The operation plan of Electric Multiple Units (EMUs) is a critical component in fulfilling transportation tasks for high-speed railways. It plays a crucial role in reducing operational costs and enhancing EMU utilization efficiency, with its rationality directly determining the overall quality of railway operations. Currently, formulation of EMU operation plans in railway practice heavily relies on manual experience, leading to inefficiencies and prolonged planning cycles. Consequently, improving the efficiency and quality of EMU operation planning through mathematical models and algorithms has become a focal point of academic research. This paper provides a comprehensive review of domestic and international studies on EMU operation planning. First, it defines the concept of EMU operation planning and examines the distinctions and interconnections among routing, allocation, and maintenance plans. Second, it systematically reviews the current state of research on EMU routing plans from three perspectives: operational conditions, models and algorithms for routing plan formulation, and EMU rescheduling under emergency scenarios. Additionally, it summarizes the modeling methodologies for EMU allocation planning and analyzes the latest advancements in models and algorithms for high-level maintenance scheduling. Finally, the paper identifies existing challenges in the formulation of EMU routing, allocation, and maintenance plans and outlines potential directions for future research. The research results indicate that early studies on EMU operation optimization primarily focused on improving routing and allocation plans to enhance utilization efficiency. With the expansion of railway network and the increase in the number of EMUs, issues related to real-time rescheduling and high-level maintenance planning have gained increasing attention. Future research should explore the collaborative optimization of maintenance and routing plans, the coordinated adjustment of train schedules and routing plans during emergencies, and precise prediction methods for high-level maintenance scheduling windows.

This study addresses the coordination imbalance between daily operation and major maintenance of Electric Multiple Units (EMU). First, from the perspective of automated scheduling, the coupling process of EMUs in variable marshaling mode is clarified, and the inherent relationship between EMU operation plans and major maintenance cycles is explored. Building upon this analysis, an integrated optimization model is developed by incorporating balanced maintenance scheduling and comprehensive operational costs into the optimization objectives, while considering constraints such as maintenance capacity, timetable requirements, and operational coupling. Then, heuristic rules are integrated into the feasible solution generation algorithm, which, combined with the optimization strategy of the simulated annealing algorithm, results in an algorithm designed to solve the EMU operation plan and optimize the solution process. Finally, case validation is conducted using timetable data from the Wuhan-Guangzhou high-speed railway. Results demonstrate that the variable marshaling mode reduces the required number of EMUs by 16.9% and decreases routine maintenance frequency by 18.7%, compared to the fixed marshaling mode, effectively lowering overall operational costs. After considering balanced maintenance constraints, the balance of major maintenance in the EMU operation schemes increases by 65%, alleviating issues such as concentrated maintenance demands and insufficient depot capacity in practical operations, thereby enhancing the organizational efficiency of EMU deployment.

To support the goals of low-carbon development and reduced logistics costs, this study investigates coordinated air-rail transport allocation strategies. First, passenger transport shares across different travel distances are determined based on traveler preferences. These shares are used to calculate the number of reserved carriages on high-speed rail and the available belly cargo capacity on passenger aircraft. Next, focusing on freight demand along a specific route, a bi-objective optimization model is developed that simultaneously considers transportation cost, carbon emissions, and timeliness-related costs. A nested steady-state genetic algorithm is proposed to obtain the Pareto-optimal solutions for freight flow allocation. Taking demand along the Jingkun high-speed railway corridor as a case study, the model’s validity is demonstrated through practical simulation. Finally, the study explores optimal freight allocation strategies under different carbon emission tax scenarios. Results indicate that the nested steady-state genetic algorithm improves solution accuracy by 59% compared to traditional algorithms. The integrated air-rail freight scheme outperforms single-mode transport in terms of cost efficiency. Moreover, belly cargo transport on passenger aircraft is more suitable for time-sensitive, long-distance shipments compared to high-speed rail. When the carbon emission tax ranges from 0.5 to 3 yuan/kgCO₂, each 0.5 yuan increment leads to an implicit cost increase of 130,000 to 150,000 yuan, while explicit costs decline. However, when the carbon emission tax rises to 3 to 4.5 yuan/kgCO₂, both implicit and explicit costs increase. This research makes full use of the residual capacities of high-speed rail and aircraft belly holds to ensure efficient, low-cost freight transport. It offers a balanced approach to managing carbon emissions and logistics expenses, and provides valuable insights for promoting air-rail cooperation and informing government carbon taxation policy.

To investigate the competitive impact of the rapid expansion of high-speed rail on civil aviation and to further enhance the structural stability of the air-rail intermodal transportation network, this study proposes a vulnerability assessment and restoration strategy that incorporates competitive effects. First, based on the dynamic equilibrium relationship between passenger travel demand and transportation supply, an air-rail competition effect index is introduced to improve the network vulnerability assessment model. Second, in response to various attack strategies and failure scenarios, a restoration model is developed with the dual objectives of minimizing both restoration cost and network vulnerability, and is solved using a particle swarm optimization algorithm. Finally, a multi-scenario comparative analysis is conducted using China Eastern Airline-China Railway intermodal network as a case study to determine optimal restoration sequences under different scenarios. The results indicate that incorporating the air-rail competition effect index increases the average node vulnerability index by approximately twofold, significantly improving the identification and accuracy of critical nodes. By classifying city nodes according to their vulnerability levels, 15 severely vulnerable cities, including Shanghai, Nanjing, Guangzhou, Shenzhen, and Xiamen, are identified. Under deliberate attack scenarios, node and regional failure restoration strategies yield the best results, with more balanced traffic distribution, a 23% increase in restoration cost, and more than a twofold improvement in overall network performance.

In the event of the execution process of the delivery and retrieval plan encountering abnormal conditions, shunting commanders are required to formulate a revised plan within a designated timeframe. To address this issue, this study investigates the real-time rescheduling of wagon delivery and retrieval plans at branch-shaped cargo operation sites under abnormal conditions. Firstly, based on an analysis of the interference caused by various conditions, the research explores decision-making criteria and operational constraints involved in real-time planning at branch-shaped cargo operation sites, and formulates a mathematical model for plan generation under abnormal conditions at such sites. Secondly, to solve the model, an initial solution is generated using a greedy insertion method, followed by the random selection of one of two strategies to produce neighborhood solutions. The fitness value of the solution is calculated by dividing the batch and determining the moment of operation. The model is solved using the Tabu Search algorithm as the main framework. Finally, the task of delivering and retrieval wagons at a railway station within a certain planned period is used as the experimental object. Results demonstrate that the Tabu Search-based approach is both feasible and effective for handling real-time planning under abnormal conditions. The average computation time is approximately 0.817 seconds, which satisfies the time and quality requirements for real-time rescheduling of wagon delivery and retrieval plans at railroad stations. Compared with the Simulated Annealing algorithm and Genetic algorithm, the Tabu Search algorithm exhibits superior performance in terms of computational efficiency and solution stability.

This study investigates the dynamic request matching problem in the ridesharing operations of electric Vertical Take-Off and Landing (eVTOL) aircraft, with a focus on matching and route planning. First, a dynamic route planning model based on ridesharing fairness is first developed, aiming to maximize the benefits of both passengers and eVTOL operators. The model incorporates key constraints such as vertiport capacity, eVTOL payload, and battery energy consumption. Second, two solution approaches, the basic insertion algorithm and the linear insertion algorithm, and two request handling strategies, namely first-come-first-served and request priority, are compared for their effectiveness in matching new requests to available eVTOLs. Finally, a case study is conducted using real geographic data from five train stations and one airport in City T, designated as vertiports. The results show that the linear insertion algorithm reduces computation time by over 60% compared to the basic insertion algorithm, demonstrating its computational efficiency. Furthermore, compared to the first-come-first-served strategy, the request priority approach decreases the average passenger payment by 0.87% and increases operator ridesharing revenue by 5.86%, achieving a more optimal matching between new requests and eVTOLs while balancing the interests of passengers and operators. The proposed dynamic route planning model provides valuable insights for the development of shared eVTOL operation systems.

To address the limited consideration of trip chain information and the correlation between joint travel choices in intercity passengers’ decision-making, this study investigates the key factors influencing intercity travelers’ joint travel decisions and constructs a corresponding decision-making model. First, taking the Beijing-Taiyuan travel corridor as the research object, this study employs the Stated Preference (SP) method to design a set of comparative scenarios incorporating various levels of trip chain information, and analyzes the characteristics and influencing factors of intercity joint travel behavior. Furthermore, to effectively capture the correlation between discrete variables (travel mode) and continuous variables (departure time), a Copula joint model is developed by integrating disaggregate travel behavior theory with Sklar’s theorem. Finally, parameters are calibrated using the gradient descent method, and their significance is assessed through Wald statistics, resulting in the development of a joint travel decision-making model for intercity passengers. The constructed Copula joint model is then applied to identify key factors that significantly influence intercity travel mode choice and departure time. The results demonstrate that, compared to travelers with partial or no trip chain information, those with complete trip chain information show stronger preferences for high-speed rail and later departure time, evidencing the significant impact of trip chain information in shaping joint travel decisions. The statistical tests of Copula parameter further confirm a significant correlation between travel mode and departure time. The Copula joint model also outperforms alternative models in terms of goodness of fit and predictive accuracy.

To address the limitations of mixed-vehicle Global Positioning System (GPS) trajectory data in supporting fine-grained transportation demand analysis and modeling, this study develops a vehicle classification model based on the Gaussian Hidden Markov Model (Gaussian HMM). First, travel characteristic indicators are extracted from trajectory data in both spatial and temporal dimensions. A comparative analysis of travel behaviors between trucks and private cars is conducted to identify distinctive classification features. Then, the model is trained and tested using the Baum-Welch (BW) and Viterbi algorithms, and a classification algorithm based on travel characteristics is designed. Finally, an empirical study is conducted using travel trajectory data from trucks and private cars in Beijing. The results indicate significant differences between trucks and private cars across seven indicators: travel start time, travel end time, total travel duration, average dwell time, average trip time, trip frequency, and travel distance. The proposed Gaussian HMM-based vehicle classification model achieves an accuracy rate of 83% for private cars, a recall rate of 82% for trucks, and an overall model accuracy of 79%, demonstrating its effectiveness in vehicle type identification. The research results offer valuable support for refined carbon emission estimation, differentiated demand management policy development, and fine-grained traffic management.

To investigate the interaction between the built environment and ride-hailing carbon emissions, this study uses ride-hailing order operation data from Nanjing. The built environment indicators are characterized based on factors such as population size, land use, distance to the city center, and housing prices. An Extreme Gradient Boosting (XGBoost) model is established, incorporating built environment factors at both the origin and destination of trips. The model aims to identify key factors affecting ride-hailing carbon emissions and reveal the nonlinear relationships and variable interactions between them. Additionally, the regression results of the XGBoost model are compared with those of the traditional Gradient Boosting Decision Tree (GBDT) model to verify the former’s advantage in regression fitting. The results indicate that the XGBoost model outperforms the traditional GBDT model, with R-squared, mean absolute error, and root mean square error values of 0.541, 0.364, and 0.275, respectively. The distance between ride-hailing trip origins and destinations and the city center contributes significantly, with contribution rates of 20.544% and 29.127%, respectively. Furthermore, the distance from the metro station to the origin and destination exhibits opposing feedback mechanisms on carbon emissions, indicating an asymmetric impact of metro station proximity on emissions. The nonlinear relationship between the distance from the origin to the city center and ride-hailing carbon emissions follows a U-shaped distribution, with significant threshold effects at 7 km and 20 km. Additionally, there are notable interaction effects between the distance from the city center and road density at the trip origin on ride-hailing carbon emissions.

To address the issues of poor generalization and susceptibility to local optima when using a single Back Propagation (BP) neural network for predicting excavation-induced deformations, this study employs Genetic Algorithms (GA) and Particle Swarm Optimization (PSO) for optimization and integrates an Attention mechanism to construct hybrid GA-Attention-BP and PSO-Attention-BP neural network models. The Nanjing Twin Towers excavation project is used as a case study, with PLAXIS 2D simulating the deformation characteristics of the retaining structure and ground surface under 680 different conditions. Additionally, 20 sets of field monitoring data from foundation pits in the Nanjing area are included in the dataset. The prediction results of different neural networks are compared with actual monitoring data under evaluation metrics including Mean Squared Error (MSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). The results demonstrate that the GA-Attention-BP and PSO-Attention-BP models achieve MSE values of 3.47 and 3.22, MAE values of 1.59 and 1.47, and R² values of 0.93 and 0.96, respectively, indicating significant performance improvements over the standard BP and Attention-BP neural networks. Furthermore, the attention-based weight allocation results indicate that excavation depth and diaphragm wall width have the most substantial influence on retaining structure deformation, with weight coefficients reaching 1.33 and 1.17, respectively.

To optimize the grouting parameters for tunnels in urban sand layers and improve grouting performance, a laboratory test system is independently developed in response to practical engineering challenges. The experimental design considers grouting pressure, soil porosity, and slurry water-cement ratio as independent variables, while the strength of the solidified body, slurry diffusion radius, and grouting volume are selected as dependent variables. Regression analysis is conducted on the test results to derive equations describing the relationships between these variables. Based on the resulting linear regression models, optimized on-site grouting parameters are determined. The effectiveness of these parameters is evaluated through observations of excavation and slurry distribution at the tunnel face, core sample strength of solidified body, and surface settlement behavior. The results indicate that the derived equations exhibit strong linear correlations and statistically significant regression coefficients, as confirmed by analysis of variance and correlation tests. Among the influencing factors, grouting pressure has the most significant impact on grouting volume, followed by slurry diffusion radius and solidified body strength. Soil porosity most strongly affects grouting volume, followed by strength, with the least impact on diffusion radius. The slurry water-cement ratio has the greatest effect on the strength of the solidified body, followed by grouting volume, and the least on diffusion radius. Field application of the optimized parameters reveals abundant and evenly distributed slurry veins during excavation. The tunnel face remains stable and upright, with no groundwater seepage. Core sample strength of solidified body meet design requirements, and surface settlement remains within acceptable limits. These outcomes indicate that the optimized grouting parameters significantly enhance grouting performance and effectively ensure the quality and safety of tunnel construction in sandy urban strata.

Cracks in the wall of railway tunnel portals pose a serious threat to the operational safety of rail transport. To address the limitations of manual inspection, such as long inspection cycles and low accuracy, this study proposes a crack detection method for railway tunnel portal walls based on Unmanned Aerial Vehicle （UAV） inspection and the RFA-YOLOv8 model. First, high-resolution image data of railway tunnel portals are captured through UAV inspection. The collected images are preprocessed, and cracks in the tunnel portal wall are annotated to construct a training dataset for wall cracks. Second, the YOLOv8 object detection model is enhanced to better accommodate the characteristics of tunnel portal cracks. The Receptive Field Attention Convolution (RFAConv) is integrated into a newly designed C2f_RFA module, replacing the original C2f module in the backbone network to improve the model’s focus on crack-prone areas. The BiFPN structure is introduced in place of the original feature fusion network to enhance the model’s detection effect for targets of different scales. Additionally, the EIoU loss function is adopted to replace the CIoU, minimizing the differences in height and width between predicted box and ground-truth bounding boxes, thereby improving the model’s detection accuracy. Finally, the RFA-YOLOv8 model is validated and evaluated from three aspects: Comparative experiments, ablation experiments, and visualization of detection results. Experimental results demonstrate that, compared to the original YOLOv8 model, the RFA-YOLOv8 model reduces the missed detection of small cracks, increasing recall by 3.8% and mean average precision by 2.5%. The proposed method effectively leverages UAV-captured tunnel portal images for accurate crack detection.

To address the high cost, long duration, and low efficiency associated with identifying mud pumping defects, this study proposes a novel identification method that integrates Grey Wolf Optimization (GWO) with Variational Mode Decomposition (VMD). First, statistical analysis of mud pumping occurrences is conducted to determine the typical length of affected sections in ballastless tracks. Then, based on track irregularity data collected by dynamic inspection vehicles, the performance of Empirical Mode Decomposition (EMD), Wavelet Decomposition (WD), and VMD in decomposing irregularity signals in mud-pumping areas is compared. Additionally, the effectiveness of several intelligent optimization algorithms, including Sparrow Search Algorithm (SSA), Artificial Bee Colony (ABC), Whale Optimization Algorithm (WOA), Particle Swarm Optimization (PSO), and GWO, in adaptively selecting the key parameters k and α of VMD is evaluated. The proposed GWO-VMD method is employed to decompose the measured track irregularity data, and the characteristics of the resulting Intrinsic Mode Functions (IMFs) are analyzed. The kurtosis values of the IMFs are used as feature vectors, and a maximum likelihood function based on envelope spectrum entropy is calculated to determine the defect identification threshold, which is found to be 3.51. Finally, the effectiveness of the GWO-VMD model is validated through a case study. The results indicate that each IMF component derived from GWO-VMD decomposition of track irregularity data exhibits distinct frequency and amplitude characteristics, corresponding to different spatial scale information. Compared with on-site defect data, the GWO-VMD method achieves an identification accuracy exceeding 90%, enabling effective localization and detection of mud pumping defects in ballastless tracks. The findings support refined service condition management of ballastless tracks and provide technical guidance for “condition-based maintenance” of high-speed railway lines.

To address the problem of thermal buckling at joints in longitudinally connected slab ballastless track structures caused by excessive thermal stress accumulation under high-temperature conditions, this study proposes an optimization method for the inter-slab joints in longitudinally connected slab ballastless track. A thermal damage calculation model for longitudinally connected slab ballastless track structures on bridges is established. The matching relationship between the width and elastic modulus of the elastic material used for joint replacement is analyzed. The effects of the equivalent elastic modulus of the joints on thermal stress, thermal deformation, compressive damage at the joints, and interfacial damage are systematically investigated, and a recommended value for the equivalent elastic modulus is proposed. The results indicate that, for a given width of elastic material, a lower elastic modulus leads to greater release of longitudinal thermal stress, while resulting in greater longitudinal deformation of the track structure, increased vertical deformation reduction at the joint, and higher compressive and interfacial damage. It is recommended that the equivalent elastic modulus of the joints be set at 28 400 MPa, corresponding to a total elastic material width of 50 mm and an elastic modulus of 17 324 MPa. Compared to the original track structure, the replacement of both sides of the wide joint with elastic material reduces the longitudinal thermal stress by 14.2%, the compressive damage at the joints by 7.4%, the interfacial damage by 3.2%, and the vertical deformation of the joint by 2.8%.

The welded seams of the rolling stock body are subjected to complex multi-axis random loads during service. To address the issue of dynamic fatigue reliability assessment, a dynamic stress-strength interference model is established for precise calculation in this paper. First, the vertical and longitudinal load spectra for the intermediate car body are compiled, and an agent model method is employed to transform the multi-axis stochastic loads acting on the car body into dynamic structural stresses at the weld seams. Second, the equivalent structural stress range for the welded joints of the car body is adjusted using the rainflow counting method, and the stress probability distribution is obtained by fitting a two-parameter Weibull distribution. Finally, the strength data are derived from the Stress-Number of cycles (main S-N) curve of aluminum alloy welded joints, and a dynamic stress-strength interference model is established by integrating a damage-based strength degradation model with the stress probability distribution function and the probability density function. This model is used to study the fatigue strength reliability of the dynamically changing car body. The results indicate that the fatigue strength reliability of the critical parts of the car body is initially high but gradually decreases as the operating mileage increases. Over 14 million kilometers of service, the fatigue strength reliability of the lap weld, located near the vertical load input position of the underframe, decreases the most, reaching 0.116 5. The failure rate of the critical points of the car body first decreases and then increases with the increase in operating mileage, exhibiting the characteristics of the bathtub curve, which includes an early failure period followed by a wear-out failure period. The polynomial fitting method for calculating the dynamic stresses of large-scale vehicle body structures proves to be both feasible and efficient. This approach can serve as a reference for the operation, maintenance, and safety and reliability assessment of the car body.

To reduce the aerodynamic drag in the bogie area of high-speed trains, three different bottom deflector designs are proposed to efficiently redirect the airflow beneath the train, thereby achieving energy conservation and consumption reduction. Numerical simulations are conducted using the Improved Delayed Detached Eddy Simulation (IDDES) method based on the SST k-ω model to compare the aerodynamic drag reduction effects of high-speed trains equipped with different structural deflectors. The study also analyzes the impact of these deflectors on the flow field in the bogie area and the wake vortex characteristics. The research results show that, although the total aerodynamic drag of the train body increases after the installation of the deflector, the aerodynamic drag of the bogies is significantly reduced. The deflector’s drainage effect slows the airflow beneath the train, reducing the impact of the air on the protruding components of the bogies. The positive pressure area in the bogie area is notably reduced, and the pressure difference between the front and rear of the bogies decreases, which is the main reason for the reduction in the bogies’ aerodynamic drag. When the train operates at a speed of 350 km/h, compared with the original model, the aerodynamic drag reduction rates for Model Ⅰ, Model Ⅱ and Model Ⅲ are 8.1%, 9.8% and 12.1%, respectively. Furthermore, due to the influence of the deflector, the three-dimensional vortex structures at the rear of the three train models exhibit noticeable changes, and the turbulent kinetic energy near the tail car’s nose decreases. This indicates a reduction in the drag effect of the wake vortex on the train, thereby decreasing the overall aerodynamic drag of the high-speed train.

With the continuous increase in train operating speeds, higher demands are placed on operational safety. This study conducts an in-depth investigation into the operational response of high-speed trains traveling at 350 km/h and above under the influence of crosswinds to solve the safety issues. First, based on fluid dynamics and vehicle dynamics theories, computational fluid dynamics CFD software and UM multibody dynamics software are respectively employed to establish aerodynamic and dynamic models of the high-speed train. Second, aerodynamic loads are applied as external excitations to the train body to calculate the full process of train operation on straight tracks under crosswind conditions. Finally, the influence of steady-state wind load patterns and wind speed on the operational safety of high-speed trains are comprehensively analyzed. The results indicate that under crosswind conditions, the lateral displacement of car body, lateral wheelset forces, vertical wheel-rail forces, derailment coefficients, and wheel load reduction rates increase significantly compared to no-wind scenarios. When the wind speed is 5 m/s and the train speed ranges from 350 to 400 km/h, safety indicators remain relatively low, with gradual increases that stay within safe limits. When wind speeds reach or exceed 10 m/s and train speeds range from 350 to 420 km/h, rail compression by the wheels occur, and there is significant wheel-rail lateral interactions, accompanied by transient separation on the windward wheel-rail interface. These effects are especially pronounced at speeds above 400 km/h. These findings provide important references for ensuring the operational safety of high-speed trains under crosswind conditions.

To investigate the influence of wheel profile evolution and surface hardening on wheel wear during braking, this study employs measured data on metro wheel wear profiles and tread hardness. A finite element model of the wheel-brake shoe system and a rigid-flexible coupled dynamic model of a vehicle system with a flexible wheelset are established to analyze the effects of different worn wheel profiles on wheel-rail contact characteristics, wheel-rail dynamic responses, wheel-brake shoe contact stresses, and braking-induced temperature rise under emergency braking conditions. Based on the consideration of varying wheel wear profiles and tread hardness, an improved Archard wear model is used to calculate wheel-rail wear, wheel-brake shoe wear, and total tread wear. The results indicate that with the increase of tread wear, both the contact stress and contact area between wheel-rail and wheel-brake shoe change significantly. Compared to the wheel-rail vertical force, the lateral and the longitudinal creep force are more sensitive to changes in the wheel profile. The wheel-rail wear is mainly affected by the wheel-rail contact stress, while the wheel-brake shoe wear is primarily affected by the initial hardness of the tread. Under emergency braking, total tread wear is dominated by wheel-brake shoe wear.