PurposeThe purpose of this study is to investigate the effects of spatial variability and parameter correlation on the structural responses and reliability of moderately buried tunnels during excavation. The study aims to clarify how cross-correlated random fields (CRF) and stratigraphic anisotropy influence tunnel failure mechanisms and to provide quantitative insights for reliability-based tunnel design.Design/methodology/approachA stochastic numerical framework was developed based on Copula theory to generate CRF for the cohesion and internal friction angle of surrounding rocks. The random fields were incorporated into a two-dimensional numerical model, and 65,100 Monte Carlo simulations were performed to evaluate tunnel responses under various stratigraphic dip angles and Kendall correlation coefficients. The probability of tunnel failure was quantified using two indicators: crown settlement and the proportion of the plastic zone.FindingsThe results show that parameter correlation strongly affects tunnel reliability. As the Kendall correlation coefficient increases from negative to positive, the dominant failure mode shifts from crown-settlement-controlled to plastic-zone-dominated behaviour. The smaller of the horizontal and vertical scales of fluctuation (SOF) governs the lower bound of tunnel failure probability. Although the stratigraphic dip has limited influence on settlement, it significantly affects plastic zone development, revealing the anisotropic nature of spatial variability in rock masses.Originality/valueThis study provides a quantitative framework for integrating parameter correlation and spatial variability into tunnel reliability analysis. The findings enhance the understanding of anisotropic effects in red-bed rock masses and offer practical guidance for the design and risk assessment of tunnels in heterogeneous geological environments.

Large-scale slope engineering projects are typically characterized by prolonged construction periods and extended service lives. The long-term deformation of rock masses can lead to geological disasters, such as mass collapse and sliding. Therefore, accurate prediction of long-term slope deformation is essential for both engineering design and disaster mitigation. This study proposes an integrated prediction framework that combines a rheological constitutive model, Bayesian estimation, and numerical simulation, and applies it to the left-bank slope of the Baihetan Hydropower Station. Triaxial creep compression tests with step loading were conducted, and the experimental data were used as observation inputs. The Burgers model was used as the theoretical foundation, and Bayesian estimation was employed to update the rheological parameters. Subsequent sensitivity and correlation analyses were conducted to evaluate parameter behavior. Finally, the updated parameters were incorporated into a numerical model to simulate the long-term deformation of the slope. The results demonstrate that the Bayesian estimation method effectively reduces and quantifies the uncertainty associated with rheological parameters. Additionally, the simulation results based on the calibrated parameters exhibit good agreement with the field monitoring data. These findings provide valuable insights for quantifying rheological uncertainty and enhancing the prediction of long-term deformation in slope engineering.

Hydrodynamic erosion is widespread in colluvial and reservoir-bank deposits, where gap-graded soil-rock mixtures (SRMs) are highly susceptible to erosion under hydraulic conditions. However, the mechanisms governing hydrodynamic erosion in gap-graded SRMs remain unclear. In this study, a coupled discrete element method and computational fluid dynamics framework is developed to investigate the hydrodynamic erosion of gap-graded SRMs, in which the discrete element method tracks particle motion and contact forces, while the computational fluid dynamics module computes pore-scale fluid flow based on Darcy's law. Two-way coupling is achieved by exchanging porosity and fluid drag forces at each time step, allowing the fluid field and particle dynamics to be updated. The numerical approach is validated through two laboratory tests for simulating hydrodynamic erosion. A series of numerical simulations are conducted to investigate the influence of rock content (RC), rock particle shape, and hydraulic gradient on hydrodynamic erosion behavior. The results indicate that higher RC generally leads to greater cumulative fine particle loss, accompanied by highly heterogeneous flow fields with preferential erosion near boundaries. Irregular rock particles further restrict pore connectivity, suppressing continuous seepage channels while enhancing localized clogging.

Landslide-generated impulse waves (LGIWs) occurring in mountain reservoirs exhibit obvious topography-related characteristics during long-term evolution. This paper conducts a comprehensive study on the run-up characteristics of impulse waves on the dam built downstream of the river bend. Large-scale physical similarity model experiments are carried out based on the potential LGIWs hazards in the Rumei (RM) reservoir to investigate the run-up process of impulse waves on the dam. Numerical simulations are then conducted using a hybrid discrete element method-smoothed particle hydrodynamics and shallow water equations model to further analyze the LGIWs disaster and the run-up evolution of impulse waves. The reasons for the complex run-up characteristics are analyzed, the influence of bends under different dam surface inclinations, and the dam position is also discussed. The results show that under the influence of the bend, the run-up process exhibits distinct temporal and spatial effects. Three wave motion patterns gradually appear at the dam surface due to the influence of the bend in front of the dam: reflected waves between the flanks, incident waves from the river, and water rebound under gravity. The maximum run-up height occurs when various factors are superimposed simultaneously, which may cause even a small incident wave to generate a large run-up height. As the inclination of the dam surface decreases, the influence of the bend on the run-up process gradually diminishes. Under the effect of the bend, the disaster of LGIWs may be amplified. This study recommends that the whole lifecycle of the LGIWs should be studied based on high-resolution terrain to get a more accurate assessment of the LGIWs disaster.

Flood discharge atomized rainfall can rapidly alter the hydraulic and mechanical states of fractured rock slopes, yet its physical simulation is constrained by the difficulty of preparing materials that reproduce both mechanical and hydraulic similarity. To address this limitation, this study develops and validates a solid–liquid coupled similar material for fractured rock slope model testing. An orthogonal design was conducted using barite powder and quartz sand as aggregates and cement-gypsum as the binder to regulate density, stiffness, strength, and permeability. The prepared materials covered elastic modulus of 35⁓171 MPa, uniaxial compressive strength of 0.11⁓0.88 MPa, tensile strength of 17⁓123 kPa, cohesion of 24⁓88 kPa, internal friction angle of 26⁓49°, and permeability coefficient of 1.78 × 10–6⁓7.7 × 10–5 m/s. Sensitivity analysis showed that the mass ratio of binder to aggregate predominantly controlled stiffness and strength, whereas the mass ratio of water to aggregate and aggregate composition mainly governed pore structure and permeability. Regression models with R2 values of 0.81⁓0.88 were established to support targeted material proportioning for bedrock, unloaded rock, and fracture zones. The developed materials were further applied to a physical model of the BDa Hydropower Station slope subjected to atomized rainfall. Monitoring results revealed a fracture-controlled hydraulic-mechanical failure process. Preferential infiltration induced rapid moisture accumulation at W2 and W3, pore pressure peaks of 2.0 and 1.8 kPa at P2 and P3, localized strength attenuation, erosion-gully coalescence, and progressive collapse from the slope toe toward the middle slope. These findings demonstrate that the developed material can effectively reproduce fracture guided seepage, water-induced softening, and progressive instability in fractured rock slopes, providing a practical experimental basis for evaluating slope stability under flood discharge atomization.

This study develops a mechanics-informed sparse-observation strategy for saturated muddy sandstone under cyclic hydro-mechanical loading. Cyclic triaxial loading-unloading tests were conducted at confining pressures of 3, 5, 7, and 10 MPa, with pore pressure fixed at 1 MPa. Confinement significantly affected strength, plastic deformation, stiffness evolution, post-peak degradation, and residual load-bearing capacity. The peak point of each cycle was extracted to construct the outer envelope. These points represented pre-peak hardening, peak-strength mobilization, post-peak degradation, and residual stabilization. They were assimilated into an elastoplastic damage model using the iterative local updating ensemble smoother. Under 10 MPa confinement, observation schemes containing 3, 7, 10, and 15 points were compared against the complete 20-point peak dataset. Results showed that reconstruction accuracy depended more on coverage of key mechanical stages than on observation number alone. The stage-based 10-point scheme used only half of the complete peak dataset, yet closely reproduced the full outer-envelope response. The same strategy was then applied independently at all four confining pressures. The reconstructed responses captured the main experimental features and agreed well with withheld peak observations. Parameters governing stiffness, strength mobilization, and residual response were more strongly constrained. Parameters associated with detailed plastic-damage interactions retained greater posterior uncertainty. The proposed strategy provides a practical basis for reduced-data constitutive updating and uncertainty quantification under limited laboratory and field observations.

Time-dependent large deformation of surrounding rock poses significant challenges to the safety and stability of tunnels excavated in soft rock formations. Accurate prediction of such deformation remains difficult due to the complex rheological behavior of soft rock and the uncertainty in model parameters. This study develops a Bayesian framework that integrates a viscoelastic-viscoplastic analytical model with field monitoring data to improve both the prediction accuracy and mechanical interpretability of deformation behavior in soft rock tunnels. Based on Markov Chain Monte Carlo (MCMC) method, posterior distributions of model parameters are inferred, followed by comprehensive parametric investigation including uncertainty quantification, correlation analysis, and sensitivity assessment. The results demonstrate e that the proposed framework can accurately capture both the magnitude and timing of deformation within a 95% credible interval. Notably, the inferred temporal evolution of the plastic zone radius, treated as an implicit intermediate variable in the analytical model, provides insight into the time-dependent migration of the elastoplastic boundary associated with stress redistribution during tunnel deformation. Compared with the classical Burgers model, the analytical model exhibits more stable predictive behavior and improved mechanical interpretability under the same Bayesian framework, underscoring the practical value of the proposed approach for tunnel engineering applications.

The construction of large hydropower projects in southwestern China has produced numerous high-steep rock slopes; however, accurately assessing their long-term creep-induced degradation remains challenging. This study addresses this issue by examining the high-steep slopes at the upper reservoir inlet of the Lianghekou pumped storage power station (PSPS). Satellite-based interferometric synthetic aperture radar (InSAR) technology, optimized for steep mountainous terrain, was employed to extract slope deformation data. The InSAR-derived data were integrated with multi-point extensometer measurements. These resulting multi-source datasets were then jointly applied to constrain a numerical model, enabling the quantitative evaluation of the evolution of key mechanical parameters under the combined effects of slope excavation and reservoir impoundment. A generative adversarial network (GAN) was subsequently employed to augment the dataset. Furthermore, a dual-layer ensemble learning regression model featuring adaptive lightweight architecture and threat-aware dynamic topology optimization was developed. The results demonstrate that the strength and stiffness of the rock mass decreased by approximately 4.43

PurposeIn order to analyze the influence of uncertainty geological structures on the stability of reservoir slope.Design/methodology/approachAn uncertainty analysis method of reservoir slopes, accounting for uncertain geological structures, is proposed. This approach employs an unsaturated constitutive relation for unsaturated mechanical simulations and uses geostatistical simulation to generate maps of uncertain geological structures. The proposed method is applied to analyze the stability of the Zhoujia landslide, concluding that the slope becomes unstable when the water level rises from the original to the design normal level. Additionally, the effects of the weak layer's inclination and the rock block content on slope stability are examined using this method.FindingsIt was found that weak layer inclination may not significantly influence slope stability. And the simulations show that higher rock block content leads to decreased slope stability due to increased gravitational loading.Originality/valueFindings of this research highlight the importance of considering geological uncertainties and rock block characteristics in slope stability assessments and in designing effective stabilization measures.

Impulse waves generated by partially submerged landslides usually occur in mountain reservoirs, which may endanger human lives and the safe operation of the reservoir. The objective of this study is to investigate the characteristics of waves generated by partially submerged landslides using a large-scale physical similarity model based on the Rongsong deposit. River cobbles are used to replicate the landslide body, and three sliding velocities are achieved by changing the material of the sliding surface. The results show that wave heights are significantly influenced by topographical features of the river. The second wave is higher than the leading wave upon generation, but the crest of the third wave is the highest at the entrance of the river channel. The waves running up at the dam surface reflect and superimpose with the waves propagating from the river, which could lead to higher runup height on the dam. There may be a great risk in considering only the leading wave for waves generated by a partially submerged landslide. Based on the topographic characteristics of V-shaped channels, the effective still water depth is proposed to obtain an accurate celerity. The results are useful in understanding interactions between waves generated by partially submerged landslide and dam in mountain reservoirs, and can provide an important reference for the treatment of potential landslides and the design of dams.

This study investigated the deformation behaviours, trigger mechanism and deformation pattern of Dawanzi (DWZ) slope, a bedding rock structure in the Baihetan reservoir area, based on the field geology surveys, site monitoring and geospatial statistical analysis methods. Monitoring results indicated that the displacement behaviours of measuring points TP01 and TP02 near the DWZ tunnel entrance area exhibited a creep deformation stage, while data from the other monitoring points remained stable, thereby confirming the effectiveness of the emergency anti-slide pile reinforcement. Furthermore, the correlations between slope stability and slope structure, gradient and lithology were analysed based on the statistical characteristics of 28 historical landslides in the Baihetan reservoir area. The statistical results revealed that the DWZ slope is a bedding rock slope with an inclination of approximately 35 degrees and a lithology characterised by a soft-hard interbedded sequence primarily composed of siltstone, silty mudstone and argillaceous dolomite, identifying it as a prone stratum. The rapid rise of the water level triggered bank collapse at the leading edge of the DWZ slope, which subsequently pulled the trailing edge rock mass forward along the slope, leading to continued deformation. The displacement normalisation method was applied to analyse the deformation pattern, indicating that the slope is currently in a steady-state deformation stage. This study provides a theoretical reference for the prevention and control of geological disasters in the reservoir area.

As a natural material, rock is inherently fractured with discontinuities exhibiting pronounced spatial variability in geometry and high nonlinearity in mechanical property-stress relationships. Each fracture involves multiple interdependent parameters that collectively form a high-dimensional, heterogenous feature space, making their identification and characterization a fundamental challenge in rock mechanics. Conventional Kalman-based ensemble smoother (ES(K)), based on linear covariance updates, is inadequate for capturing the complex influence of fracture networks on the overall mechanical behavior of rock masses. To address this challenge, this study develops a deep learning (DL)-based ensemble smoother (ES(DL)), in which a feature-weighting block is introduced to explicitly account for the heterogeneity of fracture-related parameters during the update process. By enabling adaptive reweighting of high-dimensional fracture features within the DL-based update operator, the proposed method enhances the representation of strongly nonlinear parameter-response relationships in rock fracture systems. Numerical experiments indicate that the proposed approach provides more accurate posterior estimation of high-dimensional and strongly heterogenous fracture parameters than the conventional ES(K) method, accompanied by a more effective reduction in parameter uncertainty. The feature-weighting enhanced ES(DL) is capable of reproducing the full stress-strain response of fractured rock and exhibits distinct advantages in the peak strength regime, where fracture-induced process nonlinearity is most pronounced, the method is examined through an engineering application involving rock specimens from the F115 fault damage zone, in which the proposed ES(DL) consistently captures fracture-governed mechanical behavior and associated parameter variability, providing clear evidence of its applicability and robustness for practical fractured rock masses under complex geological conditions.

The deformation prediction of fractured rock slopes is a key challenge in hydropower engineering due to the heterogeneity and stochastic distribution of fracture systems. Traditional approaches often struggle to identify dominant fracture parameters and maintain predictive accuracy while preserving geological interpretability. This study proposes a hybrid feature selection and prediction framework that integrates Discrete Fracture Network (DFN) modeling, Finite Difference Method (FDM), Global Sensitivity Analysis (GSA), and Extreme Gradient Boosting (XGBoost). Eighteen fracture-related geometric and mechanical parameters were initially considered, with maximum slope displacement from DFN-FDM simulations as the output. Sensitivity ranking was performed using Pearson correlation, Sobol variance-based index, and PAWN density-based method (Probabilistic Analysis of Whisker Numbers). The results indicate that 11 parameters dominate slope deformation, reflecting both geometric configuration and mechanical heterogeneity. Based on ranked features, XGBoost surrogate models were established and evaluated through 10-fold cross-validation. The GSA-XGBoost achieves the lowest error, outperforming both the original model and PCA-based (Principal Component Analysis) surrogate model, while reducing dimensionality and retaining geological interpretability. Application to a fractured slope at BDa Hydropower Station demonstrates the framework’s capacity for parameter prioritization and reliable deformation prediction. This approach provides practical guidance for slope stability evaluation and support design in fractured rock masses.

Landslide-generated waves involve large deformations of the landslide and a mixture of solid material and water, posing significant challenges for numerical simulation. This study proposes a smoothed particle hydrodynamics and discrete element method (SPH-DEM) coupling framework based on mixture theory to study water-soil coupling in landslide-generated waves. The water is modeled by the SPH method as a weakly compressible Newtonian fluid, while the soil landslide is directly simulated at the particle scale with DEM. The governing equations for the fluid phase are derived from mixture theory using the intrinsic fluid density. The model explicitly captures the spatiotemporal variation of fluid volume fraction resulting from landslide deformation and mixing with water. The delta-SPH formulation is incorporated to enhance the accuracy and stability of fluid simulations, and a U-tube seepage test is conducted to validate the model's reliability for seepage computations. The proposed method is further applied to three experiments on granular landslide-generated waves. The results accurately capture the evolution of the free water surface, landslide deformation, and the mixing between granular and water, while also revealing the evolution of the inter-particle interaction chains within the landslide during motion. Compared with previous methods, the proposed approach demonstrates improved accuracy in wave prediction and better performance in modeling landslide dynamics.

Uncertainty in rock rheological parameters is a pervasive issue in soft rock engineering and poses significant challenges for both laboratory-scale parameter calibration and engineering-scale application. This study proposes a Bayesian uncertainty estimation framework for rock rheological parameters based on the stable stage of laboratory step-loading creep tests and Markov Chain Monte Carlo (MCMC) sampling. The rheological parameters of the Burgers model were inferred using creep test data of calcareous mudstone from the Liujiacun section of the Dianzhong Water Diversion Project. The performance of the proposed framework was evaluated in terms of convergence behavior, posterior distributions, parameter correlations, and sensitivity characteristics, demonstrating its ability in quantifying parameter uncertainty. The inferred parameter distributions were then incorporated into finite element simulations to predict long-term tunnel deformation, including crown settlement and horizontal convergence. The predicted deformation envelopes show good agreement with field monitoring data within the 95

Landslide-generated impulse waves (LGIWs) are among the most catastrophic natural hazards, particularly in reservoir areas where they often cause devastating consequences. This study investigates the impacts of LGIWs through a combined experimental and numerical approach. A large-scale three-dimensional (3D) prototype physical model of the Meilishi (MLS) landslide in China's Gushui (GS) Reservoir is constructed at a scale of 1:150. Based on the physical experiment results, a 3D numerical model that couples the granular flow model and renormalisation group (RNG) turbulence model in FLOW-3D is established to analyse the generation and propagation of LGIWs in reservoirs. The numerical simulations show strong agreement with experimental data under identical conditions. Simulated wave amplitudes closely match the time-series data from the physical model, with relative errors below 15% in the wave generation and propagation areas and under 5% near the dam area. These results demonstrate the numerical model's capability to accurately reproduce the complex nonlinear processes of wave generation and propagation induced by subaerial landslides, particularly for large-scale, long-distance LGIW propagation analysis.

The impoundment of large reservoirs often reactivates ancient landslides, posing significant risks to infrastructure such as tunnels that traverse these unstable slopes. This study investigates the deformation behavior of the Dawanzi (DWZ) tunnel, which orthogonally passes through the Tuandigou (TDG) landslide in the Baihetan Reservoir Region, China. Field monitoring, numerical simulation, and theoretical analysis were employed to examine the interaction between the landslide and the tunnel during reservoir impounding. Results indicate that the TDG landslide, initially stable, experienced accelerated deformation when the reservoir water level exceeded 810 m, with cumulative surface displacements reaching 299.1–1146.8 mm by January 2023. Correspondingly, the tunnel section within the landslide (K8 + 479—K8 + 553) exhibited significant distress, including concrete cracking, spalling, and diameter expansion. Numerical simulations revealed a reduction in the landslide safety factor with rising water levels, and an elastic beam model was applied to analyze the tunnel-landslide interaction mechanism. To ensure long-term safety, a rerouting scheme for the affected tunnel section is proposed, bypassing the landslide area. The findings provide valuable insights for the design and risk mitigation of tunnels in reservoir-affected landslide regions.

This study presents a comprehensive numerical investigation of the bearing behavior of a novel high-pile cap foundation for offshore wind turbines. The foundation consists of a central monopile surrounded by six inclined piles and is subjected to complex loading conditions. Using Fast Lagrangian Analysis of Continua in Three Dimensions (FLAC3D) with advanced liner elements and interface modeling, the analysis incorporates realistic geological stratification and load combinations. Firstly, the load response and the bearing characteristics of pile foundation under a single working condition are examined. Then, six distinct working conditions are compared. The results show that horizontal and moment loads dominate the system response. The inclined piles act effectively as tension anchors, significantly reducing shear forces and bending moments in the monopile and improving the overall horizontal resistance. The concrete cap plays a critical role in stabilizing vertical displacements and redistributing moments. The maximum Mises stresses in the steel pile (214.18 MPa) remains well below the material yield strength, confirming the structural adequacy of the designed foundation system. The study provides validated numerical insights and practical design equations for optimizing similar high-pile cap foundations in offshore environments.

Rainfall infiltration plays a critical role in controlling pore pressure evolution and triggering instability in natural slopes composed of heterogeneous soil-rock deposits. A probabilistic framework combining interval approximation reformulation with adaptive Kriging-assisted Monte Carlo simulation is developed to assess rainfall infiltration-induced instability in deposit slopes. The adaptive Kriging is employed using a U-learning enrichment criterion to surrogate the limit-state response. The approach avoids the dimensional limitations associated with kernel density estimation in high-dimensional probability spaces. The model is applied to the Dahua deposit landslide located along the Lancang River in Yunnan Province, China. Five gradient rainfall scenarios are considered, and the surrogate framework is calibrated against a physical model test. The results show that compared with conventional Monte Carlo approaches, the proposed framework significantly reduces computational cost while preserving accuracy in probabilistic predictions. The interval approximative adaptive Kriging based on Monte Carlo method reproduces the experimental displacement response with a mean relative error of approximately 0.05%, where the simple interval adaptive method based on Monte Carlo method produces an error close to 3%. The computational time is reduced from nearly 18 h to approximately 2 h under single-thread execution on a standard processor. The probabilistic analysis further shows that the maximum failure probability increases monotonically with cumulative infiltration depth as the rainfall return period increases from 25 to 100 years. The study provides insights into uncertainty propagation in rainfall-driven fluid-solid systems and offers a practical tool for probabilistic assessment of infiltration slope instability.

The rock digital twin is a virtual system that continuously learns real-time observations of the physical entity, allowing for the dynamic updating of the mechanical parameters in the numerical model. This paper develops a digital twin framework based on data from triaxial rock mechanics test, numerical simulation, and data assimilation (DA) algorithm, focusing on the mechanical behavior of metamorphic silty sandstone from Mogu tilting deformation body under three different confining pressures. Mechanical tests on the rock samples were conducted and numerical simulation integrating the Iterative Local Updating Ensemble Smoothing (ILUES) algorithm was implemented. This integrated approach combined modondel with experimental data, achieved the iterative updating of parameters. The updated parameters were subsequently fed back into the numerical model to predict its response. We evaluated the performance of the digital twin model by assessing the mechanical parameters and model prediction. The findings of this research are of significant importance in reducing uncertainties associated with model parameters in rock materials, thereby enhancing our insight to evaluate the risk assessment of geotechnical engineering.