Multi-level railway stations are characterized by strong inter-level coupling, high passenger densities, and complex evacuation processes. To reduce the reliance on empirically specified parameters in large-scale crowd evacuation modeling, this study proposes a multi-objective parameter calibration framework and evaluates its applicability in multi-level railway station evacuation scenarios. Calibration objectives are formulated based on three typical sub-scenarios, namely corridors, bottlenecks, and staircases, and NSGA-III is employed to perform joint multi-objective calibration. Representative calibrated parameter sets selected from the Pareto front are then applied to four representative multi-level railway station configurations to examine performance consistency under controlled high-density conditions. In addition, a real-station evacuation application is conducted using observed peak-period density from a large railway station in Shenzhen, to construct the initial loading condition. The results show that the calibrated parameters consistently improve evacuation performance across the representative configurations while preserving the performance ranking primarily governed by spatial structure. In the real-station application, the calibrated parameter set also improves evacuation performance and cross-layer discharge under observed-density-based initial loading. These findings indicate that the proposed calibration framework enhances the reliability and applicability of large-scale evacuation simulation and supports comparative layout assessment, bottleneck identification, and safety-oriented design optimization in multi-level railway stations.

The growing demand for greener and smarter urban underground construction has driven the development of an internationally pioneering technique known as “Close-fit, twin tunnel pipe jacking technology for subway station construction”. This technique was first applied in the “Shasan Station”, the world’s first prefabricated metro station constructed by twin closely-spaced rectangular pipe-jacking boxes. This study presents the first investigation into the seismic performance of this novel structural system, focusing on its response characteristics under transverse seismic loading. A detailed three-dimensional (3D) numerical model is established. The numerical model has been validated by comparing its results with data obtained from full-scale joint tests and analytical solution of the wave theory. Using this model, a systematic comparison is conducted between the mechanical responses of the new prefabricated station and a conventional cast-in-place station. Furthermore, the interaction and load-transfer mechanisms between the column ring and the standard ring are investigated in detail. The results indicate that the twin closely-spaced rectangular pipe-jacking prefabricated station structure exhibits satisfactory overall seismic performance. The joints reduced the global stiffness, which altered the seismic force distribution and significantly decreased the internal forces. For instance, the bending moment at the column base was reduced by approximately 49%. The deformation of the joints and longitudinal seams under transverse seismic action is far below the design limits. The internal forces at the column ring are higher compared to those at the standard ring, but the deformation of the standard ring is greater than that of the column ring. This study offers practical insights and design guidance for the seismic design and optimization of similar prefabricated metro stations.

During shield tunnelling beneath existing twin-line tunnel structures, tunnel face instability and ground disturbance are key factors affecting construction safety and structural stability. Based on a series of controlled laboratory model tests, this study systematically investigates the effects of tunnel face position, structural spacing, and overburden depth on soil displacement fields, shear strain distribution, and surface settlement patterns. Digital image correlation (DIC) was employed to visualise the failure modes, and sensor data were integrated to analyse ground response characteristics under different working conditions. The results indicate that the tunnel face position governs the symmetry and propagation of failure zones, while structural spacing and overburden depth jointly influence the depth of the settlement trough and the magnitude of shear strain. Ground response curve (GRC) shows that the ultimate support pressure is significantly higher when the tunnel face is located between the twin tunnels than when it is directly beneath one of the tunnels, with increases of approximately 37 and 73 %, respectively. Regarding surface settlement, when the tunnel face is positioned beneath an existing tunnel, the peak of the settlement trough shifts forward owing to the shadowing effect, with the peak settlement location moving by approximately 0.53D times the tunnel diameter. The results also show that voids are prone to form beneath the existing tunnels during undercrossing, which further increases the risk of structural instability. These findings provide a theoretical basis and experimental reference for safety assessment and control strategies during shield tunnelling in complex urban environments.

When shield tunneling crosses beneath existing underground rectangular structures, tunnel face instability has a significant impact on construction safety and ground stability. However, current research offers an insufficient understanding of the instability mechanisms and influencing factors, especially under the constraints imposed by existing structures. To address this gap, this study conducts laboratory model tests to investigate the failure mechanism of tunnel face during shield tunneling beneath existing underground rectangular structure. Stress and displacement sensors are used to monitor changes in earth pressure and surface settlement, while the Digital Image Correlation (DIC) technique is applied to analyze displacement fields and shear strain distributions in the soil. The results indicate that, during the tunneling process, the active failure of the tunnel face is primarily influenced by the position of the existing structure, the spacing between structures, and the thickness of the overburden. The soil arching effect gradually forms in the early stage of excavation and provides a stabilizing effect on the surrounding soil within a certain range. In particular, the limit support pressure was about 0.35P0 at the left side, 0.20P0 at the centerline, and 0.15P0 at the right side beneath the rectangular structure (P0 = initial earth pressure acting on the tunnel face), attained at a normalized tunnel face retreat of about 0.02D (D = tunnel diameter). However, as excavation displacement increases, the arching effect weakens, leading to an expansion of the instability zone. Furthermore, surface settlement and earth pressure distribution exhibit nonlinear variations, and are significantly affected by the constraints of the existing structures. The findings of this study provide important references for ensuring the safety of shield tunneling beneath underground structures and for controlling ground stability.

The construction of large-diameter shield tunnels in soft soil under shallow overburden and ultra-close spacing (0.7 D) poses significant challenges to ground and existing tunnel (ET) stability. This study systematically investigates the interactions between subsequent tunneling (ST), ground response, and existing tunnel behavior across three distinct ground conditions—Full Reinforcement (FR), Grid Reinforcement (GR), and Unreinforced (UR)—based on comprehensive field monitoring from the Hengqin Mangzhou Tunnel project. A multi-method monitoring system was employed to track ST operational parameters, ground deformation, and ET segment convergence and internal forces. The results reveal that shield parameters exhibited pronounced fluctuations at the GR-UR transition zone, identified as a high-risk area due to the severe stiffness contrast, with strength of UR soil approximately 80 to 200 times lower than GR soil. Unstable grout mixtures—characterized by overfill ratios of 1.8–2.4 times and a 1-day strength as low as 0.25 MPa—combined with the GR-UR transition zone, caused excessive segment flotation and extreme ground heave up to 84.7 mm in the GR zone, while the UR zone experienced progressive settlement of 121.35 mm. The ET was displaced away from the ST in reinforced zones but moved toward it in the UR zone, with a critical spacing threshold of approximately 1.2 D at the GR-UR interface. Ground reinforcement reduced horizontal convergence by approximately 63 %, with stabilized convergence accounting for 22–24 % of the peak in the GR zone versus 45 % in the UR zone. These findings provide practical references for deformation control and grout volume management in similar soft ground tunneling projects with reinforcement transitions.

In pipe jacking with quasi-rectangular tunnel boring machines (RTBMs), optimizing muck flowability is crucial for minimizing surface settlement and reducing jacking resistance. While both static (fixed) and dynamic (revolving) stirring rods are employed to enhance muck flow, their comparative effectiveness and underlying mechanisms remain inadequately understood. This study employs a one-way discrete element method–finite element method (DEM-FEM) coupling approach to numerically investigate the mechanical responses of the surrounding stratum and the RTBM structure under these two stirring rod configurations. The results demonstrate that static stirring rods contribute to better control of surface settlement during the machine starting phase by locally increasing the earth pressure at the tunneling face. In contrast, dynamic stirring rods significantly improve muck discharge efficiency and reduce long-term thrust resistance during continuous operation. Based on these findings, a strategic deployment of both rod types during different construction phases to optimize overall performance is proposed. This research provides practical insights for the design and operation of stirring mechanisms in large-section pipe-jacking projects, aiming to balance settlement control with tunneling efficiency.

With shield tunnelling rapidly progressing toward extra-large diameters, greater burial depths, and special-shaped cross-sections, accurate bearing-capacity evaluation of complex tunnel linings has become a key issue in tunnel engineering. Full-scale tests are among the most reliable approaches for revealing the true mechanical behaviour of tunnel linings and validating design theories. However, conventional reaction frames usually feature fixed configurations and lack modular reconfigurability, thereby limiting their dimensional and cross-sectional adaptability in tests involving different tunnel geometries. To address this limitation, this study develops a modular full-scale test system for shield tunnel linings based on a prefabricated steel reaction frame. The system can be flexibly assembled to accommodate different cross-sectional profiles, including circular tunnel linings up to 15 m in diameter. The flange-joint and nested-column designs improve assembly efficiency, while the prestressed steel-strand restraint system enhances the load-bearing capacity of the reaction frame. Numerical analyses and a full-scale test were conducted to evaluate the mechanical performance and practical feasibility of the system. The numerical results indicate that the system can satisfy the loading requirements for 15-m-class circular and quasi-rectangular tunnel linings, corresponding to equivalent burial depths of 45 and 35 m, respectively. A full-scale test conducted on a 6.7-m-diameter shield tunnel lining at a burial depth of 35.7 m and a lateral earth pressure coefficient of 0.81 showed that the maximum deformation convergence of the reaction frame was 4.5 mm, while the maximum stress in key components reached only 18% of the material yield strength. The proposed test system improves the dimensional and cross-sectional adaptability of full-scale shield tunnel lining tests and provides a practical experimental platform for future mechanical investigations and design validation of shield tunnel linings.

Revealing the evolution mechanism of soil arching above the trapdoor under high geostress conditions and establishing an accurate predictive model for deep earth pressure are crucial for ensuring the safety of underground engineering and achieving design optimization. This study investigated the deformation and failure patterns of soil under high geostress using a self-developed trapdoor model testing apparatus. Subsequently, the particle flow code (PFC) discrete element method was employed to reveal the mesoscopic characteristics of particle movement restrained under high geostress and to clarify the influence of dilatancy on the evolution mechanism of soil arching. Based on the results of the model tests and numerical simulations, a theoretical model composed of a self-weight stress zone, a parabolic end-bearing arch zone, and a loosened zone was proposed to predict the variation of earth pressure above the trapdoor under high geostress conditions. The results indicate that: (1) the high geostress environment provides strong lateral confinement, which restricts the lateral expansion and horizontal propagation of the loosened soil, resulting in a parabolic development pattern of the loosening zone; (2) under high geostress conditions, the ground reaction curve exhibits a three-stage evolution pattern characterized by a rapid decrease, followed by a gradual reduction, and finally stabilization; (3) the increase in initial stress level restrains particle movement and dilatancy, leading to a delayed formation of soil arching and a reduction in its early load-bearing capacity; and (4) the theoretical calculation results are in good agreement with the model test and discrete element simulation results, demonstrating that the proposed model can accurately capture the variation trend of overburden earth pressure on the trapdoor under high geostress levels. The findings provide theoretical support for the design optimization and safe construction of deep underground engineering.

Further investigation into the evolution of soil arching influenced by initial anisotropy remains essential. In this study, the discrete element method is employed to analyze trapdoor models with varying degrees of initial anisotropy, which are calibrated using data from previous experimental studies. The results indicated that the overall trend of the ground reaction curves remains consistent across samples with varying initial bedding angles, with three distinct stages. The variation in the ultimate soil arching ratio closely parallels that of the minimum ratio, also exhibiting an approximately symmetric "M"-shaped pattern, with the minimum value occurring at theta = 45 degrees. Compared with isotropic spherical materials, anisotropic materials show a weaker soil arching effect. The evolutionary pathways of soil arching can be categorized into two distinct patterns: the quasi-symmetrical arch and the deflection arch. The dominant direction of force chains progressively aligns with the imposed bedding angle. Increasing the initial bedding angle leads to progressively more concentrated contact force distributions and increasingly robust force chain structures. The distribution of normal forces, tangential forces, and contact number is strongly influenced by the initial bedding angle.

This paper presents the results of a study that investigated the lateral response of open-ended pipe piles in sandy soil profiles based on a series of three-dimensional (3D) finite-element (FE) analyses. The analyses, which used an advanced two-surface plasticity constitutive model, were performed for anisotropic sandy soil deposits with varying values of relative density and at-rest lateral effective stress ratio K-0. The inclusion of both D-R and K-0 into predictive relationships leads to more accurate estimation of pile lateral load and pile head motion, which in turn makes it possible to achieve more economic foundation designs. In addition, we investigated the effect of load eccentricity on the lateral load response of pipe piles. A generalized method to estimate the full lateral load response of pipe piles in layered soil is presented. The proposed method incorporates the effect of horizontal effective stress, and is applicable to a wide range of pile geometries, sand relative density, and any load eccentricity. The method fills a gap in knowledge left by studies that primarily focused on either very large-diameter, rigid pipe piles, such as monopiles, or a narrow range of soil conditions. Validation was done by comparison to field results: close agreement was obtained between the predicted and measured pile responses obtained as part of the Pile Soil Analysis (PISA) project. It is demonstrated that the method eliminates the need for independent determination of in situ horizontal effective stress or relative density if cone resistance is available, because these two parameters can be coupled through an equation used to calculate cone resistance; cone resistance can then be used directly in calculations. This is particularly useful given the challenges of measuring K-0 for sand in practice.

Accurately predicting the tunnelling-induced axial response of adjacent piles remains a critical challenge in urban underground construction. Although conventional two-stage analysis methods are widely adopted, the mutual influences between tunnel-soil and soil-pile interactions are neglected, often resulting in unconservative predictions. To address this limitation, this paper proposes a coupled tunnel-soil-pile interaction model that explicitly incorporates cross-interaction effects within a unified analytical framework. Governing equations for the axial response of pile foundations are derived by coupling the stiffness equations of the tunnel, soil, and pile, while ensuring force equilibrium and displacement compatibility. The proposed model degenerates to the conventional two-stage method as the tunnel stiffness approaches zero, whereas an infinitely rigid tunnel yields an upper-bound response, thereby establishing a unified and generalizable theoretical framework. For the solution, a regularized Mindlin solution based on elastic continuum theory is employed to construct the soil flexibility matrix, which eliminates the source-point singularity while preserving far-field accuracy, thus enhancing numerical stability and efficiency. The method is validated through comparisons with existing analytical solutions, centrifuge model tests, and finite difference simulations incorporating HS-Small constitutive model, all of which demonstrate excellent agreement. Parametric studies reveal that neglecting cross-interaction for piles in close proximity to the tunnel underestimates the maximum axial force by approximately 20–40%. Three critical influencing parameters are identified. Based on pile length Lp, pile-to-tunnel distance Dp, tunnel depth H, and tunnel diameter Dt (where Dt=2R), a screening chart is proposed to determine the necessity of the coupled model: specifically, only when the pile lies in the crown zone with Lp > 0.8 (H−R) or in the near-field zone with Lp > H. The tensile forces induced above the tunnel crown are analyzed, with comparisons of axial force ratios and magnitudes obtained from both methods under typical scenarios. The proposed method accurately captures the tunnelling-induced mechanical response of piles, offering a physically meaningful tool for safety assessment.

Understanding soil arching and failure modes of ground subsidence above buried pipelines is crucial for developing effective mitigation strategies to reduce sinkhole hazards. Although previous studies have examined soil arching using discrete or continuum methods, the soil-pipeline interaction under localized subsidence remains insufficiently quantified. To address this gap, the present study employs a discrete-continuum coupled (DEM-FEM) trapdoor model, which was quantitatively validated against previous trapdoor tests. Using this validated framework, the influence of pipeline diameter (D) on the evolution of soil arching was systematically analyzed for five D/B ratios ranging from 0.5 to 1.5. The results reveal that increasing the pipeline diameter broadens the disturbed zone and induces three distinct soil-arching evolution patterns: closure type, parallel open-ended type, and divergent open-ended type. A critical transition occurs at D/B = 1.0, corresponding to the minimum soil-arching ratio, beyond which arching efficiency improves. Larger trapdoor displacements reduce the magnitude and spatial extent of high-pressure zones, while varying D/B produces staged pressure redistribution and anisotropic stress transfer above the pipeline. Moreover, both the coordination number and fabric anisotropy analyses highlight microstructural degradation and directional instability with increasing pipeline diameter, revealing multiscale mechanisms governing stress redistribution and soil-structure interaction that have not been previously reported in trapdoor studies.

Using a self-developed test platform, indoor model tests were conducted on shield tunnels in complex sandstone strata with different cutter track layouts. These tests revealed rock damage morphology, macroscopic wear characteristics and microstructural features examined by FE-SEM. Disc cutters wear was quantified by mass loss measurement through cutting length, and cutterhead vibration was recorded using a wireless three-axis sensor. The study investigated disc cutters wear and vibration responses under soft rock (20 MPa), hard rock (60 MPa), composite rock (20 MPa upper / 60 MPa lower), rotational speeds (18, 24, 30 rpm), and advance speeds (0.01, 0.02, 0.03 mm/s), and discussed their correlation. The results show that rock failure is divided into four stages by advance distance: powder generation (0–2 mm), dense core formation (2–4 mm), small block detachment (4–6 mm), and large block detachment (6–10 mm). Macroscopic wear is asymmetric with edge rounding and furrows on the outer side, while microscopic examination confirms two-body and three-body abrasion as the dominant mechanism. In composite rock, powder generation depends on the soft layer while the breaking load is governed by the hard layer; unit wear (4.52 mg/m) lies between soft (0.81 mg/m) and hard rock (6.50 mg/m) but closer to the hard rock value, and vibration exceeds that of either homogeneous rock. Wear growth accelerates beyond 24 rpm, while vibration growth diminishes. The wear-vibration relationship is dominated by the circumferential and radial axes (>84% contribution) and is condition-dependent, with slopes varying by approximately 6.4 times across rotational speeds: low strength with low speed minimizes both, hard rock exhibits severe wear without proportional vibration escalation, and high speed drives strong vibration even under moderate rock strength.

Accurately predicting fire temperature fields is crucial for building safety and emergency response. However, trained data-driven models often struggle to generalize across scenarios with different geometries, fire locations, and ventilation conditions. To overcome this challenge of generalization across domains, this paper presents a domain-adaptive learning framework, FPDA-LSTM-ResLap, which integrates adversarial training with explicit physical priors. This method instructs the model to learn domain-invariant spatio-temporal features while considering physical mechanisms such as distance attenuation and obstacle effects. When evaluated across ten diverse fire scenarios and a real-world case study of Shenzhen North Station, the framework demonstrated superior performance, outperforming supervised and transfer learning baselines in key metrics. The framework reliably predicts unseen, complex environments without retraining, showcasing its practical applicability to domain shift. This work provides an effective solution for predicting fire temperature fields across scenarios, enhancing the deployment potential of data-driven models in intelligent fire safety systems.

Real-time reconstruction of high-resolution fire temperature fields remains a major challenge in building fire safety. While many data-driven methods have emerged, they predominantly predict temperature at the ceiling level, neglecting the safety-critical height (e.g., 1.5-1.8 m) that is paramount for occupant safety and evacuation planning. To bridge this gap, a cascaded deep-learning framework that synergizes temporal prediction with spatial super-resolution is proposed. First, the model employs a bidirectional gated recurrent unit (BiGRU) to infer cross-layer temperature trends from sparse ceiling sensors down to the occupant level. Then, it applies a residual network (ResNet) to reconstruct high-resolution temperature fields at this height. When evaluated on a nuclear power plant fire dataset and a large-scale transportation hub scenario (Shenzhen North Station) that was not included in the training dataset, the proposed framework yielded predictions that closely match CFD benchmarks, while its inference speed surpasses that of conventional CFD simulation by orders of magnitude. Furthermore, the method maintains stable performance under sensor failure conditions, with only marginal degradation observed when multiple sensors are inactive. This work offers an accurate, efficient, and transferable solution for building fire digital twins, enabling real-time safety assessments and evacuation guidance at heights relevant to humans.

With urbanization and expanding transport hubs, Station-City Integrated Spaces (SCIS) that combine transportation, commerce, and residence are increasingly common. However, their complex structures and high pedestrian density pose major fire safety challenges. This study develops a resilience assessment framework that integrates spatial and behavioral dimensions to improve safety management in SCIS. The framework consists of resilience indicator identification, integrated weighting, and sensitivity-based optimization. It is validated through five representative SCIS case studies. The results identified key resilience factors, including reinforcement of critical risk nodes, multi-agency coordination capacity, frequency of fire safety training, facility maintenance frequency, and spatial perception and monitoring capability. Finally, sensitivity analysis guided the development of targeted strategies for spatial optimization and behavioral management. This framework is a decision-support model for management-oriented screening and comparative prioritization of fire resilience in SCIS. It has broad applicability and provides both theoretical support and practical guidance for enhancing fire resilience across diverse station-city integrated network contexts.

Chinese high-speed rail (HSR) hubs commonly operate through centralized waiting halls, gate-specific boarding, and staged access control. Under this operating mode, pulse-surge passenger demand interacts with fixed spatial layouts and sequential bottlenecks, often generating severe congestion, queue spillback, and inefficient use of available space. Existing microscopic pedestrian simulation models can reproduce local movement dynamics and collision avoidance, but they generally provide limited support for explicitly representing the operational rules that govern passenger organization in such facilities, including ticket–gate bindings, queue eligibility, and train-specific boarding windows. To address this limitation, this study proposes a Semantic Digital Twin (SDT) framework for rule-constrained HSR hub operations. The framework integrates explicit operational knowledge representation, finite-state rule logic, and microscopic pedestrian dynamics within a unified computational environment. Using Shenzhen North Railway Station as a real-world testbed, the study evaluates dynamic routing, capacity-aware staggered release, and their coordinated application under an empirical peak-demand scenario. The results show that the integrated strategy eliminated observed stationary spillback outside the designated boarding area in the tested scenario and reduced internal frictional stagnation by 75.2 %, at the cost of a modest 2.3 % increase in full-clearance makespan. These findings demonstrate that explicit representation and computational execution of operational knowledge can substantially enhance decision support for fixed-layout, rule-constrained transport systems, extending digital twins from descriptive asset mirroring toward knowledge-driven operational control.

The finite-discrete element method (FDEM) has become an increasingly valuable tool in modelling rock mechanics and microscopic fracturing behaviours in recent years. However, a reliable calibration process for 3D FDEM models, particularly for anisotropic shale, remains lacking, hindering the efficient and accurate determination of microscopic input parameters. To overcome these challenges, this study introduces a novel calibration framework for 3D FDEM anisotropic shale models by systematically investigating the sensitivity of four key fracture energy parameters: fracture energy along bedding in Mode I ( G_I^f ), fracture energy normal to bedding in Mode I ( G_I^f' ), fracture energy along bedding in Mode II ( G_II^f ), and fracture energy normal to the bedding in Mode II ( G_II^f' ). The sensitivity analysis reveals that tensile strength in shale specimens inclined at 45° and 90° is primarily governed by G_I^f , while for specimens inclined at 0°, G_I^f' plays a dominant role. In uniaxial compression, G_II^f significantly influences the compressive strength of shale at 0° and 45° inclinations, whereas G_II^f' is more influential at 90°. The study also shows that Young’s modulus is independent from fracture energy. The novel 3D FDEM anisotropic shale model demonstrates strong agreement with experimental results, outperforming 2D simulations in replicating the nonlinear deformation and plastic behaviour of shale under uniaxial compression. Moreover, 3D fracture volume analysis reveals bedding-dependent secondary fracture patterns: shear-induced tensile fractures develop orthogonal to bedding-parallel shear planes at 45°, while combined principal stress and shear-induced tensile stresses generate oblique secondary fractures at 0°.

Compared with small and medium-diameter tunnels, super-large-diameter shallow-buried shield tunnels experience considerably amplified construction disturbance effects. This amplification leads to more pronounced longitudinal deformation of the tunnel lining and an increased risk of structural damage. Therefore, during tunnel construction, achieving advanced prediction and warning of longitudinal deformation is crucial for ensuring structural safety. On the basis of the Zhuhai Xingye Expressway Project, the longitudinal deformation of a super-large-diameter shallow-buried shield tunnel is predicted in this paper by employing a physics-informed neural network (PINN) algorithm. First, following data cleaning, a dataset was constructed using feature variables selected on the basis of correlation analysis. The longitudinal deformation of the tunnel was subsequently analogized to beam deflection, and a PINN model was developed by incorporating physical information derived from a longitudinal continuous equivalent model based on Timoshenko beam theory. Finally, for comparative analysis, five additional algorithms, namely RF, GRNN, SVR, DNN, and LSTM, were used to establish corresponding predictive models for longitudinal deformation. The results indicate that in terms of generalizability of the test set results, the performance ranking is PINN > DNN > RF > GRNN > SVR > LSTM. The PINN model outperformed all other algorithms, with an RMSE for the test set of 6.01 mm. These findings demonstrate that the PINN model offers high interpretability and strong generalizability, making it suitable for predicting longitudinal deformation in super-large-diameter shallow-buried shield tunnels. These findings provide a meaningful reference for improving construction safety in shield tunnels.

The crucial role of water in soil arching evolution remains poorly understood. This study investigates the development of soil arching with the coupled Smoothed Particle Hydrodynamics-Discrete Element Method (SPH-DEM) with consideration of submerged condition. The method was validated based on previous experimental results. The results showed that the rising water levels attenuate the soil arching effect during trapdoor displacement, increasing the minimum arching ratio compared to dry conditions. Deformation transitions from localized inverted V-shaped shear bands (dry) to global deformation with broad horizontal bands (saturated). Fluid forces concentrate near the water table at low levels but distribute more evenly with higher saturation, while contact forces decrease. The Coordination number decreases with trapdoor displacement, but water level strongly affects the rate of decline. With respect to the anisotropy degree, all hydraulic conditions exhibit a general increasing trend as trapdoor displacement advances. This study provides novel insights into the impact of submerged condition on soil arching from macro to micro perspective.