The development of traffic networks in mountainous areas has led to an increasing number of tunnels being constructed in regions of high geothermal activity. This study examined the effects of geothermal temperature, heat release rate, and fire source location on temperature distribution and smoke movement in construction tunnel fires through a series of scaled-down experiments. Results showed that geothermal conditions heat the air, creating layered flow within construction tunnels. The temperature and velocity of the induced airflow along the tunnel length were characterized. The upper airflow caused by geothermal conditions hinders the spread of smoke toward the tunnel face, resulting in a complex thermal stratification phenomenon. A model for predicting the smoke diffusion length upstream of the fire source was developed, considering geothermal temperature, heat release rate, and fire source location. Additionally, the ceiling temperature distribution was analyzed, showing that downstream temperature decay is insensitive to fire location, while upstream temperature decay can be divided into geothermal-affected and non-affected zones based on the fire source position. Prediction models for the ceiling temperature distribution upstream and downstream were established, respectively. These findings enhance the understanding of smoke dynamics in construction tunnel fires under high geothermal conditions.

During tunnel excavation, the surrounding soil experiences complex stress redistribution, which is the root cause of the ground deformation and other engineering disasters. Many researchers have studied this issue through numerical simulations, but the results depend on the soil constitutive model and simulation strategy for the excavation process. In this paper, a large-scale laboratory test is conducted using a scaled shield machine, and the three-dimensional stress state of the surrounding soil is measured by a special earth pressure cell. Test data shows that the normal stress components and principal stresses above the crown decrease, and the stress path on the normalized deviatoric plane reaches the failure envelope determined by Matsuoka-Nakai criterion. Due to the misalignment between the stress release direction and principal directions of the geostatic stresses, shear stress is generated in the physical space, which explains the principal stress rotation of the surrounding soil near the shoulder. Near the sidewall, the major principal stress σ₁ is vertical and remains basically unchanged, the intermediate principal stress σ₂ is along the longitudinal direction and increases when the cutterhead reaches the monitoring section, while the minor principal stress σ₃ is along the transversal direction and decreases. On the deviatoric plane, stress paths near the foot and invert have similar development tendencies as those near the shoulder and crown, respectively. Therefore, the influence of the complex stress state on soil behaviours should be considered to provide a reasonable analysis for the tunnel excavation problem.

Underground debris flows, arising from the complex interplay of anthropogenic activities and rainfall-induced hydromechanical processes, present significant geotechnical hazards that remain poorly understood due to their hidden nature and dynamic multiphase triggers. Focusing on underground debris flow in a mining area in Southwest China, this study advances an integrated framework combining air-ground transient electromagnetic method (AGTEM) and computational fluid dynamics coupled with the discrete element method (CFD-DEM), revealing the migration mechanism in which microscale multiphase hydraulic erosion drives the macroscopic initiation of underground debris flow. Key findings include: (1) The identification of three transport phases (rapid erosion, slow erosion, and stabilization) provides actionable thresholds for monitoring and mitigation. (2) The coupled feedback between hydraulic conductivity anisotropy and the formation of preferential flow is the primary driver of large-scale debris transportation. (3) Linking mining-induced seismic energy to vibration-induced liquefaction via DEM simulations offers a physics-based explanation for flow mobilization triggers. The integrated geophysical-numerical framework offers new capabilities for predicting initiation thresholds and developing physics-based mitigation strategies in mining-affected terrains.

Due to the unclear mechanisms behind tunneling-induced deformation of pile-raft foundations, there are strict global restrictions on tunneling beneath embankments of high-speed railways. This study conducted a series of two-dimensional tunneling model tests to investigate the tunneling-induced deformation characteristics and mechanisms of pile-raft foundations. Soil displacement field and pile settlement were measured using particle image velocimetry and displacement transducers. The changes in soil displacement and the flexure of the pile-raft foundation in response to varying tunnel-pile distances, ground surface loads, and tunnel volume loss were analyzed. The results indicate that the tunneling-disturbed zone can be categorized into a loosened zone and an arch zone as identified by the propagation and separation of shear bands, with significant soil settlement occurring in the loosened zone. The maximum settlement of piles in a pile-raft foundation is greater than that in greenfield due to the larger loosened zone. However, the settlement width at the ground surface in pile-raft foundations is reduced due to the blocking effect of the piles. According to the relative position between the piles and the formed arch structure, three patterns of tunneling-ground-pile systems can be identified. As the tunnel-pile distance increases, the maximum settlement of the piles decreases. Increasing surface loads hardly affects the maximum settlement value of the pile, while the tunneling-induced arch zone expands significantly. This study provides a fundamental understanding of pile settlement behavior for tunneling beneath the pile-raft foundations of high-speed railways.

Tunnel construction in karst terrain is influenced by water-filled karst caves and stratigraphic layers, which involves failure characteristics of water-resistant structures and complex fluid-solid interaction (FSI) processes. To cope with this challenge, this paper used coupled discontinuous smoothed particle hydrodynamics (CDSPH) method for investigating water inrush of tunnels considering stratigraphic layers and karst cave positions. Hydraulic fracturing test and sliding-induced impulsive wave test were carried out to verify the accuracy of the CDSPH method. Moreover, a comprehensive analysis of inrush events in the field-scale Qiyeshan (QYS) karst tunnel was conducted, considering different layer dip angles and cave positions on the evolution characteristics of inrush disasters, with quantitative parameters proposed for predicting water/mud inrush from local to overall disaster. The simulation results indicate that CDSPH karst model has been verified to faithfully capture the progressive failure of water-resistant structure during inrush in stratigraphic layers. Water/mud inrush in QYS tunnels can be divided into four stages based on vertical/horizontal stress characteristics, encompassing hydraulic fracturing of karst caves, local inrush, rock collapse, and overall inrush. The dip angle of the bedding planes affects the hydraulic failure characteristics of karst caves. When the cave is located at the top of the tunnel, the water-resistant structures with a dip angle (θ) of 45° poses the highest risk, while θ = 0° provides the most stability. Furthermore, the decrease in water pressure and the occurrence of the maximum flow velocity within the cave can serve as vital indexes to predict the transition from local inrush to overall inrush disaster. These findings emphasize the importance of the CDSPH tunnel model considering stratigraphic layers and karst cave positions when predicting water/mud inrush, and provide guidance for the prevention of inrush flow in karst tunnels.

Severe damage to the Daliang high-speed railway tunnel during earthquakes primarily results from the dynamic interplay between fault dislocation and intense seismic forces near fault lines, accompanied by their complex feedback mechanisms. This study introduces a novel hybrid finite element model to explore the impact of fault dislocation-induced earthquakes on tunnel lining integrity. The influence of seismic characteristics on factors such as peak ground acceleration, tunnel structure form, and the shear modulus of surrounding rock is analyzed. Extensive numerical simulations investigate the coupling effects of faults and various seismic motions on tunnel structures. Additionally, a rapid resilience assessment model for tunnels crossing strike-slip faults is developed using the Adaboost algorithm. This model evaluates the seismic fragility and resilience of such tunnels, offering insights into the anti-seismic behaviors of three distinct tunnel lining configurations under the combined stresses of fault dislocation and significant seismic activity. Furthermore, the fault damage characteristics of the crossing-fault high-speed railway tunnel are assessed, based on real earthquake damage classification and current seismic codes. Findings demonstrate that the evaluation model is both highly accurate and efficient, serving as an effective alternative to traditional nonlinear time-history analysis of tunnel structures. Research shows that critical factors influencing seismic fragility and resilience include the structural design of the tunnel, shear modulus of the surrounding rock, peak ground acceleration, and tunnel height. Simulations reveal that tensile and compressive damage are significantly reduced in circular tunnels with a shock-absorbing joint compared to original tunnel prototypes. Overall, damage assessments from actual faults in crossing-fault high-speed railway tunnels correlate well with numerical predictions, providing essential references for structural recovery and safety evaluations post-earthquake.

Earth pressure balance (EPB) shield tunneling in sandy cobble strata often encounters challenges such as muck stagnation, severe tool wear, difficulties in chamber pressure control, and low excavation efficiency. To address these issues, this study proposes a novel gradient stress construction strategy based on rigid wall boundaries by integrating the finite difference method (FDM) and the discrete element method (DEM), and establishes a refined FDM-DEM coupled shield tunneling model. Using this model, the pressure distribution and load transfer mechanisms at the excavation face and within the chamber, as well as the motion trajectories, velocities, and spatial distribution of muck particles, are analyzed in detail. The results indicate that: (1) The pressure at the cutterhead spokes is lower than that at the cutterhead openings; the muck pressure within the chamber exhibits significant radial gradient variations, with distinct differences between the left and right sides. (2) The average pressure in the upper regions of both the left and right sides of the chamber is nearly equal, with a stable pressure transmission coefficient of approximately 0.8. An under-pressure advancement strategy is recommended to avoid ground heave. (3) The muck particles follow spiral trajectories, forming dual-vortex stagnation zones in the central region of the cutterhead (0-0.2D, where D denotes the cutterhead diameter) and the support column region of the chamber (0-0.25D). The installation of radial mixing rods on the cutterhead shaft is suggested to improve muck flowability. This study provides new insights for optimizing cutterhead and chamber design, offering significant implications for enhancing the efficiency of shield tunneling construction.

Foundation pit excavation for underground space development inevitably disrupts the surrounding soil, raising safety concerns for adjacent buildings. To address the need for an intelligent and trustworthy warning of the excavation-induced risk for adjacent buildings, this study develops a hybrid deep learning framework for probabilistic modeling (PM) with a long short-term memory (LSTM) neural network (termed as PM-LSTM). The proposed framework incorporates Bayesian reasoning and a bidirectional mechanism to enhance its predictive capabilities. The forward learning process enables the dynamic estimation of the probability that adjacent buildings will experience varying levels of risk over time, as new data is incorporated. Meanwhile, it can precisely calculate the first exceeding probability of the adjacent building entering an extremely high-risk level daily, facilitating early warning triggers. Besides, the reverse learning process leverages Bayesian reasoning to identify the most influential response parameters of the foundation pit, serving as key checkpoints for excavation monitoring. It further calculates the posterior probabilities and their intervals for each response parameter under the assumption of a specific risk state for adjacent structures. These insights enable the formulation of proactive risk mitigation measures. The proposed PM-LSTM framework is validated through a case study of the excavation project at Zone A of Jing’an Temple Station on Shanghai Metro Line 14. Comparative analyses further demonstrate the robustness of the framework, underscoring its potential as a reliable decision-making tool for risk analysis and management in complex and uncertain underground engineering projects.

The macro mechanical behavior of rock material is attributed to the meso/mineral characteristics. To deeply reveal the mechanisms of strain rate effect on mechanical properties and crack propagation, a series of unconfined compression experiments and simulations for exploring the meso-scale characteristic were conducted at different strain rates. Based on the micro-loading equipment with microphotography capabilities and the numerical grain-based model method, the meso-scale crack propagation and energy evolution characteristics of granite during the pre-peak loading process were analyzed. The results indicate that with the increase of strain rate, the crack distribution entropy value increases, which means that cracks are more evenly distributed among various minerals. The differences in stored elastic strain energy among different minerals decrease, resulting in more uniform energy release. In addition, cracks associated with biotite transits from intergranular to transgranular modes. Therefore, the increased strain rate can prompt more minerals to participate in deformation, thereby enhancing the mechanical properties. This study deeply reveals the mechanisms of strain rate on granite crack propagation at the meso-scale, offering valuable insights for the stability and safety of underground space engineering.

Twin-tunnel construction inevitably interacts under complex geological conditions, inducing highly complex hydraulic-rock-structure interactions. This study proposes a micro-electro-mechanical systems (MEMS)-based automatic monitoring system for in-situ measurement of rock and structural responses. It measures pore pressure, earth pressure, rock displacement, and additional stress and displacement of segments. Test results reveal three evolutionary stages: pre-shield arrival, shield passage, and post-shield passage. The final distribution and disturbance extent of these responses correlate with tunnel distance. A 3D refined numerical model incorporating the fluid-solid coupling and detailed construction process is developed. Numerical results analyze excess pore pressure, vault settlement, lining response, and key construction parameter effects (face and grouting pressure). Findings enhance understanding of twin tunnel interactions and hydraulic-rock-structural response mechanisms, providing insights for similar projects.

Trace recognition is essential for rock discontinuity characterization of tunnel excavation faces. Traditional methods of trace identification based on three-dimensional (3D) point cloud curvatures require manual fine-tuning for curvature detection and lack consistency with orientation grouping results. This paper proposes a new automatic method for trace identification from 3D point cloud. An adaptive vector method based on neighbor assignment is proposed to accurately generate both normal vectors and directional vectors on sharp points. A principal component analysis-based oriented contraction (PWI-OC) method is presented to extract point cloud skeletons with good iterative conformality. A sparse growing method is proposed to generate extensive trace segments. Two rock excavation face cases, from a mining tunnel and a railway tunnel, are adopted for analysis. The significance of adaptive normal vectors is validated for improving the quality of orientation grouping, and the iterative conformality of PWI-OC is validated to generate more accurate and robust trace skeletons than the traditional method. The results show that the proposed method can achieve a more accurate trace identification than traditional methods, consistent with orientation grouping results, robust to overlapping traces, and automates curvature point detection.

In complex jointed rock masses, wedge blocks are likely to form on the tunnel sidewalls after excavation, and the mechanical properties and stress environment of the surrounding rock have a significant impact on their stability. In this study, cubic rock specimens with prefabricated wedge blocks and arched tunnel features were tested under biaxial compression. Acoustic emission (AE) and digital image correlation technologies were used to monitor crack propagation and specimen failure in real-time. The results showed that supported specimens exhibited higher strength during both the peak and post-peak stages, with a slower strength decline after the peak. The support regulated AE hit rates and enhanced energy storage capacity. Different specimens displayed varying strain evolution, with supported specimens generally having higher lateral strain than shear strain. In unsupported specimens, tensile and shear stresses were concentrated at the wedge block apex, while supported specimens showed more complex stress variations, especially under the influence of wedge blocks. Stable specimens experienced shear sliding failure, while extremely stable specimens experienced both shear sliding and tensile fracture. As horizontal stress (σ₃) increased, specimen strength and wedge block failure both increased. Triangular wedge blocks played a decisive role in tunnel stability, with extremely stable triangular blocks providing greater safety. In addition, a typical stability analysis method for wedge blocks was proposed. The findings provide a scientific basis for rock mass stability assessment and support measure selection in tunnel design.

To investigate the structural stress conditions during the excavation and failure stages of subway stations under adjustable water and soil pressures, a 1∶10 scaled model was created based on similarity theory. Considering the equivalent soil pressure load, the loading procedures that controlled the excavation and failure of a metro station created via the cover excavation reverse construction method were evaluated. Additionally, an excavation unloading device and an external soil pressure-based graded loading device were developed for a metro station created via the cover excavation reverse construction method. By comparing the experimental results with the finite element simulation results, the axial force variations in the balance props during the excavation process were revealed, and the crack development process of the metro station was summarized. The external soil pressure remained unchanged; furthermore, the increase in the axial force of the balance props was negatively correlated with the distance to the previous balance prop and positively correlated with the axial force of the previous balance prop at the time of unloading. According to the graded soil pressure load and the corresponding crack initiation, development, and structural failure states, the model failure process was divided into four stages: the no-crack stage, initial cracking stage, crack penetration stage, and local damage stage. The first cracks in the station structure appeared at the corners and centers of the excavation openings. The first penetration of transverse cracks appeared in the middle of the basement first-floor wall. The cracks at the excavation opening corners and middle locations developed obliquely, forming an overall horseshoe shape. Localized damage first occurred at the corners where concrete spalled, exposing the reinforcement.

Ground penetrating radar (GPR) has been extensively applied in tunnel engineering for the non-destructive assessment of lining structures. However, the interpretation of GPR images remains a time-consuming and expertise-dependent task. To address this challenge, this study proposes tunnel ground-penetrating radar mask region-based convolutional neural network (T-GPRMask), a deep learning-based instance segmentation model designed for the automated detection of tunnel lining defects and components. By integrating a convolutional block attention module (CBAM) and feature pyramid network (FPN), T-GPRMask enhances multi-scale feature extraction, enabling the detection of small, low-contrast defects that are commonly encountered in GPR images. The model was pretrained on a domain-specific dataset containing a diverse set of GPR images related to underground structures and then fine-tuned on a dataset specifically designed for tunnel inspections. The model achieved recognition accuracies of 83.18%, 88.24%, 92.84%, and 91.56% for detecting poor compactness, voids, steel arch supports, and initial lining thickness, respectively. A comparative study further demonstrated T-GPRMask’s superior performance over traditional models, such as YOLOv7 and RetinaNet. Field experiments on real-world tunnel inspection data validated the model’s high spatial accuracy and highlighted its practical applicability in tunnel maintenance.

Maintaining the integrity of sewage networks is crucial for sustainable urban development. Despite extensive research on inspection tools, machine learning applications, and condition assessment for sewer defects, a holistic review of these elements remains absent. This paper addresses this gap by presenting a comprehensive review within a unified framework, employing a mixed-method approach that includes bibliometric, scientometric, and systematic analyses. Our findings reveal that integrating in-pipe and out-pipe inspection methods enhances outcomes. The current literature identifies modified RegNet, dilation segmentation with conditional random field (DilaSeg-CRF), you only look once (YOLO) models, and faster region-based convolutional neural network (Faster R-CNN) as effective algorithms for classification, segmentation, and object detection, both on-site and off-site, respectively. However, machine learning is an evolving field, and future algorithms may surpass these models. Identifying key challenges, we propose recommendations aimed at advancing research and enhancing replicability: notably, the expansion of international research collaborations, particularly in underrepresented regions such as the Middle East, Africa, Asia, and South America; applying the latest version of YOLOv11 in object detection; and investigating defect patterns in polyvinyl chloride (PVC) sewer and rehabilitated pipes using advanced diagnostic methods. This review anticipates aiding policymakers in adopting informed strategies, thereby contributing to the development of smarter, more sustainable cities.

Lithology identification is of vital significance for fundamental geological research and engineering applications. Traditional methods rely on rock image features and often cause confusion among visually similar rocks. To enhance identification accuracy, spectral features are integrated to better represent rock composition. Nonetheless, spectral testing inevitably damages samples and is prone to challenges of the occurrence of similar spectra for different materials. This study explores the advantages of hyperspectral imaging technology, enabling the integration of spectral and image data without damage or contact. A novel spectral-image fusion method is proposed for lithology identification. It uses a dual-channel residual neural network that combines spectral and texture feature channels. Additionally, key features are effectively captured by spectral-spatial attention mechanisms. Finally, a customized transfer learning approach is utilized to fine-tune the pre-trained network on a small dataset for lithology identification at the tunnel site, facilitating rapid model adaptation. The research indicates that employing the ResNetX2-50 network for integrating spectral-image information yields optimal identification results, with a fusion accuracy of over 99% on the test set, which is 2 percentage points higher than the accuracy of a single spectral model and about 20 percentage points higher than the accuracy of a single texture model. Research findings provide robust technical support for non-destructive, non-contact, fast lithology identification in field investigations and construction projects.

To enhance the accuracy of point cloud semantic segmentation in tunnel face construction areas, this study proposes a novel model named enhanced dual attention-tunnel construction net (EDA-TCNet). EDA-TCNet introduces a 3D enhanced dual attention module (EDAM), which employs a parallel channel and spatial attention mechanism to strengthen the model’s focus on critical features. Additionally, a loss function named CELDAM is designed, combining cross-entropy loss and label-distribution-aware margin loss to effectively address data imbalance issues and improve the prediction capability for minority classes. Experiments conducted on three ongoing tunnel projects in Northwest China demonstrate that EDA-TCNet achieves a mean intersection over union (mIoU) of 0.8816 and an overall accuracy (OA) of 0.9406 on the test set. Compared to PointNet, PointNet++, DGCNN, and PointMLP, EDA-TCNet improves mIoU by 18.20%, 3.00%, 8.61%, and 32.23%, and OA by 15.98%, 1.74%, 5.48%, and 22.38%, respectively. Furthermore, the optimization of the balancing coefficient μ in CELDAM further enhances the model’s generalization capability. In conclusion, EDA-TCNet demonstrates exceptional performance in point cloud semantic segmentation tasks for tunnel construction areas and shows great potential for engineering applications.

With the intensive and rapid development of urban underground space, there are more and more adjacent construction disturbances to the existing shield tunnels, posing serious challenges to their safety operation and maintenance. Resilience is an integrated representation of the ability of the engineering systems to resist disaster disturbances and recover function, and it can comprehensively reflect the impact of adjacent construction disturbances on the whole disaster chain of shield tunnels. However, the field currently faces challenges related to vague definitions of resilience, diverse evaluation indicators and measures, and an emphasis on evaluation rather than enhancement. Hence, this paper firstly summarized the definition of engineering resilience, especially the resilience of shield tunnels, and proposed the resilience definition of shield tunnels under adjacent construction disturbance, considering the basic connotation of resilience and disturbance characteristics. Secondly, the existing resilience evaluation frameworks for shield tunnels were summarized and analyzed, and the applicability of the existing framework for the shield tunnel under adjacent construction was discussed in light of the disturbance characteristics. Building on the mechanism and deformation characteristics, the resilience evaluation indexes and evaluation methods were reviewed based on the indicators of influencing factors and indicators of effectiveness of the entire disaster chain. Afterwards, the synergistic enhancement technology of shield tunnel resilience was summarized into 4 aspects: optimal structural design, disturbance transmission blocking, structural performance enhancement, and stratum grouting. Finally, research prospects in this field were analyzed systematically. This paper is intended to provide a meaningful reference for the in-depth research and application of structural resilience of shield tunnels subjected to adjacent construction disturbances.

The initial water content significantly affects rock mechanics, especially with swelling minerals. However, the effects of initial water content on the mechanical properties and failure mode of sulfate rocks remain unclear. This study explores these effects by conducting unconfined compressive strength (UCS) experiments and discrete element method simulations on sulfate rocks. The results indicate that as the initial water content increased from 0 to 9%, the Young’s modulus and Poisson’s ratio of sulfate rock exponentially decreased by 48.9% and 290%, respectively. Additionally, the crack initiation stress ($\sigma _{ci}$), crack damage stress ($\sigma _{cd}$), and UCS decreased by 62.4%, 51.5%, and 53.3%, respectively. The stress responses to initial water content follow linear functions. Notable decreases were also observed in the normal and shear stiffness parameters ($k_n$ and $k_s$ of contact, diminishing by 46.53% and 46.54%, respectively; peak cohesion decreased by 69.70%; peak friction angle by 17.39%; peak tensile strength by 124%. Rising initial water content leads to increased damage and softening of sulfate rock, causing decreased mechanical properties. It can be observed that as the initial water content increases, the proportion of micro-tensile fractures in the total number of fractures increases, and the dominant failure mode of sulfate rock gradually transitions from shear to tensile failure.

Shield tunneling often gives rise to excessive long-term horizontal displacement in consolidating soft ground, posing risks to the safety of adjacent structures. This study investigates the characteristics of long-term horizontal displacement induced by shield tunneling in consolidating soft ground, with the aim of providing practical guidance for optimizing ground treatment strategies. Firstly, a three-dimensional numerical model, validated by a case history in Shanghai, is employed to analyze the horizontal displacement of the soft ground. Comparisons are conducted between the horizontal displacements in normally-consolidated and consolidating cases. Subsequently, the influence of the consolidating state on the horizontal displacement is investigated by numerical analyses. The simulation results indicate that the short-term horizontal displacements follow a similar trend and comparable magnitude in both normally-consolidated and consolidating soft soil. However, the long-term horizontal displacements display a quite different pattern. The maximum discrepancy between normally-consolidated and consolidating cases is observed at the ground surface, where the long-term horizontal displacements of the two cases orient toward entirely opposite directions. The discrepancy at the ground surface increases as the degree of consolidation or the tunnel depth decreases, while it is relatively insensitive to the thickness of the newly filled layer. Finally, an empirical estimation method is proposed to predict the long-term horizontal displacement at the ground surface for shield tunneling in consolidating soft ground.