
Abstract The distribution of empty railcars is a critical challenge in railway operations, as inefficient distribution can undermine system-wide economic performance. This study develops a multi-constraint mixed-integer linear programming (MILP) model that integrates railcar-type substitution and carbon emission costs within a low-carbon optimization framework. Unlike conventional models with fixed supply–demand classifications, the model allows station roles to change dynamically according to previous distribution decisions, and substitution relationships among seven car types are explicitly considered. Using real operational data from 17 stations of the Jinan Railway Bureau over a 7-day period, exact solutions obtained via C#-invoked CPLEX are compared with those from a genetic algorithm (GA), a traditional microevolution algorithm (MEA), and an improved MEA. Results show that CPLEX yields the lowest objective value, while the improved MEA achieves solutions within 1.14% of the CPLEX objective value and reduces computation time to 145.87 s, corresponding to a computation time reduction of approximately 15.47%, consistently outperforming the traditional MEA. Carbon emission costs account for approximately 4.8% of the total cost for the improved MEA under the current pricing scheme, providing a low-carbon cost incentive with a moderate impact on distribution structures. Reliability tests based on 50 independent runs confirm the robustness of the improved MEA. Overall, the proposed approach offers a practical and scalable decision-support tool for empty railcar distribution under increasingly stringent carbon regulation.
Virtually coupled train formations have gained increasing attention due to their potential to enhance transportation efficiency. This paper investigates the aerodynamic behaviors of virtually coupled trains with various platoon spacings under crosswind conditions. As the platoon spacing approaches D/H = 2.5, the mean aerodynamic coefficients Cd, Cs, and Cl(where Cd, Cs, and Cl denote the drag, side-force and lift coefficients, respectively.) for the tail of the leading train approach 1. However, the ratios of forces for the head of the trailing train remain below 1 due to the extended wake of the leading train. The sums of Cd and Cs for the tail car (T3) of the preceding train and the head car (T4) of the following train consistently decrease as the platoon spacing decreases within the considered range, while the sums of time-averaged Clfor T3 and T4 exhibit the opposite trend. Instantaneous slipstream on the leeward side displays significant run-to-run variation, in contrast to the windward side. The double-unit region contributes to a nearly constant mean value and standard deviation of the streamwise slipstream velocity component when the platoon spacing exceeds 0.625H. Flow field structures are visualized, revealing a strong shielding effect introduced by the double-unit region. As the platoon spacing increases, T4 remains within the wake of T3, but the velocity deficiency in this region recovers. The findings from this research provide fundamental insights for supporting control strategies aimed at reducing energy consumption and ensuring operational safety for virtually coupled trains operating under crosswind conditions.
Abstract Accurate segmentation of rail surface defect regions is critical to the safe operation of railway systems. To achieve precise rail defect segmentation, a segmentation approach that integrates deformable convolution with a spectral deformable transformer (SDT) was proposed. A multi-scale strip-aware orientation-constrained deformable convolution was employed in the upper encoder block, while an SDT was introduced in the deepest encoder block. In the decoder network, a feature-gated fusion module was constructed by combining an Efficient Up-Convolution Block with a Mixed Local Channel Attention module. In the training stage, the pixel-wise classification loss was calculated by a label smoothing weighted cross-entropy function, while the object-level loss was calculated by a weighted generalized Dice function. On this basis, a unified quality score was introduced to enable coordinated optimization of the two independent loss functions. Based on a self-developed railway inspection platform, rail surface defect images were collected from in-service railway lines and a defect segmentation dataset was constructed. Experimental results on the test dataset show that the proposed method achieved 74.89% mIoU and 85.38% F1. These results demonstrate that the proposed framework can accurately and robustly segment slender rail surface defects under complex inspection conditions, showing practical potential for intelligent railway inspection and maintenance.
Abstract This study examines whether a local top-view variation in the cab region of high-speed train heads can be converted into a measurable design variable that maintains pedigree continuity while supporting aerodynamic decision-making. A cab top-view variant gene, θ, was defined from late-stage Shinkansen models as the angle between the local cab-sidewall trend line and the reference trend line of the main car-body sidewall. Six E2-Series-based models with θ values from 0° to 15° were evaluated using transient numerical simulations at 350 km/h under open-air and tunnel conditions. Increasing θ reduced the three-car drag coefficient from 0.336 to 0.317 and the tail-car lift coefficient from 0.122 to 0.108. Tunnel pressure responses were strongly location dependent: monitoring points near the cab and streamline-transition region were substantially more sensitive to θ than mid-car points. Surface-pressure results further showed weakened positive-pressure and suction regions around the cab transition. θ is not merely a styling descriptor, but a local pedigree-variation parameter with an interpretable aerodynamic effect. It links pedigree recognition, local slenderness control, sectional-area regulation, and aerodynamic assessment, thereby supporting early-stage high-speed train head-shape design.
Abstract The massive accumulation of debris following the 2023 Kahramanmaraş earthquakes necessitates sustainable resource recovery strategies for resilient infrastructure reconstruction. This study establishes a multi-criteria decision-making framework to evaluate the Overall Material Sustainability (OMS) of pavement base layers incorporating earthquake-induced Recycled Concrete Aggregates (RCA) at varying replacement ratios (0% to 100%). An adapted hierarchical Sustainability Assessment by Fuzzy Evaluation (SAFE) model was developed to overcome the limitations of traditional engineering thresholds. First, 30 primary engineering and ecological indicators, derived from standard testing and techno-economic analysis, were objectively weighted using the Shannon Entropy method. These were structured into five core dimensions: Mechanical Performance, Physical Durability, Environmental Safety, Economic Efficiency, and Social Impact. Subsequently, a Mamdani fuzzy inference system (FIS) aggregated these into Technical Performance and Socio-Eco Efficiency, yielding crisp OMS scores. The defuzzified results, mapped against the Prescott-Allen Sustainability Barometer, revealed a parabolic penalty mechanism for extreme designs. Mix-0 (100% natural aggregate) and Mix-100 (100% RCA) fell into potentially unsustainable zones with scores of 0.380 and 0.320, respectively, failing to balance ecological preservation with structural safety. Conversely, among the five investigated mixtures, Mix-50 (50% RCA) achieved the highest model-derived score (0.740) and the best-performing balance between load-bearing requirements, landfill diversion, and carbon reduction. This model-specific optimum depends on the selected indicators, Shannon Entropy weights, and fuzzy rule base and is not a universal RCA replacement ratio. Ultimately, this framework provides a robust, data-driven tool to optimize disaster waste management and accelerate sustainable urban recovery.