Purpose Traditional diagnostic methods are, in most cases, based on single-sensor data; however, this configuration may fail to capture the complex dynamic signatures associated with multi-component degradation. This paper presents a lightweight deep learning architecture for the fusion of multi-modal sensor data, typically vibration and acoustic signals. Design/methodology/approach The proposed model represents a dual-branch one-dimensional convolutional neural network (1D-CNN) designed to simultaneously process six axis vibration signals and dual-channel acoustic emissions. To ensure the model's generalizability under multiple operating conditions, the study evaluates two distinct data-splitting strategies: a standard random split and a more rigorous sample-based split that accounts for specific data acquisition conditions and trial-to-trial variability. Findings Experimental results, validated on the open-source Beijing Jiaotong University - Rail Autonomous Operations (BJTU-RAO) subway bogie dataset, demonstrate that the fusion-based approach presents significant results compared to single-modality baselines. The architecture achieves high classification accuracy for the 12 distinct health conditions, which precisely identify both single and multiple-component defects. Originality/value These results highlight the potential of data fusion integration to provide a stable, scalable and frugal framework for predictive maintenance in railway systems, combining high diagnostic performance with a minimal computational and memory footprint.
Purpose This study reviews green high-speed railway (HSR) development in China from a full-life-cycle perspective, focusing on key environmental practices, challenges and future priorities. Design/methodology/approach A structured review of peer-reviewed studies, policy documents, industry standards and railway-sector reports is conducted across six dimensions: land use and biodiversity, embodied carbon, operational energy and emissions, noise and vibration, water and soil environment, and solid-waste management. Findings China has made progress in route optimization, ecological mitigation, prefabricated construction, noise control and environmental monitoring. However, evidence remains fragmented regarding life-cycle trade-offs, project-level indicator application and waste-treatment technology selection. Existing green railway indicator systems require clearer definitions, data support, weighting methods, thresholds and empirical validation. For station and on-board waste, source reduction, separation and recycling should be prioritized, while refuse-derived fuel, pyrolysis and gasification require site-specific feasibility and life-cycle assessment. Originality/value This study integrates environmental impacts, policy frameworks and waste-management strategies into a full-life-cycle evidence framework, highlighting key methodological, data and governance priorities for future green HSR research and practice.
Purpose This study asks whether the Skywalk Concourse of the planned Thammasat University terminus on the SRT Red Line Phase II extension can act as a passive crowd buffer and tests the peak-demand assumptions behind the answer. Design/methodology/approach A whole-station discrete event simulation, checked against a microscopic AnyLogic model, is run for six stress scenarios at three peak-hour demand concentrations (500 iterations each), with capacity, terminus and headway sensitivity tests. Demand concentration is benchmarked against twelve months of fare data from the thirteen operating stations. Findings Every operating station concentrates more demand in its peak hour than originally assumed, so the scenarios were also run at the observed levels. In all 19 configurations, the concourse stays at level of service A and meets the density, holding and drainage criteria, peaking below 290 passengers against a service limit of 464. The binding elements are the platform up-escalator, which breaches a two-minute wait criterion in most morning peak runs, and the platform, which reaches level of service C under a delayed train or an event surge. Research limitations/implications The two tiers agree, but external validation is impossible before opening. Platform results depend on terminus operation and headway; evacuation, local density, guidance and accessibility criteria are not established. A Heat Index projection to 2050 is reported only as a planning-horizon observation. Originality/value The study turns passive crowd buffering into a testable property with eight criteria and separates what a pre-operational assessment can establish from what it must assume.
Purpose Train delay heterogeneity between service categories is well known but rarely characterized at large scale from operational data, and rarely used directly to drive the design of graph-based delay-propagation models. This study uses the real International Railway Network Dataset (IRD2025) – 2,678,291 operational records from 3,324 trains across 1,397 stations on the Italian railway network, January–June 2024 – to empirically characterize this delay heterogeneity and to motivate and evaluate a heterogeneous graph-based delay-prediction framework (HTW-STGN). Design/methodology/approach Exploratory analysis is conducted on the full real IRD2025 dataset, rather than the smaller synthetic validation sample used in earlier work, to characterize delay heterogeneity across the seven Italian train service categories and to test the empirical assumptions – cross-category delay correlation, weather sensitivity and category-level delay ranking – that originally motivated the heterogeneous-edge design of HTW-STGN. HTW-STGN is implemented as a relation-aware graph-feature aggregation model, with heterogeneous edges (same-category/cross-category) grounded in real, physically observed infrastructure sharing and evaluated on a held-out June 2024 test set. Findings Results confirm and correct earlier reported patterns: regional services (96.4% of all records) have lower delays than long-distance categories, the opposite of the result reported from the earlier synthetic sample; the weather effect is real but moderate rather than strongly multiplicative as earlier estimated; and cross-category delay correlation is consistently weak (r = −0.05 to 0.37) rather than the high correlation (up to 0.89) that had originally motivated the choice of heterogeneous graph edges. Only 4.6% of station-to-station links in the network are used by more than one train category, and this real infrastructure-sharing structure – rather than delay correlation – is used to root HTW-STGN's heterogeneous edges. This design reduces mean absolute delay-prediction error by 12% beyond a homogeneous-graph baseline (95% CI [0.49, 0.63] min, p < 0.001) and by approximately 40% beyond features that ignore graph structure entirely, consistently across all seven categories. Originality/value This is the first large-scale empirical characterization of delay heterogeneity across the full real Italian Railway Network Dataset, and the first to test – and correct – the correlational assumptions behind a heterogeneous graph-based delay-propagation architecture using real operational data rather than a synthetic proxy. HTW-STGN's heterogeneous-edge design is re-grounded in observed infrastructure sharing rather than delay correlation, offering a first empirical test of the heterogeneous-edge hypothesis for delay propagation in Italian railways.
