Deformation prediction during pre-support tunnel construction remains a major challenge: conventional methods lack adaptability, while deep learning models often sacrifice physical interpretability for accuracy. This study addresses the research question of how to achieve predictions that are both accurate and physically consistent under dynamic tunnel construction. We propose a hybrid physics–data prediction approach (HPDPA) that integrates a physical–mathematical model grounded in Winkler elastic foundation beam theory with a physics-informed attention long short-term memory (PI-ALSTM) network. Through a bidirectional collaboration mechanism, the PI-ALSTM supplies data-driven variables to refine the physical–mathematical model, while the latter informs the learning process of the data-driven model, ensuring consistency with physical mechanics and adaptability to the updated site-specific conditions. When validated on real-world pipe-roof tunneling data, HPDPA achieves high predictive precision (MAE = 1.11 mm, RMSE = 0.97 mm, MRE = 0.07) and markedly outperforms purely data-driven baselines. These results demonstrate the framework’s capability to provide physically trustworthy, high-resolution, and real-time guidance for tunnel construction, thereby contributing to the digitalization process of risk management in pre-support tunnel projects.

T-joints represent a critical yet distinct structural feature in shield tunnels, exhibiting fundamentally different sealant performance and leakage behaviours compared to continuous joints. This study presents a novel pathfinding algorithm-based approach integrating contact stress analysis to evaluate localised sealant behaviour and leakage mechanisms affected by T-joints under various deformation patterns. The approach enables simultaneous assessment of waterproof capacity and identification of potential leakage paths. The effectiveness of the proposed approach is validated by comparing with laboratory testing data. It is further demonstrated that the waterproof capacity of T-joints can be effectively characterised by an ellipsoid function accounting for joint offset and openings in both circumferential and longitudinal directions. Joint offset substantially reduces waterproof capacity and creates localised lift-up areas with critically low contact stress due to gasket-to-groove interaction. Longitudinal joint openings exhibit less impact on waterproofing performance compared to circumferential openings, with residual sealing capacity maintained even at full opening displacement. A newly proposed leakage centralisation ratio reveals four distinct leakage regions, whose formation depends on joint-specific contact behaviour governed by opening and offset parameters. These insights provide a quantitative basis for optimising T-joint design in shield tunnels, with particular relevance for waterproofing system performance under complex loading conditions.

Rockbursts in deep tunnels traversing stratified rock under impact loading pose significant safety challenges. The split Hopkinson pressure bar experimental system was combined with finite element modeling in this study to explore the effects of weak interlayer position and crustal stress on the rockburst. Three simplified mechanical models with varying interlayer configurations were subjected to impact loading under different crustal stresses, and failure processes were captured using high-speed photography and digital image correlation. Numerical simulations were validated experimentally and provided insights into stress and energy dynamics. It was revealed that the position of the weak interlayer governs the location and intensity of the rockburst. The lateral weak interlayers increase the brittleness index at the tunnel shoulder and foot by 197.75% and 143.40%. A nonlinear influence of crustal stress on damage was observed, accompanied by strain hardening effects.

Aiming at the problem of predicting the displacement impact on an existing metro station adjacent to the new foundation pit, this paper establishes a quantitative coupling relationship among the support displacement of the foundation pit, the metro station displacement, and the lateral earth pressure by equating the finite width soil (abbreviated as finite soil) between the foundation pit and the station to horizontally independent soil springs. A trigonometric function model is employed to describe the displacement-dependent lateral earth pressure, and the hyperbolic tangent function is introduced to achieve a smooth transition between ultimate and non-ultimate lateral earth pressure states. Based on Euler–Bernoulli beam theory, a closed-form solution framework is constructed, leading to the development of a theoretical model for displacement response that considers the dynamic structural interaction effects between the new foundation pit and the existing metro station. This model overcomes the limitation of the traditional two-stage analysis method, which neglects the bidirectional interactions between structures. Model validation shows the theoretical maximum horizontal displacement (d_max) of the station under various conditions aligns closely with simulated values, with a maximum difference of 8.30 mm. Under actual conditions, the deviation between theoretical calculations and literature data for d_max is only 1.11 mm. Theoretical solutions for lateral earth pressure also match literature data well under both finite and infinite soil conditions. Parameter analysis indicates that increasing the finite soil width can effectively mitigate the impact on the metro station. A greater station cover depth facilitates a transition of the lateral earth pressure from the active state of finite soil to the passive state of infinite soil, ultimately driving d_max toward zero. Furthermore, an increase in the soil-structure friction angle enhances horizontal friction at the station boundary, leading to a monotonic decrease in d_max.

The exploitation of essential mineral resources is crucial to human progress and the development of modern civilization. As surface and shallow-depth mineral reserves have been extracted or depleted, the mining industry has been increasingly moving to deep and ultra-deep operations, facing unprecedented geological and engineering challenges. This study provides a detailed evaluation of the key challenges associated with deep mining, including high water pressure, rockbursts, squeezing ground, heat, transportation, and the in-situ stress field. The combination of these risks not only undermines the stability of the structures in depth and increases the likelihood of accidents, but also creates a necessary demand for the use of more adaptable, advanced mining methods and ground support systems for risk management in deep mining operations. This paper also reviewed innovative operational solutions, including smart technologies and automation, ventilation and cooling systems, as well as alternative rock-cutting methods and tools. Furthermore, the paper summarizes various mining and design strategies, and it is emphasized that conventional theories of rock mechanics are inadequate for explaining the intricate behaviors of rock masses at depth, requiring new adaptive strategies for deep and ultra-deep mining. Ultimately, this paper advocates for a multidisciplinary and innovation-driven approach to enhance safety, sustainability, and efficiency in deep mining operations by examining several global case studies and their corresponding specifications.

This study introduces a shield attitude control method aimed at improving the accuracy and stability of tunnel boring machine (TBM) in synchronous excavation and segment assembly (S-TBM) along the design tunnel axis. The proposed hybrid deep learning model integrates bidirectional long short-term memory (BiLSTM) and Kolmogorov–Arnold network (KAN) cells to accurately predict S-TBM attitude. This model is further combined with the multi-objective grey wolf optimizer (MOGWO) method for determining the optimal proportional-integral-derivative (PID) parameters, thereby minimizing attitude deviations. The experimental results indicate that: (1) The S-TBM reduces the construction time per single lining ring to 39 min, which is 34 min faster than the conventional TBM. (2) The improved BiLSTM model can effectively estimate shield attitude, with an RMSE of 1.030, an MAE of 0.695, and an R² score of 0.998. (3) The proposed multi-objective optimization (MOO) framework is able to reliably control S-TBM attitude, achieving an overall improvement percentage of 25.66%. (4) Compared to other methods, the proposed approach enhances R² score by 0.03 and the overall improvement percentage by 5.72%. The novelty of this research lies in the development of an intelligent control framework for shield attitude under the synchronous excavation and segment assembly construction mode, which achieves reliable prediction and precise regulation of S-TBM attitude by synergistically coupling a nonlinear attitude response model based on an improved BiLSTM with an MOGWO-driven PID parameter optimization strategy.

Microseismic signals in complex geological environments are frequently contaminated by multiple non-stationary noise. Current denoising models often struggle to effectively balance noise suppression with accurate signal preservation, compromising the reliability of early warning systems. A conditional denoising U-Net (CDU-Net) model is proposed in this study to address this challenge. The model augments the U-Net backbone with a Transformer block to capture long-range temporal dependencies and a feature-wise linear modulation (FiLM) layer to adaptively adjust feature representations in different signal-to-noise ratio (SNR) environments. Microseismic monitoring data obtained through field measurements at the Hanjiang-to-Weihe River Diversion Project were employed to establish the research dataset. The denoising performances of five methods, wavelet transform (WT), variational mode decomposition (VMD), convolutional neural network (CNN), U-Net, and CDU-Net, are systematically compared from two perspectives: quantitative evaluation metrics and qualitative analysis. Significant advantages were demonstrated by the CDU-Net model across all quantitative evaluation indicators and waveform reconstruction quality, as shown by the experimental results. Particularly excellent performance was achieved, with the peak signal-to-noise ratio (PSNR) exceeding 30 dB and the structural similarity index measure (SSIM) reaching 0.82. At the same time, the denoising performance of the CDU-Net model was verified through visual analysis. Furthermore, ablation studies further confirm the complementary benefits of the U-Net backbone, Transformer, and FiLM components, reducing loss and synergistically improving output SNR. This study provides a solid theoretical foundation and reliable technical support for the long-term stable identification and high-precision early warning of microseismic signals in deep engineering.

Ground-penetrating radar (GPR) has been widely used to detect concealed defects behind tunnel linings. However, the limited penetration capability of traditional GPR in complex media such as concrete constrains tunnel-lining inspection. This limitation makes it difficult to obtain clear images of deeper structures. Furthermore, achieving rapid, multi-directional, and multi-angle inspections with traditional methods is challenging, especially in operational tunnels. In this paper, a vehicle-mounted multi-antenna synthesized ultra-wideband (UWB) GPR system is designed for tunnel inspection. First, we introduce a synthesized UWB signal design method that uses a field-programmable gate array (FPGA) to dynamically generate transmit signals from 0.3 to 1.0 GHz. This approach balances inspection depth and resolution. Second, we propose a delayed phase-locked synchronization technique to achieve high-precision time synchronization between the transmitter and receiver. Additionally, we design a semi-elliptical antipodal Vivaldi antenna array that outperforms conventional designs and obtains higher-quality inspection data in real tunnel environments. We systematically verified the feasibility and effectiveness of the proposed system through four experiments, including reflection analysis of multi-material targets, non-contact detection of buried objects in sandy media, dynamic scanning-based imaging in tunnel structures, and identification and imaging of typical defects in complex tunnel environments. The experimental results demonstrate that the system can effectively detect various types of subsurface targets under non-contact conditions. In the future, the system is expected to provide an efficient solution for multi-directional, multi-angle, deep, and rapid tunnel defect detection, particularly in complex engineering scenarios such as operational tunnels.

Thermal cracks significantly affect the physical and mechanical properties of granite, with implications for underground engineering and energy exploitation. This study combines macroscopic fluorescence visualization with image processing, innovatively integrating Otsu and Bradley threshold segmentation to achieve large-scale, accurate extraction and parameterization of cracks. Using the extensive database of thermal crack parameters obtained, we analyzed the statistical characteristics of thermal cracks at different temperatures (25–500 °C). The results reveal that thermal cracks undergo a ‘‘coalescence-rupture” cycle as temperature increases, reflecting the accumulation and release dynamics of thermal stress. Crack length distributions exhibit log-normal behavior, highlighting nonlinear evolution and the emergence of long-tail features at intermediate temperatures. Crack orientation shows limited temperature dependence and is mainly controlled by the intrinsic fabric and mineralogy of the granite. The orientation distributions are broadly consistent across temperatures with only slight peak fluctuations, and are well described by statistically robust multimodal normal fits. These findings provide valuable insights into the thermal behavior of granite and offer robust statistical parameters for numerical modeling and engineering applications. These results provide statistically parameterized inputs that help bridge laboratory observations and engineering-scale thermo-mechanical assessments, supporting safer and more efficient design and operation of high-temperature underground energy systems and subsurface infrastructure.

Metro-led underground spaces (MUS) have gained significant importance in addressing deteriorating urban issues in high-density built environments. However, existing planning techniques for MUS lack could enable the increasingly complex spatial morphology and function assignment, resulting in poor performance of MUS development in an unintegrated manner. To bridge the research gap, an enhanced layout planning approach for MUS (ELPA-MUS) was systematically formulated. ELPA-MUS incorporated a digital interpretation framework for MUS layout, enabling simultaneous analysis of spatial morphology and function. The model transformed the layout planning task into a multi-objective optimization (MOO) problem with nine objective functions. The non-dominant sorting genetic algorithm III (NSGA-III) was employed to find the Pareto front in high dimensions. To enhance the practicality of ELPA-MUS, an ensemble method was proposed, combining subjective expertise and objective computational analytics. The model was applied to a case study in Jinan, China to demonstrate its applicability and rationality. Overall, the ELPA-MUS model provided a modifiable paradigm for intelligent layout planning of complex underground spaces and expanded the data-driven planning toolkits towards a more sustainable MUS development.

The stress release factor with the range [0, 1] is employed to reflect the stress release degree of the tunnel excavation. Elastic analytical solutions of stratum stress and displacement are obtained considering the influence of the stress release degree of the tunnel excavation based on the complex variable method. An intelligent method for determining the stress release factor is proposed through a machine learning (ML) model. Random forest (RF) is chosen as the ML model, and the sparrow search algorithm is utilized to optimize the hyperparameters of RF. The rationality of the proposed analytical method is validated by means of calculation results of FEM and field monitoring data. The influence of stress release factor and tunnel geometric parameters on the distribution laws of the stratum tensile zone and plastic zone is analyzed. The influence of tunnel stress release on stratum displacement and tunnel cross-section deformation is discussed. A dual-driven framework for the intelligent prediction of stratum displacement based on the developed analytical method is proposed and conceptually formulated. The proposed method can provide guidance in the conceptual stage of the design process of shallow tunnels.

This paper presents a state-of-the-art review on the recent theories and models for the coupled heat and salt transport, i.e., thermo-haline transport (THT), in engineered geomaterials and subsurface systems. The primary progress in simulating coupled THT in soils under both positive and subzero temperature conditions is discussed, with the critical role of water phase transitions in driving THT emphasized. Furthermore, existing theories and models for describing THT in both rock matrices and fractures are systematically categorized, highlighting the significant influence of fractures on THT, which, however, has been less discussed before. The potential applications of THT in various engineering are outlined, clarifying the pivotal role of THT mechanisms in addressing challenges across these engineering. Finally, this paper identifies key future research directions for advancing coupled THT theories to resolve critical geo-engineering challenges, including the influence of porosity variations on THT in geomaterials, fracture-dominated THT interactions, the impact of THT on the efficiency of subsurface operations associated with salt transfer, and the need for conducting experimental studies to validate and improve the performance and accuracy of the existing THT models.

During the construction of new Austrian tunnelling method (NATM) tunnels, inherent multi-source uncertainties may lead to substantial economic losses and project delays. To address this issue, a simulation system enabled by digital twin (DT) technology was developed. Guided by the concept of five-dimensional DT, a four-layer system architecture was established, including perception, transmission, modelling, and alternation layers. An integrated discrete event simulation (DES) model was then developed as the core of the system using structured system analysis. The model accounts for both geological and construction-related uncertainties, including routine delays and accident recovery durations. Geological uncertainties are represented through a stochastic geological modelling approach based on the Markov random field (MRF). Construction-related uncertainties are addressed using a tunnel vulnerability evaluation method based on the analytic hierarchy process – decision-making trial and evaluation laboratory (AHP–DEMATEL) method, along with the fitting of delay probability density functions. Finally, the proposed system was validated and evaluated in a real-world tunnel project. The results demonstrate that the system effectively analyzes multiple uncertainties, accurately estimates the construction schedule, enhances communication with the construction site, and ensures the safe and efficient completion of the project.

Understanding temperature field evolution during construction is essential for thermal control in high-geothermal tunnels. In this study, field monitoring was conducted to obtain the distributions of rock temperature (T_R), ambient temperature (T_e), and wind speed (v_w), while laboratory tests were performed to characterize the temperature-dependent thermal properties of surrounding rock, initial shotcrete, and secondary lining concrete. Based on tunnel ventilation conditions, a transient heat transfer model incorporating the thermo-temporal effect (TTE) of material thermal properties was developed and validated through numerical simulations. The results show: (1) The surrounding rock, initial shotcrete, and secondary lining concrete all exhibit clear temperature-dependent thermal behavior. Their specific heat capacity increases with temperature, while the thermal diffusivity decreases across all materials. In contrast, thermal conductivity shows material-dependent trends, remaining nearly constant in the surrounding rock but increasing significantly in both concrete types. (2) The influences of T_R, T_e, and W_v on peak tunnel temperature, peak heat dissipation, and the temperature difference between the lining center and edge were quantified, revealing distinct response patterns across construction stages. (3) Compared with constant-property models, the proposed TTE model reduces the prediction error of secondary lining temperature by 17.6%. Additionally, a multivariate prediction model was developed, enabling accurate estimation of temperature extremes and heat dissipation demands during construction.

Tunnel boring machines (TBMs) have become essential in modern tunnelling projects due to their efficiency, safety, and ability to manage complex geological conditions. Accurate prediction of TBM performance is critical for project planning, risk management, and cost estimation. This study focuses on the performance assessment of a hard rock double-shield TBM used in the exploratory tunnel of the Brenner Base Tunnel (BBT) underground system, specifically within the Central Gneiss unit. Several prediction models from the literature, including those based on rock mass classification systems such as RMR and GSI, were applied and compared with actual TBM performance data collected over approximately 3200-m stretch. The relationships between TBM operational parameters, such as penetration rate, instantaneous cutting rate, specific energy, boreability index, power consumption, specific penetration, and geomechanical properties were statistically analysed. Two new predictive models based on RMR and GSI were also proposed and evaluated. The results demonstrate good correlations between rock mass quality indices and TBM performance indicators, highlighting the importance of detailed geomechanical characterisation for reliable performance forecasting.

The spatial distribution of compressive strength in shield tunnel concrete segments and its age-dependent dynamic evolution are vital to the safety and longevity of tunnel structures. However, traditional compressive strength assessment relies heavily on destructive testing, which is labor-intensive, time-consuming, costly, provides only localized information, and cannot capture the inherent spatial variability across the segment. To overcome these limitations and enable efficient, non-destructive, and spatially comprehensive evaluation, this study proposes a novel non-contact method integrating hyperspectral imaging (HSI) analysis with random field modeling (RFM). A hyperspectral camera (900–1700 nm) captured spectral data of concrete segments at various ages. Coupled with a proposed Spectral 3D ResNet deep learning model, this enabled high-precision compressive strength prediction, achieving an average cross-validation coefficient of determination of 0.918 and an average ratio of performance to deviation (RPD_cv) of 3.78, markedly superior to traditional models (partial least squares regression (PLSR), random forest regression (RFR), and convolutional neural network (CNN)). Further analysis revealed that segment strength development progresses through three distinct stages: the plastic stage (0–7 days) characterized by rapid growth but high variability and uneven distribution; the setting stage (7–15 days) with slowing growth and more concentrated distribution; and the hardening stage (15–33 days) where strength stabilizes with significantly enhanced uniformity and stability. Based on these findings, random field models of compressive strength at various ages were established, revealing the dynamic evolution of strength distribution from heterogeneity to homogeneity. At plastic ages, the random field exhibits pronounced spatial variability and longer correlation lengths; with increasing age, spatial correlation diminishes, resulting in a more uniform distribution with shorter correlation lengths at hardening ages. This study provides crucial theoretical backing and a data basis for non-destructive testing and comprehensive performance assessment of concrete shield segments, offering valuable guidance for quality control and structural design optimization in tunnel engineering.

Comprehending the shear behavior of rock fractures plays a vital role in reducing risks in underground rock engineering. An extensively utilized experimental method to investigate this issue is the direct shear tests of fractured rock specimens under a variety of mechanical conditions. Acoustic emission (AE) monitoring, as a powerful tool, can provide real-time insights into the cracking and failure processes during shear tests. In recent decades, a multitude of studies delved into the AE characteristics in rock shear testing, and demonstrated that AE responses were strongly correlated with initiation and propagation of microcracks. Various methods focusing on waveform or amplitude characteristics of AE signals have been developed to interpret AE data, linking AE parameters to shear behaviors of rock fractures. In this review, the previous achievements on AE responses during direct shear tests on rock fractures are first summarized, and a unified method is proposed to reinterpret the representative data and better understand the correlation between AE parameters and the shear strength of rock fractures. The results highlight the potential of AE monitoring for predicting rock failure and improving the safety of underground rock engineering. Finally, the developments and challenges of AE monitoring techniques and interpretation models in fractured rocks are emphasized.

Accurately evaluating the nonlinear rotational stiffness (NRS) of segmental joints is critical for predicting the load-induced transverse mechanical behaviors of shield tunnels. Existing analytical methods and machine learning techniques for joint NRS are limited by simplified assumptions and data volume, respectively, making accurate evaluation challenging. Furthermore, existing transverse mechanical solutions for tunnels are not applicable to arbitrary distributions of joint positions and loads. To this end, this study first develops physics-data hybrid-driven neural networks (PDNNs) for evaluating the NRS of segmental joints. The proposed analytical solutions for constructing physical constraints have been significantly improved in generality compared with existing methods, as they do not rely on known joint deformation paths and can easily incorporate complex material stress–strain relationships. The developed PDNNs are then employed for the iterative calculation of joint NRS in a beam-spring model resting on a tensionless Winkler foundation. Using an adaptive relaxation-iteration strategy and the state-space method, a simplified solution for the transverse mechanical response of shield tunnels with arbitrary joint layouts under arbitrary external loads is proposed. The effectiveness of the proposed methods is validated by comparing their results with those from high-fidelity finite element models (FEMs) in two application scenarios. Furthermore, the effects of concrete constitutive models on joint flexural performance, as well as those of key block position and top loading on tunnel mechanical performance, are investigated. The main conclusions drawn are as follows: (1) the developed PDNNs outperform purely analytical solutions and data-driven neural networks in predictive performance. When the training set contains only joint rotation angles corresponding to one set of axial force cases, the coefficients of determination (R²) of the PDNNs increase by approximately 2% and 46% under sagging and hogging moments, respectively, compared with the purely analytical methods, while the relative L₂ errors decrease by about 7% and 33%, respectively. (2) The simplified solution proposed exhibits good agreement between its predictions of displacement and internal force and the FEM results. Optimal deformation control is achieved when the key block is positioned 60° above the tunnel waist.

Conventional tunnel engineering designs have notable limitations in addressing the fault stick–slip and seismic effects (FSS-SE). To overcome the constraints of quasi-static fault-dislocation model tests, this study employs a dual-shaking table setup to impose non-uniform excitation. A controlled loading framework is established to combine a prescribed permanent fault dislocation with pulse-type near-fault ground motion. The validity of the fault stick–slip simulation is examined using the measured displacement histories of the model box, observed deformation, and failure characteristics of the surrounding rock. The response of a tunnel with flexible joints is analyzed in terms of deformation, damage patterns, acceleration, and strain. The results indicate that the most severe damage is concentrated in tunnel segments located within the fault fracture zone and moving block. The load components for tunnel structures under FSS-SE are conceptually decomposed into three interrelated components: dislocation load, dynamic dislocation effect, and seismic load. Fault dislocation provides the fundamental cause of tunnel damage, while the fault slip rate plays a critical role in governing deformation modes and failure characteristics. With increasing slip rate, the dynamic dislocation effect becomes more pronounced, driving the tunnel response from an overall bending-dominated pattern toward a combined bending-shear mode with increasingly prominent shear-type features. Seismic loading further aggravates damage, particularly for tunnels that have already been weakened by dislocation load and dynamic dislocation effect. The proposed framework helps clarify potential failure mechanisms of cross-fault tunnels and offers engineering insight for coordinated fault-resistance and seismic design.

In high-intensity earthquake zones, seismic mitigation for pipeline tunnels passing through soft-hard strata is of importance to the safe operation of pipeline systems. This study proposed a combined seismic mitigation approach integrating novel flexible joints and buffer layers for tunnel boring machine (TBM) pipeline tunnels. Large-scale (1∶5) shaking table tests were conducted to simulate TBM pipeline tunnels passing through soft-hard strata, with realistic modeling of bolt-connected segmental structures and consideration of buried pipelines within tunnels. Dynamic response and failure mechanism investigation verified the efficacy of the proposed mitigation measures. Test results indicated that the buffer layer effectively reduced peak acceleration and high-frequency components in the Fourier spectrum of the tunnel, while promoting greater consistency in its dynamic responses in soft and hard rocks. The novel flexible joints demonstrated superior deformability to adapt to forced displacements induced by soft-hard strata, thereby alleviating deformation in conventional circumferential joints near the interface. The combined measures reduced the maximum bolt axial forces of the tunnel under seismic motion, with reduction rates ranging from 14.47% to 23.71% in the soft rock and 11.91% to 23.08% in the hard rock. The peak ground acceleration (PGA) threshold for tensile strain of the tunnel exceeding failure limits increased from 0.4g to 0.8g. Furthermore, seismic damage was substantially mitigated or eliminated, including backfill cracking, bolt hole fractures, concrete spalling, and segment dislocation. Although pipeline strains exhibited an increasing trend due to the flexible joint deformation compared to the unmitigated case, measured strains remained well below allowable limits, confirming ample safety margins for the tunnel-protected pipeline system.

In recent years, with the rapid expansion of transportation infrastructure, shield tunnels have encountered increasingly severe soil and water conditions, which present significant challenges for shield tail sealing. In this study, a comprehensive shield sealing failure test device is developed, integrating the tail brush, sealing grease, and external water pressure. The sealing failure evolution is investigated through quantitative analysis of leakage rates and internal pressure distribution. The results show that the leakage evolution can be categorized into three stages: the sealed stage, the micro water leakage stage, and the gushing stage. The compression force of the shield tail brush is a direct factor affecting the breakdown pressure, and a large compression deformation model of the shield tail is proposed to evaluate the breakdown pressure. The findings of this study offer valuable insights for preventing sealing failures, such as guiding the selection of appropriate tail brushes and managing the tail clearance of shield machines.

Deep-buried subway stations are increasingly adopted worldwide, but their long vertical egress paths create compound risks under fire due to environmental hazards (temperature, smoke/visibility, toxicants) and physiological fatigue from sustained stair ascent. This study establishes a unified evacuation framework that integrates a stair fatigue–height velocity model with fire-dependent speed reduction factors to quantify egress performance and derive risk-informed time benchmarks for deep-buried subway stations. A stair-climbing experiment with 200 participants was conducted to quantify fatigue-related speed degradation, and a regression-based model was established to represent evacuation velocity as a function of vertical height. Fire and evacuation scenarios were simulated using PyroSim and Pathfinder to assess the combined impact of fire products (temperature, visibility, toxic gases) and human fatigue on evacuation efficiency. Results indicate that the synergistic effects of fire hazards and physical fatigue extend the total evacuation time to 975.8 s, which exceeds the regulatory standard by 62.6% and is 34.0 s longer than the fire-only condition. The time to reach the station hall exit is 586.8 s, 63.0% beyond the standard limit. Based on these findings, the study recommends adjusting safety benchmarks for deep subway stations to 10 min from platform to station hall and 17 min to the outdoor ground level: 4 min longer than for shallow-buried subway stations. This research contributes a quantitative, simulation-based framework for evaluating evacuation performance under compound crisis conditions. It provides insights for the revision of evacuation standards, the design of resilient underground infrastructure, and the development of decision support tools for emergency response planning.