The use of Unmanned Aerial Vehicles (UAVs) for defect detection on railway slopes is becoming increasingly widespread due to their ability to capture high-resolution images over large, inaccessible, and topographically complex areas. However, current UAV-based detection methods face several critical limitations, including constrained deployment frequency, limited availability of annotated defect data, and the lack of mature risk assessment frameworks. To address these challenges, this study introduces a novel approach that integrates diffusion models with Large Language Models (LLMs) to generate high-quality synthetic defect images tailored to railway slope scenarios. Furthermore, an improved transformer-based architecture is proposed, incorporating attention mechanisms and LLM-guided diffusion-generated imagery to enhance defect recognition performance under complex environmental conditions. Experimental evaluations conducted on a dataset of 300 field-collected images from high-risk railway slopes demonstrate that the proposed method significantly outperforms existing baselines in terms of precision, recall, and robustness, indicating strong applicability for real-world railway infrastructure monitoring and disaster prevention.

Deep learning and fatigue life prediction remain focal research areas in rail vehicle engineering. This study addresses the vibration fatigue of wheelset lifting lug in Chengdu Metro Line 1 bogies, aiming to develop a fatigue life prediction method for critical bogie components using deep learning models and measured track load spectra. Extensive field tests on Chengdu Metro Line 1 were conducted to acquire acceleration and stress response data of the wheelset lifting lug, generating training samples for the neural network system. Component stress responses were calculated via time-domain track acceleration and validated against in-situ stress measurements. Results show that neural network-fitted dynamic stress values exhibit excellent consistency with measured data, with errors constrained within 5 %. This study validates the proposed small-sample deep learning approach as an effective and accurate solution for fatigue life prediction of critical bogie components under operational load conditions.

The stator of the maglev track plays a crucial role in the operation of the maglev system. Currently, the efficiency of maglev track inspection is limited by several factors, including the large span of elevated structures, manual visual inspection, short inspection window times, and limited GPS positioning accuracy. To address these issues, this paper proposes a deep learning-based method for detecting and locating stator surface damage. This study establishes a maglev track stator surface image dataset, trains different object detection models, and compares their performance. Ultimately, YOLO and ByteTrack object tracking algorithms were chosen as the basic framework and enhanced to achieve automatic identification of high-speed maglev track stator surface damage images and track and count stator surface localization feature images. By matching the identified damaged images with their corresponding stator segment and beam segment sequence numbers, the location of the damage is pinpointed to the corresponding stator segment, enabling rapid and accurate identification and localization of complex damage to the maglev track stator surface.

As structural damage patterns and service environments become more complex, digital twin-based structural health monitoring, with its unique advantages, can compensate for the limitations of data-driven methods regarding data dependency and model interpretability. However, it still faces challenges in modeling complexity, simulation accuracy, and discrepancies between real and virtual features. This study proposes a balanced fidelity digital twin for structural damage monitoring based on Lamb wave multilevel feature enhancement and adaptive space interaction. Firstly, multilevel refined features are extracted from few-shot guided wave signals obtained in physical and digital space, and the adversarial synthetic balancing algorithm is proposed for feature enhancement. Additionally, the learning phase of the damage monitoring model based on the feature-mapping convolutional network is driven by virtual samples of readily accessible balanced fidelity in digital space. To reduce the feature distributional difference between the two spaces, an interactive transfer approach is introduced to establish a shared feature digital twin space. Overall, this study provides a feasible technique to enhance the accessibility and generalizability of digital twins for real engineering structures.

This paper studies the structural response of high-speed train wipers under the combined action of complex flow fields and scraping actions. The stress concentration areas are determined through simulation analysis, and the stress and aerodynamic load measurement points are reasonably arranged accordingly. The actual measurement is carried out in combination with the operating conditions of the existing lines. The stress variations and spectral characteristics of the train under different speed levels (80 , 160 , 180 , 200 km/h), tunnel entry and exit, and scraper action conditions were compared and analyzed. The stress amplification factors under tunnel intersection and scraper action were obtained, providing boundary conditions for the design of wipers for high-speed s. The research results show that the maximum stress of the wiper structure obtained through simulation calculation is concentrated at the connection of the wiper arm. Structural stress increases with the rise of speed grade. The stress increases by 1.11 times when the tunnel meets. When the scraper operates, the stress on the scraper arm increases by 4.1-7.6 times. Due to the broadband excitation effect of the aerodynamic load, the spectral energy of the structure is relatively high at the natural frequency, which excites the natural mode of the wiper.

Research on high-speed railways is a relatively new yet highly significant field in Vietnam. Among its key components, train control signaling plays a critical role, as it directly affects various interconnected systems, including infrastructure, traction power supply, operational planning, and overall railway safety. This article focuses on evaluating the capacity of the line based on the types of signals suitable for high-speed railways that have been effectively implemented in several European countries and successfully adapted in China. The research and simulation are conducted using MATLAB software, a reliable and widely adopted tool in the scientific community. The findings demonstrate that under normal conditions, the European Railway Traffic Management System/European Train Control System (ERTMS/ETCS) Level 2 signaling can support up to 23.7485 trains/hour/direction. Meanwhile, ERTMS/ETCS Level 3 with full moving block can accommodate up to 30.8735 trains/hour/direction, and ERTMS/ETCS Level 3 with fixed virtual blocks up to 29.4694 trains/hour/direction. In emergency scenarios, ERTMS/ETCS Level 3 with full moving block reduces headway by 33.27 % compared to CTCS Level 3, while ERTMS/ETCS Level 3 with fixed virtual blocks achieves a 28.78 % reduction. Overall, the ERTMS/ETCS Level 3 emerges as a state-of-the-art signaling technology offering high capacity and operational efficiency, and is recommended as a forward-looking solution for future implementation in Vietnam.

The environmental wind tunnel of high-speed railway trains serves as a crucial experimental facility for the research and development of high-speed railway technology. The refrigeration system within the wind tunnel is an important subsystem. However, the design of the wind tunnel refrigeration system management program presents significant scientific challenges and limitations. Traditional management approaches in wind tunnel refrigeration systems suffer from prolonged decision-making times and reliance on experiential knowledge, necessitating the need for intelligent transformation. This paper aims to address these issues by exploring existing intelligent management methodologies and defining the concept of a wind tunnel intelligent laboratory along with its primary modules. Furthermore, we propose a water cooler failure prediction model based on the existing equipment model of the wind tunnel’s refrigeration system. This model effectively predicts the Remaining Useful Life (RUL) of the water cooler in the case of fouling failure, contributing to enhanced efficiency, cost reduction, and safety improvements in laboratories.

Knowledge graphs, which combine structured representation with semantic modeling, have shown great potential in knowledge expression, causal inference, and automated reasoning, and are widely used in fields such as intelligent question answering, decision support, and fault diagnosis. As high-speed train systems become increasingly intelligent and interconnected, fault patterns have grown more complex and dynamic. Knowledge graphs offer a promising solution to support the structured management and real-time reasoning of fault knowledge, addressing key requirements such as interpretability, accuracy, and continuous evolution in intelligent diagnostic systems. However, conventional knowledge graph construction relies heavily on domain expertise and specialized tools, resulting in high entry barriers for non-experts and limiting their practical application in frontline maintenance scenarios. To address this limitation, this paper proposes a fault knowledge modeling approach for high-speed trains that integrates structured logic diagrams with knowledge graphs. The method employs a seven-layer logic structure—comprising fault name, applicable vehicles, diagnostic logic, signal parameters, verification conditions, fault causes, and emergency measures—to transform unstructured knowledge into a visual and hierarchical representation. A semantic mapping mechanism is then used to automatically convert logic diagrams into machine-interpretable knowledge graphs, enabling dynamic reasoning and knowledge reuse. Furthermore, the proposed method establishes a three-layer architecture—logic structuring, knowledge graph transformation, and dynamic inference—to bridge human-expert logic with machine-based reasoning. Experimental validation and system implementation demonstrate that this approach not only improves knowledge interpretability and inference precision but also significantly enhances modeling efficiency and system maintainability. It provides a scalable and adaptable solution for intelligent operation and maintenance platforms in the high-speed rail domain.