The amplification of seismic waves by soft soils, known as site effect, makes the mapping of Quaternary deposits and geomorphic settings essential for seismic hazard zoning. We propose a framework for primary-level site effect assessment in urban areas. Our approach utilizes historical aerial photos and Structure from Motion (SfM) to reconstruct a high-resolution digital elevation model (DEM) of the pre-urban landscape. The integration of the high-resolution DEM with an orthophotomosaic and 3D visualizations provided a robust basis for identifying and classifying Quaternary deposits and tectonic landforms. Applied to Urmia, Iran, a high-seismic-hazard city, the method revealed that 74
The weak interlayer zone (WIZ), characterized by tectonic damage and weak cementation, is significantly impacted by load-induced damage accumulation and stiffness degradation under cyclic loading and unloading conditions, thereby posing a serious threat to the stability of underground cavern excavations. To address this engineering geological issue, a synergistic reinforcement technique combining enzyme-induced calcium carbonate precipitation and fiber reinforcement (EICP-FR) was employed to reinforce the WIZ. The macro-meso correlation between mesoscopic structural characteristics and macroscopic mechanical responses was elucidated through uniaxial cyclic loading–unloading tests combined with quantitative SEM-MATLAB analysis. The peak strength of the WIZ increases to 340
Columnar jointed rock masses (CJRMs) exhibit pronounced structural anisotropy and time-dependent deformation under long-term loading, posing significant challenges for the stability evaluation of large hydropower slopes. Conventional rheological constitutive models generally assume isotropy or neglect the influence of joint orientation, limiting their applicability to columnar basalt formations. To address this issue, this study develops an anisotropic rheological damage constitutive model for CJRMs based on microstructure tensor theory. By introducing an aging damage variable, the model quantitatively links rheological strain evolution to time-dependent damage processes, enabling unified characterization of both anisotropic and rheological behavior. The proposed model is implemented in FLAC3D through secondary development, and its validity is verified through comparison with triaxial rheological tests on Baihetan basalt sample. The numerical results reproduce the whole creep stage and closely match laboratory measurements, confirming the model’s capability to capture directional deformation. The model is further applied to the right-bank slope of the Baihetan Hydropower Station as a case study. The analysis reveals that incorporating structural anisotropy significantly improves agreement with field monitoring data compared with isotropic assumptions. The maximum calculated damage factor is 0.16, which is well below the adopted threshold value and corresponds to a safety factor of 1.48, indicating that the slope remains stable during long-term operation. This study establishes a validated constitutive framework and numerical approach for simulating anisotropic rheological damage in columnar jointed rock masses. The results enhance the reliability of long-term deformation prediction and provide practical guidance for the design and safety evaluation of large hydropower slopes dominated by CJRMs.
Alternating soft and hard layers are common in geotechnical engineering, but their interface shear mechanics remain poorly understood. This study combined direct shear tests on gypsum–based soft–hard joint analogues with three–dimensional discrete element method (DEM) simulations to investigate the effects of normal stress, wall–strength contrast, and joint roughness coefficient (JRC) on shear strength, acoustic emission (AE) response, crack evolution, and post–failure fracture topology. Both experimental and numerical results exhibited a four–stage shear response comprising elastic, damage, softening, and residual stages. Within the tested experimental matrix, normal stress produced the largest observed changes in peak and residual strengths and was associated with a greater relative contribution of tensile–type cracking and more distributed damage in the weaker wall. Increasing JRC enhanced asperity interlocking and shear–related damage, whereas increasing wall–strength ratio promoted asymmetric damage localization toward the weaker wall. AE monitoring and DEM bond–break analysis showed qualitative consistency in damage staging and relative crack–type proportions, although the DEM underestimated residual strength in some cases and did not reproduce the early compaction stage. Post–shear fracture maps showed broader damage zones at higher normal stress, more tortuous fracture paths with increasing roughness, and stronger one-sided localization with increasing strength contrast. These findings provide a multiscale basis for interpreting the shear failure of the tested soft–hard interfaces, while their broader applicability requires further validation using replicated tests, wider parameter ranges, and natural rock interfaces.
The unconfined compressive strength (UCS) is critical for evaluating frozen soil mechanics. Traditional empirical models often fail to capture the complex nonlinear relationships between UCS and varying key factors (temperature, water content, density) and ignore inherent prediction uncertainties. To address this, we propose a novel Partial Bayesian Neural Network (PBNN) model to predict frozen soil UCS, explicitly quantifying natural uncertainties through iterative sampling. Results demonstrate: (1) The PBNN model outperforms traditional machine learning algorithms, achieving a high Coefficient of Determination (R2) of 0.98. (2) Incorporating soil type as a categorical input significantly enhances predictive performance, with the unified U-UCS model showing optimal results. (3) The PBNN model effectively quantifies prediction uncertainty, providing robust reliability assessments. Feature analysis identifies temperature, mass water content, and dry density as the most critical factors influencing UCS. Additionally, we developed an open-source graphical user interface to enhance practical applicability. This study provides a reliable, data-driven approach for UCS prediction, facilitating safer cold-region infrastructure design.
In engineering applications, machine learning (ML) provides a promising approach when direct characterization of subsurface properties is challenging, time-consuming, and costly. This study developed a multi-class ML framework for classifying rock mass permeability based on Lugeon values using 86 observations from drilling and geophysical investigations at the Kırklartepe dam site in Bayburt, Türkiye. Seven predictors were considered: depth, Rock Quality Designation (RQD), joint frequency, Schmidt hammer hardness, P-wave velocity, S-wave velocity, and electrical resistivity. The observations were assigned to four permeability classes: very low, low, moderate, and high. Six classifiers were evaluated: Polynomial Kernel Support Vector Machine (P-SVM), Linear Discriminant Analysis, k-Nearest Neighbors (k-NN), AdaBoostM2, Bagged Trees, and Radial Basis Function SVM. Sequential Forward Feature Selection and Leave-One-Out Cross-Validation were used for feature selection and model evaluation, respectively. Robustness was further assessed using bootstrap confidence intervals, class-imbalance treatments, feature-selection stability, and learning curves. Among the baseline models, P-SVM using all seven predictors achieved the highest accuracy (79.07
In cold regions, repeated freeze–thaw disturbances can weaken saline soils and compromise long-term infrastructure performance, highlighting the need for durable, low-carbon-oriented stabilization. This research investigates the mechanical behavior and microstructural evolution of saline soils treated with a combined system of ionic soil stabilizer (ISS), lime, and fly ash, prepared under field-relevant compaction and subjected to controlled freeze-thaw conditions. Consolidated drained triaxial tests, mercury intrusion porosimetry, and scanning electron microscopy were performed to evaluate strength development and pore-structure evolution. The results show that the peak deviator stress increases progressively with freeze-thaw repetitions, increasing from 1707 to 2093 kPa before cycling to 2420–2502 kPa after 20 cycles under a confining pressure of 200 kPa, with the 9
This study presents an integrated approach that combines laboratory testing, numerical modeling, and artificial intelligence to enhance the assessment of soil behavior and slope stability. Direct shear tests were conducted on soils stabilized with varying proportions of lime and pozzolana under normal stresses of 50, 100, 200, and 300 kPa, with curing periods of 7 and 28 days. The results demonstrate a significant increase in shear strength, cohesion, and the internal friction angle with increasing stabilizer content and curing time, highlighting the synergistic effect of the lime–pozzolana mixture. The improved geotechnical parameters were subsequently incorporated into PLAXIS 2D simulations to evaluate the stability of an earthen dam using the c–φ reduction method within the framework of the Mohr–Coulomb model, resulting in a substantial improvement in the factor of safety. To mitigate the high computational cost associated with repeated finite element simulations, a Physics-Informed Neural Network (PINN) was developed as a surrogate model. The PINN accurately predicted displacement and stress fields while incorporating the governing elastostatic equations, achieving errors below 3
This study explores how grain size, temperature, and moisture affect the shear resistance and microstructure of granitic rocks. The granite samples were categorized into three grain sizes: Fine (F), Medium (M), and Coarse (C). Specimens were prepared as 3 cm-thick slabs with overall dimensions of 7 × 7 cm to assess their basic friction angle. The experimental program examined the effects of temperature at 25 °C, 275 °C, and 525 °C, as well as a saturated condition at 25 °C, to evaluate variations in the basic friction angle from tilt tests. The changes in grain size across different temperatures were analyzed using thin section microscopy. Results show that the basic friction angle (φb) increases with temperature, with coarse-grained samples experiencing the most significant increase. Furthermore, φb rises with grain size across all temperature levels and this trend is well represented by a proposed linear regression model. Saturation also enhances friction, leading to the highest φb under room temperature saturated conditions. The comparison of the basic friction angle for similar and dissimilar surfaces revealed that the impact of temperature could be similar or different depending on the surface combinations. This research links grain size distribution to mineralogical composition and demonstrates a synergistic effect of temperature and grain size on the basic friction angle. This integrated approach builds on previous work by concurrently considering grain size, thermal and moisture conditions, and mineralogical factors in granitic materials.
Freeze–thaw (F-T) cycling induces extensive open micro-fissures in dense fine sandstone, a lithology common in cold-region engineering, substantially prolonging the nonlinear "compaction-hardening" stage during early loading. Existing statistical damage models struggle to capture this behavior. Weibull-based formulations idealize early deformation as linear-elastic and thus overestimate the initial tangent stiffness. Compaction-modified models, in turn, typically rely on empirical piecewise segmentation or stress-driven implicit formulations, which introduce discontinuity or computational complexity. Moreover, no consensus exists on which physical quantity best defines the F-T damage variable. To address these limitations, this study proposes a strain-driven instantaneous compaction compliance model, and builds on it a generalized statistical damage constitutive model with an explicit analytical solution. Three damage definitions based on P-wave velocity, elastic modulus, and peak strain energy are introduced and systematically compared. The peak strain energy, though obtainable only after failure, yields the best pre-peak fitting accuracy and is adopted in the proposed model, raising R2 to 0.998 and reducing peak-stress prediction error to within 0.85
The Eastern Himalayan Syntaxis (EHS) is characterized by extreme climate, intense tectonism, and rugged terrain. Traditional landslide susceptibility assessment (LSA) models face limitations including negative sampling randomness, label distortion, and over-reliance on sample balance. These problems hinder the effective application of LSA models in high-steep mountainous areas. To overcome these bottlenecks, this study focuses on EHS and proposes a dynamic LSA framework integrating spatio-probabilistic co-constrained negative sampling and semi-supervised imbalanced learning. A comprehensive landslide inventory was refined utilizing Interferometric Synthetic Aperture Radar (InSAR) and high-resolution optical imagery. By combining Moran’s index and frequency ratio model, non-landslide sampling was achieved under the dual-constraints of spatial structure and local probability. Continuous probability labels were generated for non-landslide samples to reflect their potential instability. Landslide and non-landslide samples were subjected to semi-supervised imbalanced LSA through regression forest algorithm. Finally, the traditional InSAR slope projection model was optimized by integrating slope and aspect, and optimized slope deformation rate was integrated into the semi-supervised imbalanced LSA to obtain the final LSA result. Results demonstrate that the proposed spatio-probabilistic non-landslide sampling approach (the area under the curve (AUC) = 0.968) significantly outperforms traditional non-landslide sampling method. Integrating improved InSAR slope deformation rates boosted the optimal semi-supervised model’s (imbalance ratio 1:5) AUC to 0.987, accurately identifying Sedongpu Gully’s high-risk areas. Furthermore, Shapley Additive Explanations indicates that excessive vegetation coverage will have a negative hydrological feedback effect on the development of landslides. This study provides a reliable methodology for geological hazard assessment and prevention in active tectonic orogens.
Compacted loess exhibits pronounced water sensitivity, and the seepage-stress coupling interaction ultimately leads to significant deformation and degradation in structures. To investigate the seepage-shear characteristics, structural degradation mechanisms, and triggering processes under this coupling effect, this study employed triaxial seepage-shear tests along with nuclear magnetic resonance (NMR) and scanning electron microscopy (SEM) techniques. The results demonstrate that seepage pressure significantly impacts the cohesion of compacted loess, causing a notable reduction in cohesion. The permeability coefficient exhibits an evolution pattern characterized by an initial decrease followed by an increase. The specimen was partitioned into five equidistant layers (L1-L5): Horizontal cross-sections were prepared for odd-numbered layers, while vertical cross-sections were obtained for even-numbered layers to capture multidimensional microstructural features. Pore directionality is diffusely distributed across the first, third, and fifth layers, while it remains concentrated in the vertical direction in the second and fourth layers. Under seepage-stress coupling, the degradation-triggering mechanisms in compacted loess diverge depending on the dominant physical field. The stress field prevails: small and medium pores evolve into larger pores, while decreased circularity and increased elongation facilitate pore interconnectivity and microcrack initiation. Conversely, the seepage field dominates: pore size distributions shift toward the medium pores range, and a significant increase in circularity leads to the proliferation of seepage overflow points. These findings offer valuable scientific insight into the deformation and degradation behavior of compacted loess in dams and other fill projects subjected to seepage-stress coupling conditions.
In the tunnel engineering, to ensure the safety and the efficiency of the tunnel construction process, a variety of excavation methods and supporting measures were designed and adopted. To improve the tunnel stability meanwhile shorten the construction period, two novel modifications, namely advanced oblique cable supporting system and sync-three-bench method were first adopted in the Chongqing-Kunming high-speed railway project. In this paper, the excavation mechanism of tunnels under these two new modifications was revealed. Firstly, to better capture the mechanical response of surrounding rocks, a novel approach to describe the rock damage behaviors was put forward. Triaxial tests were conducted to acquire rock properties required in the proposed method, then the effectiveness of the acquired rock properties was examined through simulations. Secondly, the full tunnel excavation process was simulated, and the rock damage and deformation behaviors in both longitudinal and radial direction were explained according to the simulation results. To examine the rationality and the correctness of simulation results, the on-site monitoring data and testing data were adopted to compare with the simulation results. Thirdly, the influence of geo-stress, the rock GSI, the excavation footage and the liner thickness to the rock damage and rock mechanical parameters were analyzed systematically. The findings of this article can serve as a reference for promoting the use of these two novel modifications in drilling and blasting tunnels.
A novel constitutive model for unsaturated frozen soil is developed within the granular thermodynamics framework by incorporating the water–ice phase transition and frost heave effect. Double entropy theory and granular temperature are introduced to describe energy dissipation, particle rearrangement, and microstructural evolution during freezing. The frost heave effect is embedded into the elastic potential energy density function, enabling a unified description of coupled thermal, hydraulic, and mechanical responses. Sensitivity analysis reveals that the limiting density parameter M0 predominantly controls the volumetric strain, while the phase transition parameter B governs the freezing rate. The model shows good agreement with experimental results in predicting the unfrozen water content, volumetric deformation, and stress–strain behavior. The model also captures the transition from strain softening to strain hardening as confining pressure increases, providing a practical tool for frost heave risk assessment and stability evaluation of cold-region geotechnical structures.
Soil anisotropy in fully saturated sands has been widely studied, whereas its role in gas-bearing sands remains insufficiently understood. Free gas can stabilise loose states yet weaken dense states, and existing models rarely reproduce gas-induced weakening or unloading-driven instability. These limitations hinder assessment of submarine slope stability under tidal water-level fluctuations, where pore-pressure changes and gas exsolution may occur. This study develops an anisotropic critical state framework for gassy sands by coupling pressure-dependent pore-fluid compressibility governed by Boyle’s and Henry’s laws with evolving deviatoric fabric anisotropy. The formulation is calibrated and validated against undrained isotropic unloading, constant-q unloading, and undrained triaxial compression and extension tests across loose-to-dense states and different gas types. Implemented in ABAQUS via coupled UMAT and UEL subroutines, the model reproduces gas-exsolution-induced collapse during unloading and captures stress-path-dependent reversals of gas effects in medium-dense sands. The framework is then applied to submarine slope simulations under rapid water-level drawdown, with instability tracked using second-order work. Results show that density, gas saturation, bedding inclination and drawdown magnitude jointly control failure initiation and shear-band development, and inclined bedding markedly amplifies deformation. The proposed approach links multiphase pore-fluid behaviour, fabric evolution and unloading-induced instability, providing a mechanistic basis for assessing geohazards in gas-bearing offshore slopes.
Specimen size is a critical factor influencing the mechanical strength behavior of building stones. In this study, the effect of specimen size on the uniaxial compressive strength (UCS), Brazilian tensile strength (BTS), and point load index (PLI) of different marbles was investigated. Furthermore, correlation equations for predicting UCS using BTS and PLI were developed, explicitly accounting for the effect of specimen size. To this end, UCS specimens with length-to-diameter (L/D) ratios of 2.0, 2.2, 2.4, 2.6, 2.8, and 3.0 were prepared from the marble samples. For BTS tests, specimens with thickness-to-diameter (T/D) ratios of 0.5, 0.6, 0.7, 0.8, 0.9, and 1.0 were extracted. Similarly, PLI specimens were prepared with T/D ratios of 0.30, 0.44, 0.58, 0.72, 0.86, and 1.00. The findings revealed a noticeable decrease in UCS, BTS, and PLI values as the L/D and T/D ratios increased. Based on scanning electron microscopy (SEM) observations, this trend was attributed to the higher probability of microcracks and internal flaws in larger-sized specimens. The results indicated that specimen size has a minor effect on the accuracy of the BTS and PLI-based correlation equations for predicting UCS. Additionally, it was observed that BTS is a more reliable parameter than PLI for predicting UCS.
This study proposes a probabilistic framework for regional hazard assessment of excavation-disturbed translational rockslides. The framework integrates field investigation, laboratory testing, key stratigraphic parameter prediction, and Monte Carlo limit-equilibrium analysis to quantify slope-unit failure probability under data-limited conditions. It was applied to 139 slope units in the resettlement town of Yanshan Township, Chongqing, where slopes are mainly composed of a shallow surficial deposit, fractured rock mass, weak interlayer, and hard bedrock. After considering excavation effects, a new high-hazard zone of 0.0449 km2, accounting for 3.21
The study used ROC (Receiver Operating Characteristic)-based sensitivity analysis to analyse landslide conditioning factors and optimize landslide susceptibility. In this regard, two landslide susceptibility zone maps were prepared by the frequency ratio (FR) and WOE (Weight of Evidence) models, and model precision was estimated by AUC (Area under curve) and statistical error. Meanwhile, the sensitivity of factors was judged by removing a single layer from the original model, and AUC values were determined. The ROC-based sensitivity revealed that the AUC value increased after the removal of geomorphology, curvature, and TWI (Topographic Wetness Index), whereas the AUC value sharply decreased after the removal of DTR (Distance to Roads) and NDVI (Normalised Difference Vegetation Index). Next, the optimization process by grouping different combinations of conditioning factors reflected the combination of six factors, i.e., lithology, NDVI, aspect, DTR, (Distance to lineaments) DTL, and slope, which has optimum precision by the highest AUC value and less error.
Extreme rainfall creates concentrated catchment areas in mountainous regions, significantly increasing the risk of subgrade slope collapse. This paper focuses on the quantitative analysis of hydrodynamic erosion. Using 3D laser scanning and the SCS distributed hydrological model, reveals the evolution patterns of confluence and quantifies the intensifying effect of hydraulic gradients on erosion at the slope toe. A limit analysis method for slope stability considering the synergistic mechanism of internal seepage and external runoff scouring was developed, and its validity was verified through numerical simulation in GeoStudio. The results show that the hydraulic erosion distance increases continuously with rainfall duration, incorporating hydraulic gradient increases erosion distance by 31.5
Collapse failures have occurred in high liquid limit clay (CH) slopes along the Bengbu section due to arid climatic conditions. Given that anisotropic parameters are critical for slope stability evaluation, this study investigates the anisotropy of undisturbed CH samples collected at three orientations under two stress paths: maximum drying stress (MDS) and the combined action of maximum drying stress and spherical stress (MDS-SS), using triaxial shear tests and microstructural analyses. Results show that CH exhibits pronounced anisotropy, with the MDS and MDS-SS paths inducing vertical crack development and healing behavior, respectively. Firstly, illite-dominated CH exhibits inherent anisotropy, as the fixation of interlayer potassium ions suppresses hydration and interparticle rearrangement. Secondly, under the MDS path, the anisotropic morphology and arrangement of clay particles lead to an anisotropic distribution of matric suction, resulting in macroscopic vertical cracking, strength degradation, and a microscale “polarization” in particle orientation. Finally, under the MDS-SS path, the application of spherical stress facilitates crack closure within the soil skeleton, shifting macroscopic mechanical behavior from anisotropy (where effective stress paths may evolve into shear contraction or dilation modes) toward isotropy (characterized by a softening–partial contraction–partial dilation path). Compared to the undisturbed samples, MDS-SS samples exhibit higher strength and a microscale “depolarization” in particle morphology and alignment.
