during blasting and residual stress after blasting were distinguished to evaluate the effects of successive cycles. Anchor bolt velocity and prestress loss decreased with increasing distance from the blasting face, while the crown experienced the strongest disturbance. Prestress loss was concentrated near the blast-facing bolt tail. The lattice-girder crown and blast-facing anchor-bolt tail were identified as critical response zones and were recommended as priority locations for vibration monitoring, support-condition assessment, and stability evaluation under comparable construction conditions.
Traditional deterministic methods often assume constant soil parameters and overlook natural variability and uncertainty in soil properties, potentially leading to inaccurate or overly conservative safety estimates. To address this challenge, a deterministic computational model is developed in Excel spreadsheet to conduct a reliability-based probabilistic analysis of the soil slope using the first-order second-moment (FOSM) and Monte Carlo simulation (MCS) methods. The results indicate that the MCS provides a more reliable estimate of the probability of failure than the FOSM method. Three soft computing models, artificial neural network (ANN), particle swarm optimisation-ANN (PSO-ANN), and adaptive neuro-fuzzy inference system (ANFIS), are employed to predict the factor of safety and estimate the reliability index. The models' predictive performance is evaluated using various statistical metrics. The results show that the ANFIS model yields the best predictive performance as compared to ANN and PSO-ANN, with test results of Nash–Sutcliffe efficiency = 0.9993, root mean square error = 0.0088, variance accounted for = 99.9337, coefficient of determination = 0.9993, bias factor = 0.9999, performance index = 1.9898, and reliability index = 2.3272. Further validation using Taylor diagrams and ROC curve analysis confirms that all models perform well.
Rockfalls pose a significant hazard to transportation corridors due to their sudden occurrence, high velocity, and impact energy. This study assessed rockfall hazard along State Highway-5 in the Markundi Hills, Uttar Pradesh, India, where five representative road-cut slopes developed in Dhandraul Sandstone of the Vindhyan Supergroup were investigated through geological mapping. The slopes exhibited unfavourably oriented discontinuities and anthropogenic modifications associated with road-cutting activities. The Rockfall Hazard Rating System (RHRS) was applied to rank slope hazard, followed by two-dimensional probabilistic rockfall trajectory analysis. RHRS scores ranged from 282 to 546, identifying one critically hazardous slope. Numerical simulations indicated that detached blocks could traverse the entire road width, reaching a maximum bounce height of 9.5 m, translational velocity of 26 m/s, rotational velocity of 82 rad/s, and kinetic energy of 200 kJ, indicating substantial potential for vehicle impact. Unstable crest blocks acting as rockfall seeders were identified as the primary initiation mechanism. Considering the competent and tectonically stable sandstone, controlled removal of unstable source blocks was identified as the most effective and economically sustainable mitigation measure, offering a practical alternative to costly passive protection systems.
Deflectometer testing is a practical tool for assessing the mechanical behavior of materials in both laboratory samples and compacted or natural soil layers. However, these devices are quite expensive, and not all institutions can afford them. In this research, a low-cost, lightweight deflectometer was developed using inexpensive building materials and sensors, namely a load cell and a geophone. The working principles, connections, and calibration procedures of these sensors are presented. Furthermore, to assess the accuracy of the developed equipment, tests were performed on samples compacted in modified Proctor molds using the equipment and a standardized deflectometer as references. Three materials were employed to represent typical soils used in pavement layers: a base, a subbase, and a subgrade. The obtained moduli indicate that, although the moduli differ between the two devices, both follow the same trend across the materials. Finally, a calibration curve was obtained for the developed LWD, which, when applied, scales the results to be comparable to those measured with the standardized deflectometer. These results show that, with the right design strategies and calibration methods, low-cost, self-made equipment can be a viable alternative to more costly devices.
This study developed a fast and physically consistent surrogate of the constitutive response of the Hardening Soil model, whose stress-dependent stiffness makes finite-element parametric studies costly. A physics-informed neural network was formulated to learn the deviatoric stress-strain response of the drained triaxial test; the Kondner hyperbolic governing equation and the stress-dependent stiffness law were embedded as a physics residual in the loss function, and the zero-origin condition was imposed as a hard constraint. A material-point finite-element data engine generated training and validation curves over a wide parameter space, and the model was benchmarked against an identically configured data-driven network. Using only fifty noisy training curves, the physics-informed network reached a coefficient of determination of 0.9998 on the test set, whereas the data-driven network needed about thirty times more data to match this accuracy. In the extrapolation domain its root-mean-square error was about thirteen times lower. Against 132 real triaxial secant-modulus points from three Ho Chi Minh City soil layers, the network reproduced the modulus with a mean absolute percentage error of 1.89
This study developed a hybrid deep learning framework that combined a three-dimensional convolutional neural network (3D CNN) with a multilayer perceptron (MLP) to predict pile head settlement of bored piles under vertical load and to estimate axial bearing capacity. The 3D CNN branch encoded a fixed voxel representation of the soil–pile system that captured geometric and zone-of-influence features, while the MLP branch learned the effects of the applied load and eight soil mechanical parameters. The two branches were fused through a regression head to predict settlement and reconstruct the load–settlement curve. A dataset of 105 finite element cases in fine sand, aggregated from about 475 million raw rows, was split case-wise to prevent leakage between simulation conditions. The trained model achieved a coefficient of determination of 0.998, and a root mean square error of 8–16 mm across the dataset, and it reproduced the elastic, transitional, and large-deformation stages of the load–settlement response. Using the 10
Repeated wetting-drying can cause mechanical and microstructural deterioration of sandstone slopes, particularly in high rainfall environments. This study examines the response of silty and sandy sandstone from Aizawl, India. Specimens were subjected to repeated wetting-drying cycles under laboratory conditions. Brazilian tensile strength (BTS), P-wave velocity, elastic parameters, and slake durability were evaluated. Microstructural changes, image-based porosity, and fractal dimension were also examined. Silty sandstone showed a 79.2
A numerical study is presented for the load bearing capacity and settlement of skirted ring footings under vertical and inclined loading in loose sandy soils. The ring footings were analyzed using finite element analysis software with a detailed parametric study conducted to evaluate how the depth of embedment of the skirts (expressed in terms of the depth-to-diameter ratio, H/D = 0, 0.25, 0.5, 1.0, and 1.5), and the load inclination angle (5°, 10°, and 15°), affect the performance of ring footings. Results were validated by comparing with published experimental data and very good agreement was achieved, establishing the reliability of the numerical approach adopted. The results showed that the increment in the skirt depth has a significant effect on the bearing capacity, with a 400
This paper presented an efficient constitutive model requiring a minimal number of parameters within the hypoplasticity and generalized plasticity frameworks. The model was developed based on the Modified Cam-Clay model to predict the cyclic and dynamic behavior of soils. Similar to models proposed within those frameworks, the developed model treated soil continuously as an elastoplastic material. It did not employ a yield surface, nor was any purely elastic region assumed. Despite its small number of easily calibrated parameters, the developed model performed excellently in predicting the cyclic and dynamic behavior of soils. The model could readily be added to and implemented in existing applications for static and dynamic numerical analyses. It was calibrated for several soils, and its predictive performance was evaluated using cyclic consolidation and cyclic triaxial tests. Its dynamic performance was assessed through centrifuge tests on several embankments, where it was implemented in the numerical analyses. It was concluded that the developed model could reproduce soil behavior well under dynamic loading, as demonstrated by both the single-element tests and the embankment centrifuge tests.
Ketahun Sub-district in North Bengkulu is located in a region characterised by significant seismic activity, resulting in complex subsurface conditions. This study seeks to analyse the characteristics and distribution of subsurface dynamic parameters using data from 15 field investigation points integrating shear wave velocity, density, Poisson’s ratio, and elasticity modulus. In contrast to standard two-dimensional mapping, the novelty of this study lies in the application of a three-dimensional spatial hybrid framework through the hierarchical integration of the ordinary kriging method with radial basis function-based implicit geological modelling to present the variability of elastic parameters continuously. The results indicated that the distribution of dynamic parameters developed laterally and vertically following stratigraphic and morphological conditions, where soil stiffness was dominated by soft clay vulnerable to local seismic amplification. Statistical analysis confirmed that data distribution had low to moderate variations with strong linear relationships among the parameters. Overall, this study offers a comprehensive depiction of the geotechnical conditions and subsurface seismic characteristics. This approach is expected to serve as a reference for developing disaster mitigation studies and geotechnical evaluations in regions with high seismic activity.
A total of 162 three-dimensional numerical models were simulated to evaluate the stabilization of a two-layer undrained clay slope using a row of micropiles. The slope had an inclination angle of 26.6° (2H:1V), with soft clay (su = 30 kPa) overlying stiff clay (su = 150 kPa). The inclination of the interface at the base of the soft layer (α) varied from 0° to 20°, corresponding to α/θ ratios of 0–0.75. Six dimensionless micropile spacings (s/DMP) ranging from 2 to 12 were considered. The numerical models incorporated the two-layer slope, micropile row, soil–micropile interface, and interlayer interface. Micropile row location, s/DMP, and α/θ were varied to quantify their effects on slope displacement and stabilization performance. The results showed that placing the micropile row at the slope crest with s/DMP = 4.0 provided the most favorable stabilization response, reducing lateral displacement by 20.52
A simplified analytical-stochastic procedure was developed to examine progressive internal failure of soil nail retaining walls used in transportation infrastructure. Tensile rupture and pull-out were represented through demand–capacity safety ratios, while friction angle, cohesion, unit weight, and soil–nail bond strength were modeled as random variables. Monte Carlo simulation was used to estimate nail-level failure probabilities. A stage-based transition matrix was then constructed to summarize the evolution of a selected progressive-failure sequence after individual nail levels became non-operative. System states were defined by the number of failed nail levels, and load redistribution was represented through an assumed rule. The resulting probabilities were therefore interpreted as comparative, screening-level measures rather than direct indicators of design safety. Application to a representative wall showed that pull-out governed the probabilistic response for the examined configuration. The probability of avoiding additional nail-level failure remained high during the initial stages but decreased markedly as failures accumulated, illustrating the progressive loss of redundancy within the reinforced retaining system.
A deep neural network was developed to simultaneously identify eight parameters of cohesionless sandy soil: five constitutive parameters of the linear-elastic perfectly plastic Mohr–Coulomb model (effective elastic modulus, Poisson’s ratio, effective cohesion, internal friction angle, and dilatancy angle) and three physical state properties (unsaturated and saturated unit weights and void ratio). Three monitoring variables (absolute settlement, applied load, and loading-step duration) were augmented with three physics-based derived features: compliance, stiffness, and a logarithmic settlement transform. The six-dimensional input was mapped to the eight outputs by a network with hidden layers of 64, 128, and 64 neurons, regularized by batch normalization and dropout. A database of 418 finite-element samples covering 19 soil groups was split by group into 286 training, 44 validation, and 88 test samples. On the independent test set, the network achieved coefficients of determination of 0.61–0.84 and mean absolute percentage errors of 0.45–5.08
Rock slope instability was assessed along the Varandh Ghat corridor (NH-965DD), Western Ghats, India, using field-based geomechanical characterization of sixteen representative slopes (VS1-VS16). Lithology, slope geometry, discontinuity orientation, spacing, persistence, surface and groundwater conditions were evaluated using Rock Mass Rating (RMR), Geological Strength Index (GSI), Slope Mass Rating (SMR), Q-slope, kinematic analysis and Hazard Index (HI) methods. The GSI indicated moderate to good rock mass quality at most locations, whereas RMR values of -5 to 66 reflected reductions in effective rock mass quality where adverse discontinuity orientations were considered. SMR values ranged from −12 to 63, indicating stable to critically unstable conditions, with wedge and planar failures identified as the dominant structurally controlled failure modes. More than 60
A unified mechanistic framework was developed to relate column fill properties, load-transfer mechanisms, and system performance in geosynthetic-encased granular column (GEC)- supported embankments on soft ground. Consolidated drained triaxial tests were carried out on three marine-origin sands and three reference granular materials, giving friction angles of 36–40° and elastic moduli of 25–32 MPa. These parameters were introduced into a three-dimensional finite element model calibrated against field measurements of stress and settlement under staged construction. Parametric analyses covered the internal friction angle, elastic modulus, encasement stiffness and improvement area ratio. The friction angle governed load transfer, since raising it from 30° to 45° increased the settlement reduction ratio from 0.53 to 0.74 after consolidation, whereas the elastic modulus lost influence beyond 40–60 MPa. Encasement stiffness and improvement area ratio controlled radial confinement and arching continuity. Performance thresholds were then defined for the stress concentration ratio, the geosynthetic support ratio and the settlement reduction ratio, and expressed as design charts. Applied to marine-origin sand, the framework predicted stress concentration ratios of 2.0–2.7, geosynthetic support ratios of 1.1–1.5 and settlement reduction ratios above 0.75, comparable with river sand fill.
Buried lifelines under transportation infrastructure, such as pipelines beneath highways and railways, are highly vulnerable to joint damage induced by cyclic surface traffic and foundation loading. This study presents a parametric numerical evaluation of Expanded Polystyrene (EPS) geofoam blocks as a protective compressible inclusion to mitigate pipeline displacements beneath dynamic surface loads. Finite element simulations utilizing an advanced small-strain stiffness soil model were systematically conducted to investigate the coupled effects of geofoam thickness (t = 0.25–1.0D), width (B = 1.0–3.0D), and spacing from the pipe crown (S = 0.2–1.0D). The numerical trajectories indicate that the geosynthetic inclusion enhances structural performance by mobilizing positive soil arching, trapping shear strains, and maximizing dynamic energy dissipation. The optimized geofoam configuration restricts cumulative pipe crown settlement by up to 50
A physics-based probabilistic back-analysis of a canal-side highway embankment failure along Highway No. 105, Mae Sot, Thailand, was performed and translated into an operational early warning framework. A coupled transient seepage–strength reduction finite element model was calibrated against the failure observed on 26 July 2024 at a cumulative rainfall of 578 mm, with strength parameters estimated from standard penetration test correlations and hydraulic parameters from pedotransfer functions. One-at-a-time screening of fourteen parameters identified the friction angle of the embankment fill as dominant (sensitivity coefficient 0.984), followed by fill cohesion, canal water table (WT) elevation, foundation cohesion, and fill unit weight. Monte Carlo simulation with 10,000 trials per scenario produced daily failure probability and reliability index curves for three seasonal canal WT conditions, and spatial variability checks confirmed the adequacy of point statistics. The pre-rainfall factor of safety decreased from 1.518 under dry-season WT to 1.309 under flood-season WT, and the critical rainfall threshold decreased by 63 mm. Thresholds for four performance levels provided lead times up to 19 days and were cross-validated against documented failure and flooding events along the same corridor. The results were synthesized into a warning chart requiring only a canal staff gauge and a rain gauge.
This study conducts a large-scale, 1 g down-scaled experimental investigation and numerical analysis for evaluating the response on tunnels underneath basement excavations in sandy soils with fines. A staged excavation was conducted to evaluate the impacts of complex basement excavation-tunnel interactions in sandy soils containing fines. The 1 g down-scaled experimental investigation provided the basis for subsequent three-dimensional numerical analysis, which enhanced comprehension of the problem. At the crown of the tunnel in the longitudinal direction the normalized shear modulus of the soil for the silty sand (SM) was, on average, 50
Global climate change is altering rainfall patterns and increasing the risk of landslides, prompting this study to investigate slope stability and responses through an integrated hydrological, statistical, and geotechnical framework. A 45-year dataset from 29 rainfall stations in Pacitan, East Java, Indonesia, was analyzed using statistical distributions to identify trends and predict extreme rainfall intensities for 5 to 100-year return periods. The results showed the existence of a localized increase in rainfall intensity exhibiting the most significant rise by reaching a predicted 50-year maximum of 416 mm. Subsequent numerical modeling was applied to evaluate slope safety factors (SF) under different rainfall scenarios and soil drainage conditions. The results showed that soil drainage capacity was the most important factor controlling slope stability. Poor drainage conditions characterized by low hydraulic conductivity at 10⁻⁶ m/s led to a drastic reduction in the SF and triggered instability during extreme short-duration or prolonged rainfall events. In contrast, slopes with good drainage maintained stable SF values of 1.5–1.8. This study concluded that inadequate drainage significantly amplified slope vulnerability to provide a critical basis for targeted landslide mitigation strategies in tropical regions under a changing climate.
