To investigate the structural damage characteristics and deformation mechanisms of in-service metro tunnels in collapsible loess regions, this study conducted long-term structural deformation monitoring on a metro tunnel in Xi'an, China. Safety assessments of tunnel sections were performed based on monitoring data, with numerical simulations further revealing the causes of deformation in high-risk zones. The monitoring data show that the study line tunnels present a composite deformation pattern of arch subsidence and structural expansion dominated by arch subsidence. The Z5–Z6 interval was evaluated as a high-risk section due to the seriousness of the structural distress. Numerical simulations demonstrate distinct deformation mechanisms: saturation of the soil above the tunnel crown predominantly induces horizontal ovalization deformation ("horizontal egg-shaped" distortion), while saturation below the invert mainly triggers overall subsidence with minimal impact on structural dilation or surrounding rock pressure. By combining the geological investigation and deformation monitoring data, the monitored structural-expansion values match well with the predicted values for a 4-8 m flooding depth in simulation condition 3 (groundwater level rise), confirming that the compressive deformation triggered by the flooding of old loess beneath the invert is the main cause of structural deformation in this section. The findings provide valuable insights for the construction and operation and maintenance of urban rail transit systems in similar collapsible loess regions.

In light of the rapid expansion of underground rail transit systems and the growing requirements for tunnel operation and maintenance, the longitudinal structural resilience of shield tunnels has emerged as a critical research topic in the field of engineering. In this paper, the longitudinal differential settlement is employed as a performance index, and the ratio of residual performance to normal performance of differential settlements is defined as the resilience metric Re, thus a new model using longitudinal relative differential settlement as the index of tunnel structural resilience performance is proposed. Based on a case study in which a newly constructed shield tunnel of Shanghai suburban railway airport link line undercrosses the existing Metro Line 15 tunnel, the variations in tunnel deformation with respect to tunneling distance and construction schedule are analyzed. Then, the degradation characteristics of structural performance during both the disturbance stage and the recovery stage are effectively captured. Finally, the proposed model is applied to evaluate the structural resilience of existing metro tunnels. The results show that 32.3% of the performance loss is attributed to the new tunnel undercrossing during the first 38 days. After the completion of the grouting reinforcement, the tunnel performance is improved from 0.677 to 0.868; this improvement accounts for approximately 59.1% of the performance loss during the crossing period. Eventually, the resilience metric Re is calculated as 0.764, indicating a high level of resilience for the existing metro tunnel in this case. Moreover, other performance indices related to the longitudinal settlement of the tunnel are compared to validate the rationality and applicability of the proposed performance index.

With the increase of service life of urban rail transit, the stability of subway tunnel lining is getting worse and worse. Aiming at the problem of difficulty in detecting voids behind lining, impact echo method has been developed as a new means to detect such diseases. Among them, effectively extracting echo oscillation characteristics from complex signals of detection data has become the key to detecting voids behind lining. Based on the voids detection experiment behind "segment-surrounding rock" full-scale model, this paper proposes a denoising method based on Butterworth bandpass filter signal coarse screening strategy combined with local noise smoothing based on Daubechies8 wavelet transform. This method uses 4th-order bandpass filter and sets high and low frequency cutoff frequencies of 1 kHz and 50 kHz to effectively suppress signals and noise outside the frequency band; wavelet transform based on improved adaptive threshold method smoothes signal spikes and finely removes residual noise within the frequency band. The results show that after denoising the two-dimensional time-frequency diagram of experimental data by this method, high-frequency spikes are suppressed in the frequency domain, low-frequency discrete noise is removed, and echo signals in the passband are smoothed; the distribution characteristics of echo signals are retained in the time domain to reduce noise interference. The joint denoising method proposed in this paper significantly improves the extraction effect of the echo signal oscillation characteristics, and provides a preliminary research basis for the impact echo method to detect voids behind the lining of subway tunnels.

In order to study the influence of vehicle conditions on wheel-rail system vibration in surburban railways, a vertically coupled dynamic vehicle-track-tunnel model is established, and the influence mechanism of operating speed, train passenger load, wheelset mass and wheel polygonalization on the vibration response of vehicle, rail and tunnel is analyzed in detail. The research results show that with the train speed increasing from 100 km/h to 160 km/h, tunnel vibration source strength increases from 68.05 dB to 72.06 dB. With the increase of train passenger load, the vertical ride comfort index of vehicles gradually decreases, while the vertical wheel-rail force, rail vertical displacement and acceleration, maximum Z vibration level of the rail, and tunnel vertical displacement gradually increase, while the tunnel vertical acceleration and maximum Z-vibration level remain basically unchanged. Passenger load mainly affects the tunnel vibration response below 10 Hz. Above 10 Hz, the response is governed primarily by wheel-rail irregularities, and the tunnel acceleration level shows no significant differences among no-load, full-load, and overload conditions. Increasing wheelset mass increases the maximum Z vibration level of the rail and tunnel. During routine vehicle maintenance, the excitation frequency of wheel polygonalization should be prevented from coinciding with the wheel-rail system P2 resonance frequency.

To address bottlenecks and challenges caused by scarce fault samples, unclear failure modes, and insufficient methods for analyzing the reliability performance of regenerative braking energy feedback devices in urban rail transit, this study analyzes equipment reliability by collecting and examining fault event data for these devices. The paper establishes a fault tree model for the regenerative braking energy feedback devices and explores Gaussian kernel density estimation (KDE) to obtain a fitted failure probability distribution function, thereby constructing a reliability modeling and analysis method for such devices. This approach transforms discrete monthly mean time between failures (MTBF) data into a smooth probability density function curve, effectively narrowing the fluctuation range of reliability estimates. Furthermore, it calculates the core MTBF value range with a 95% confidence interval, significantly enhancing both the intuitiveness and scientific understanding of fault data and its distribution for such equipment. This methodology assists in precisely determining the optimal timing for scheduled maintenance, enabling dynamic maintenance adjustments, and supports future equipment selection and lifecycle management decisions.

To address the core problems of difficult evaluation of photovoltaic (PV) absorption and the poor alignment between energy storage (ES) configuration and life-cycle benefit maximization after integrating PV-storage hybrid systems into urban rail transit, this study first simulates the dynamic fluctuation characteristics of traction loads and the hourly PV power output characteristics throughout the year. On this basis, a multi-dimensional mathematical model incorporating PV characteristics, ES operation strategies, and life-cycle benefits is constructed. The model is then used to systematically analyze the annual PV absorption level, optimal ES capacity configuration, and ES charging and discharging strategy design, and proposes an optimal configuration scheme for urban rail transit PV-storage hybrid systems that accounts for time-of-use (TOU) electricity prices. The results indicate the following: Accurate assessment of PV absorption requires analyzing annual PV generation characteristics and load variation patterns; ES capacity selection should integrate PV absorption rate, investment payback period, and annual average benefits to minimize the static investment payback period while improving PV absorption, thereby achieving full-life-cycle profitability; the design of ES charging and discharging strategies needs to fully consider annual TOU electricity price differences, maximizing the use of peak-to-off-peak price spreads to enhance benefits. This research method provides a reference for the scientific planning and implementation of PV-ES integrated projects in urban rail transit.

To address the issues of poor real-time performance and low accuracy in existing subway protection area supervision models, an intelligent inspection system was developed that integrates video surveillance with image recognition and edge computing technologies. Firstly, an edge computing system for protection area inspection was constructed using edge computing terminals and surveillance cameras, effectively reducing reliance on cloud servers. Secondly, the YOLOv8l algorithm was selected as the image recognition model for the intelligent inspection system and deployed on the edge computing terminals. Finally, a B/S architecture -based inspection cloud platform was developed for managing inspection equipment, emergencies, and operations. Through training and testing with approximately 100,000 images of 6 categories of mechanical risk sources under various scenarios, the model achieves a consistent recognition rate exceeding 90% in both daytime and nighttime conditions. Field tests conducted along Guangzhou Metro Line 5 demonstrate an overall system recognition rate of 85.42%, indicating the substantial potential of this video-based intelligent inspection system. This advancement holds significant practical implications for ensuring the sustainable development of urban rail transit systems.

Focusing on the flood prevention safety assessment for rail transit operating lines during their operational period, this study takes the Nanjing rail transit operating lines as an example to establish a key technical system for flood prevention safety assessment during the operational period of these lines. This system encompasses five types of sitess along the operating lines, including subway stations, vehicle depots, substations, section ventilation shafts, and tunnel portals for vehicle access. The key technical system primarily involves dividing flood prevention zones, analyzing hazard factors, establishing a risk assessment matrix, and proposing flood prevention risk control measures. Based on this system, the Nanjing rail transit operating lines are divided into 34 level-I zones and 47 level-II zones. The hazard factors faced by the rail transit operating lines are categorized into six aspects: external floods, waterlogging, flash floods, local drainage, entrances and exits, and drainage/blocking facilities. Results show that there are 21 critical risk points, 112 significant risk points, 201 moderate risk points, and 61 minor risk points along the lines. The critical risk points are primarily concentrated in Lines 1, 2, 3, and 10. Corresponding risk control measures are proposed for different risk levels and types of points. The research findings provide theoretical guidance for ensuring flood prevention safety in urban rail transit operating lines.

Bolts on rail transit evacuation platforms are often concealed, making comprehensive manual inspection difficult and increasing the risk of loosening or detachment that could compromise operational safety. This paper proposes an intelligent inspection system integrating machine vision and laser scanning to achieve precise detection of bolt defects and clearance-intrusion trends. The system employs a dual-track autonomous robot equipped with a high-resolution camera and LiDAR, enabling comprehensive inspection through image recognition and point-cloud processing. First, high-resolution image acquisition combined with Adaptive Contrast Enhancement (ACE) is used to improve image quality. Optimized YOLOv11 and Mask R-CNN models are then applied for small-object detection and segmentation, enabling automatic identification of missing bolts, loose nuts, and insufficient bolt protrusion, with an accuracy of 94.4%. Furthermore, laser point-cloud modeling combined with a vector cross-product algorithm is employed to provide clearance-intrusion warnings with a precision of 2 cm. Field tests conducted on the Hangzhou-Haining Intercity Railway, across both elevated and tunnel sections, demonstrate an F1-score of 92.6% for bolt defect recognition and a 0% false alarm rate for clearance detection. The proposed system significantly enhances inspection efficiency and provides technical support for intelligent maintenance of rail transit infrastructure.

To promote “four-network integration” of rail transit and better support the development of urban agglomerations and metropolitan areas, this study deeply analyzes the connotation and characteristics of the “four-network integration” and summarizes its role and significance. On this basis, the study systematically reviews the four-network integration planning practice of Guangzhou's rail transit and proposes recommendations from the aspects of top-level design, operational organization, and institutional mechanisms. The research results indicate that to achieve high-quality four-network integration, it is essential to firstly establish a rail functional hierarchy system suitable for metropolitan area development based on its spatial scale, central system, and passenger flow demand. Secondly, for a certain direction of passenger flow corridor, it is necessary to fully consider the characteristics of passenger flow demand, and study the reasonable configuration of different lines within the same channel from the perspective of leveraging the advantages of various levels of track functions. Thirdly, the appropriate network integration method should be selected from a holistic perspective, balancing demand and cost while considering the applicability of interconnectivity and hub node transfers. Finally, efforts should be made to promote mutual trust in security checks and interoperability of ticket systems to facilitate passenger travel.