To address the challenges of weak feature extraction capability and poor adaptability to multi-source data encountered by by intelligent inspection algorithms in urban rail transit equipment rooms under complex environments, this paper proposes an innovative inspection scheme based on large vision models. This scheme leverages the powerful deep semantic modeling capability of large vision models to perform multi-granularity feature extraction equipment-room monitoring images, while simultaneously introducing an adaptive multi-scale fusion module to optimize feature representation capability. To balance computational efficiency and detection accuracy, a cloud-edge collaborative architecture is designed. The cloud hosts the complete large vision model and is responsible for high-precision inference in complex scenarios and model iteration, while the edge side deploys a lightweight inference module for preprocessing, real-timedetection, and fault early warning of inspection images, ensuring the timeliness and safety of equipment room operation and maintenance. To validate the effectiveness of the proposed scheme, tests are conducted using real operational data from the equipment room of Dalian North Station on Dalian Metro Line 1, covering various equipment types. Experimental results show that the algorithm achieves an mAP@50 of 0.923 and an F1-score of 0.915 in equipment state detection tasks, outperforming mainstream inspection algorithms in overall detection performance. Economic benefit analysis indicates that the scheme incurs an average annual operating cost of approximately RMB 440,000 yuan, representing an annual saving of about RMB 460,000 yuan per year compared with traditional manual and periodic maintenance approaches, demonstrating sound economic feasibility. Furthermore, the scheme adopts a general-purpose industrial hardware acquisition architecture, standardized data access protocols, and a modular algorithm design, offering strong scenario adaptability. It is not limited to urban rail transit equipment rooms and can be extended to other rail transit scenarios such as subway tunnels and stations.

To address the problems of high manual dependence, complex data processing, and low construction efficiency in constructing knowledge graphs for rail transit operation and maintenance, this paper proposes an automated knowledge graph construction method based on large language models. First, two-stage domain-specific prompt templates for rail transit operation and maintenance are designed for domain concept layer generation and operation and maintenance data extraction, respectively. Second, the zero-shot learning capability of large language models is employed to automatically generate knowledge graph data that conform to RDF syntax specifications. Finally, a prompt-driven knowledge graph construction framework is developed to automatically transform rail transit operation and maintenance metadata into RDF knowledge graphs and support integration with existing knowledge graphs. The experimental results show that, compared with the general prompt, the proposed method improves entity extraction accuracy from 81.2% to 93.8%, an increase of 12.6 percentage points; the completeness of the concept layer increases from 68.4% to 91.6%; and the accuracy of complex relation extraction increases from 46.7% to 78.9%. In the comparison with three triple extraction methods, namely OpenNRE, the Language Technology Platform (LTP), and Bert4Keras,

With advancement of intelligent urban rail transit, automatic train operation needs to achieve an effective balance between energy efficiency and punctuality. However, existing dispatching regulations are mostly expressed in natural language, making them difficult to directly incorporate into control modeling. To address this problem, this paper proposes an automated train operation control method integrating large language models (LLMs) and proximal policy optimization (PPO). The proposed method employs an LLMs to perform semantic parsing of dispatching regulations and operational experience, and automatically constructs a Markov decision process (MDP) model under the constraints of a knowledge graph (KG). The generated MDP includes the state space, action set, reward function, and safety constraints. On this basis, an action masking mechanism and an adaptive reward-weight adjustment strategy are introduced, and PPO is used to train a safe and controllable automated train operation policy in a simulation environment. Simulation results show that, compared with conventional rule-based control, Model Predictive Control (MPC), and Deep Q-Network (DQN), the proposed method reduces the average traction energy consumption by approximately 2.5%, 1.5%, and 1.0%, respectively, and reduces the average delay time by approximately 13.7%, 19.6%, and 16.3%, respectively. In addition, the average jerk is reduced to 0.59 m/s³. These results verify the effectiveness of combining LLM-based semantic modeling with reinforcement learning for automated train operation control. Further hardware-in-the-loop tests indicate that KG-based consistency checking can reduce rule omissions and physical inconsistencies in semantic parsing. The action masking mechanism embeds hard constraints, including speed limits and safe headway distances, into policy execution, while the adaptive reward-weight mechanism adjusts the trade-off among energy consumption, punctuality, and ride comfort according to operational feedback. The proposed method therefore establishes a closed-loop framework of semantic modeling, policy optimization, and feedback adjustment, providing an interpretable approach for incorporating natural-language dispatching knowledge into automated train operation control.

To assess the equity performance of an urban rail-transit-oriented public transport system and identify potential issues, this study uses mobile phone signaling data from Shenzhen to build a job accessibility equity evaluation framework integrating spatial and group equity, and proposes a "spatial-disparity identification first, population-characteristic summarization second" approach. Absolute and relative job accessibility are jointly measured under different commuting scenarios; transport-disadvantaged units are identified using the "dual-disadvantage" criterion of low absolute and high relative accessibility; K-medoids clustering is used to characterize socioeconomic attributes; and a decomposable Theil index is applied to identify sources of inequity. Results show a distinctive pattern in Shenzhen's job accessibility: unequal spatial access to job opportunities alongside convergent commuting time. Rail transit shortens commuting time, but job opportunities remain concentrated in station core areas and locationally advantaged areas, while transport-disadvantaged units increase with distance from rail stations. Four disadvantaged groups are identified—multidimensionally disadvantaged groups, experience-based labor groups, traditional labor groups, and high-income, high-education commuting-mismatch groups—revealing diverse mechanisms of disadvantage. Inequity is driven mainly by within-group disparities, while between-group differences are largely driven by education and income. The study provides new insights for equity measurement and sustainable operation of urban rail transit-oriented public transport systems.

Expansive soils, due to the water sensitivity of clay minerals such as montmorillonite and their lattice swelling characteristics, are prone to significant swelling and shrinkage deformation, which seriously threatens the safety and stability of urban rail transit engineering structures such as tunnels and subgrades. In response to issues such as high energy consumption, high carbon emissions, and insufficient durability associated with traditional cementitious material stabilization techniques, this study proposes an active performance regulation method for weakly expansive soils. By developing a novel environmentally friendly stabilizer and a cement-lime composite regulating agent, the macro- and micro-mechanical properties of weakly expansive soils are dynamically regulated, achieving comprehensive improvements in their mechanical performance, durability, and environmental friendliness. Based on multi-scale testing methods, it is revealed that the stabilizer can actively regulate the soil microstructure and mechanical properties through the synergistic effects of multiple mechanisms, including ion exchange, gel network formation, physical filling, and skeleton reinforcement. The results indicate that the cement-lime-stabilizer composite system significantly enhances the mechanical properties and durability of expansive soil. The core mechanism lies in the targeted filling of pores by cementitious materials at the microscopic level and the construction of a three-dimensional network structure. The recommended optimal mix ratio for engineering improvement is cement : lime : stabilizer = 3% : 3% : 0.02%. This ratio ensures the effectiveness of improvement while substantially reducing the amounts of cement and lime used, effectively controlling engineering costs. From the perspective of chemical-physical synergistic regulation, this study provides theoretical support and a technical pathway for the green and efficient improvement of expansive soils.

Against the background of the rapid development of autonomous train operation and virtual coupling, the traditional independent design of vehicle and signaling systems and the distributed control system adopted on the vehicle side have led to problems such as lengthy on-board Automatic Train Operation (ATO) control links, large response delays, and low control accuracy. This paper proposes a multi-system fusion-based closed-loop control method for train braking deceleration. Firstly, an integrated hardware and software platform integrating vehicle control and signaling control functions is established to construct a multi-system fusion control architecture. secondly, based on this platform, the functional redistribution of the entire process of train automatic operation control is carried out, and the core links such as ATO control, traction/braking force collaborative distribution, and closed-loop feedback control are deeply integrated to reconstruct the control closed-loop; Finally, on-vehicle tests are conducted to verify the control performance of the proposed method. The test results show that, compared with the traditional distributed train control system, the control accuracy under ATO mode is increased by 35% with a 25% reduction in response delay; the accuracy of pure electric braking is improved by 70% with a 20% delay reduction; and the accuracy of electro-pneumatic blended braking is raised by 30% with a 14% delay reduction. The integrated architecture effectively improves train braking stability and stopping accuracy, providing critical technical support for the safe and efficient intelligent and autonomous operation of trains.

Under high-density operating conditions, single-point failures in metro vehicle braking systems may escalate from equipment-level faults to operation-level risks. Taking an imported braking system applied in the Beijing Metro as a case study, this paper proposes a system analysis and optimization method for engineering implementation. Driven by a typical operational delay event and based on system decomposition, function-structure mapping, and on-vehicle fault simulation, the propagation paths and amplification mechanisms of key risks, such as exhaust solenoid valve sticking in the brake control valve, strong coupling between the main air line and the parking brake, and braking system-TCMS signal interface mismatches, are systematically identified. A combined Failure Mode and Effects Analysis (FMEA) and Fault Tree Analysis (FTA) approach is then adopted to quantitatively evaluate these risks, with the results serving as direct constraints for system reconfiguration. Coordinated optimization measures are proposed at the software, pneumatic, and human-machine interaction levels, including an HMI-triggered braking system self-test and self-reset mechanism, feedforward diagnosis of low brake cylinder pressure, reconstruction of the parking brake air supply path using check valves, and an operation-disposal-oriented TCMS fault classification and early warning strategy. Multi-line on-vehicle validation demonstrates that, compared with the pre-optimization scheme, the proposed method increases the identification accuracy of three typical single-point faults by an average of 34.3 percentage points and reduces emergency handling time by an average of 3.1 min, significantly mitigating operational impacts such as passenger evacuation, speed-restricted operation, and delays. The established "fault simulation-risk quantification-system reconfiguration" framework provides an engineering reference for resilience-oriented design and proactive risk management of metro braking systems and other critical systems.

To address the problem of elevated stray current caused by insulation deterioration in running rail sections of urban rail transit, which seriously endangers the safety of the traction power supply system and surrounding underground metallic structures, this paper investigates the dynamic distribution patterns of stray current and corresponding protection methods under such insulation deterioration. First, a high-insulation protection scheme for urban rail line sections is designed, and a stray current mitigation strategy for high-insulation sections is proposed, which includes the installation of unidirectional conducting devices and the definition of high-insulation zones. Then, a multi-train dynamic operation simulation model of the urban rail power supply system is established under section insulation deterioration, taking into account the sectional protection control characteristics of the return system. A unified chain-type system model and a power flow solution are developed accordingly. Finally, based on the parameters of an actual domestic urban rail transit line, simulations are performed to analyze the dynamic distribution of stray current and rail potential along the entire line under section insulation deterioration, and the control effectiveness of sectional protection for both stray current and rail potential is evaluated. The results show that insulation deterioration in a section leads to rail potential clamping and an increase in stray current in that section. By adopting sectional protection, which isolates the return current from outside the section while maintaining high insulation within the section, the stray current amplitude in the section can be significantly suppressed, and the rail potential clamping problem induced by insulation deterioration is effectively resolved. The proposed methods can be applied to stray current protection for critical sections of urban rail lines.

To address the insufficient fusion of multi-granularity temporal features and inadequate modeling of spatial heterogeneity among stations in short-term metro inbound passenger flow forecasting, this paper proposes the Multi-Granularity and Attention-enhanced Passenger Flow Prediction model (MGAPF). In the temporal dimension, four one-dimensional convolutional branches with different receptive fields are constructed, and a granularity attention mechanism is used to adaptively fuse multi-scale temporal features. In the spatial dimension, a graph attention mechanism is introduced to dynamically learn the correlation strengths among adjacent stations. External temporal contextual features, including holidays, weekends, and weather conditions, are also incorporated. A gated fusion structure, residual connections, and a Transformer encoder are further combined to perform spatiotemporal feature interaction and generate prediction outputs. Experimental results based on real operational data from Hangzhou Metro show that MGAPF achieves MAPE values of 16.93% and 9.52% in the 10-minute and 30-minute forecasting tasks, respectively, representing error reductions of approximately 4.5% and 7.8% compared with the second-best baseline model. MGAPF also performs better in terms of MAE, RMSE, and prediction stability, indicating that the fusion of multi-granularity temporal features, spatial correlations among stations, and external factors can effectively improve the accuracy and stability of short-term metro inbound passenger flow prediction.

In response to the problems of obscured fault characteristics of metro bogie bearings under complex conditions, the susceptibility of original vibration signals to noise interference, and the poor generalization and high complexity of traditional diagnostic methods, an intelligent bearing fault diagnosis method integrating Eulerian Feature Extraction (EFE) and shallow convolutional neural networks is proposed. This method achieves dimensionless normalization and standardization of vibration signals through adaptive interval mapping, converts linear time series signals into a polar-coordinate phase representation using geometric phase encoding, and then constructs two feature expressions, FirstEuler and SecondEuler, through dual Eulerian feature fusion to achieve deep integration of time-domain and phase-domain features, effectively enhancing the discriminability of weak fault features. The features extracted by EFE can be directly used as input to a one-dimensional convolutional neural network or converted into grayscale images as input to a two-dimensional convolutional neural network, enabling the construction of shallow diagnostic models that balance diagnostic accuracy and efficiency. Using the Case Western Reserve University bearing dataset, fault diagnosis experiments were conducted under single-condition, multi-condition, and variable-speed operating conditions, and the proposed method was compared with traditional feature extraction algorithms such as VMD, EMD, and HT. Experimental results show that in a -6 dB strong noise environment, the EFE method improves one-dimensional diagnostic accuracy by more than 7.8% compared with the raw signals. In single-condition two-dimensional diagnosis, the accuracy is approximately 19.5% higher than that obtained using the raw data, and after secondary enhancement of features extracted by VMD, EMD, and other algorithms, the diagnostic accuracy reaches up to 100%. Under multi-condition and variable-speed operating conditions, secondary feature extraction by EFE increases the diagnostic accuracy by more than 15% on average; in particular, VMD combined with FirstEuler achieves 100% diagnostic accuracy for the full variable-speed condition combination. The proposed method effectively suppresses noise interference and highlight weak fault features, demonstrating excellent generalization ability and feature extraction performance under complex conditions such as constant load, variable load, and variable speed. In addition, the shallow CNN model reduces computational complexity, providing a new technical solution for efficient fault diagnosis of metro bogie bearings.