High-speed railway (HSR) turnouts are among the most mechanically demanding components in the railway infrastructure, yet current operation and maintenance (O&M) practices remain largely reactive, experience-dependent, and disconnected from automated decision support. This paper presents a large language model (LLM)-driven expert system that bridges the gap between raw sensor data and actionable maintenance decisions for turnout defect diagnosis. Three core contributions are made. First, a data textualization strategy is devel oped to convert train body acceleration signals into structured text sequences comprehensible to LLMs, enabling domain-specific diagnosis without architectural modification of the base model. Second, an enhanced instruction fine-tuning scheme is proposed, incorporating a contrastive loss function that tightens intra-class feature clus ters and widens inter-class margins, alongside a hierarchical evaluation method that reliably extracts categorical intent from free-form model outputs. Third, a retrieval-augmented generation (RAG) module is integrated with the fine-tuned model, enabling the system to generate standards-compliant maintenance recommendations di rectly from diagnostic results. Controlled experiments across four pre-trained models and 26 experimental groups demonstrate that the proposed system reaches a peak diagnostic accuracy of 89.6%, while preserving the natural language generation capabilities essential for report production. The framework is evaluated on a physically rep resentative dataset generated by a validated stochastic vehicle-turnout dynamics model. The resulting integrated pipeline, from extracted signal features to maintenance decision output, offers a practical and scalable solution for intelligent O&M of complex railway turnout infrastructure and beyond.

High-speed railways constitute a critical component of modern transportation infrastructure. However, the longitudinal rail force, influenced by cyclic train loads and environmental conditions, poses considerable safety risks if not accurately monitored. Traditional longitudinal rail force detection techniques primarily rely on physical testing and periodic manual inspections, which significantly limit the potential for real-time and continuous monitoring. To address these limitations, this study introduces an innovative approach employing Large Language Models to predict longitudinal rail force based on historical monitoring data. The proposed method is validated through extensive long-term field monitoring of longitudinal rail force on high-speed railway lines, thereby confirming its practical applicability. In contrast to conventional time series forecasting large language models, the proposed method evaluates the prompt-free architecture. When applied to real-world longitudinal rail force data from the Beijing-Shanghai High-Speed Railway, the model achieves an average coefficient of determination (R2) of 0.932 and a root mean square error of 3.537 kN, outperforming traditional deep learning models. Furthermore, the model exhibits strong robustness under conditions of intermittent data loss. The proposed framework is seamlessly integrated into a localized intelligent system using Langchain-Chatchat, enabling expert-level recommendations based on domain-specific documentation. Overall, this study presents a practical, efficient, and scalable solution for intelligent railway monitoring, offering an advancement toward safer and more intelligent high-speed railway operations.

The track-bridge system is a key component of high-speed railways; however, evaluating the global reliability of the vehicle-track-bridge system (VTBS) under seismic excitation poses multiple challenges. Specifically, existing physical models require substantial computational costs to obtain dynamic responses; traditional surrogate models often lack sufficient accuracy in characterizing complex nonlinear mappings; furthermore, conventional reliability solution processes are inefficient and struggle to effectively account for complex system failure modes. To address these issues, an efficient global reliability evaluation method for the VTBS under earthquakes (GREMVTBE) is proposed. First, a dynamic analysis model incorporating multidimensional random parameters is established to achieve the accurate solution of random responses. Second, an equivalent extreme value event integrating multiple safety failure modes is constructed for global reliability assessment, which avoids the single failure mode limitation and the complexity of high-dimensional joint integration in traditional reliability analysis. Finally, a surrogate model is introduced to overcome computational bottlenecks. By employing the Bayesian optimization (BO) algorithm to adaptively optimize the key hyperparameters of the Extreme Gradient Boosting (XGBoost) model, a high-fidelity surrogate model is constructed to directly predict the equivalent extreme values. The results demonstrate that the proposed surrogate model substantially decreases the required number of dynamic response computations, reducing the computation time of the GREM-VTBE method to only 33.36% of that required by traditional methods, while maintaining exceptional predictive accuracy and robustness. This outcome provides a highly practical evaluation tool for the early warning and engineering protection of the VTBS during seismic events.

Track slab voids in ballastless high-speed railway systems severely threaten structural safety and operational reliability. Conventional structural health monitoring (SHM) approaches relying on the ‘repair-then-infer’ pipeline are vulnerable to error accumulation under sensor failures. This paper proposes a physics-guided sensor-robust mixture-of-experts model (PhyMoE-R) for direct void detection using corrupted data without intermediate restoration. The framework integrates three components: (1) sensor-expert binding with dynamic masking to isolate faulty signals; (2) SHAP-based physical priors for interpretable network initialization; and (3) topology-guided bidirectional distillation for cross-channel knowledge transfer. Validated via 5-fold cross-validation on full-scale track slab experiments, PhyMoE-R achieves a mean F1-score of 0.920±0.010 under sensor failures, outperforming all sequential and unconstrained mixture-of-experts baselines. Ablation studies confirm the necessity of incorporating physical priors. By integrating a residual bottleneck expert module, the model reduces parameter size by 34% while attaining a worst-case F1-score of 0.6. The proposed approach offers a robust solution for SHM under sensor failures.

In earthquake-prone areas, the operational safety of trains is under potential threat, making it crucial to conduct prediction and assessment of high-speed railway (HSR) operational safety for emergency avoidance. However, traditional numerical simulation-based assessment methods suffer from low computational efficiency and complex processes, while purely data-driven deep learning models lack physical mechanism support, resulting in poor interpretability and insufficient robustness. To address these challenges, this study proposes a physics-informed operational safety assessment framework for predicting and evaluating HSR operational safety indicators under earthquakes. First, a train-track-bridge coupled dynamic model under earthquakes was established to generate sample data of vehicle safety indicators. SHAP interpretability analysis was conducted to obtain key physical information affecting HSR safety under earthquakes. Various modular static-dynamic feature fusion mechanisms were developed to improve the utilization of high-dimensional data features. Based on the derived physical laws, a family of weighted loss functions incorporating seismic wave characteristics and prior knowledge was proposed to enhance their focus on sensitive information. Finally, the improved framework was applied to multiple baseline models to verify its accuracy in sequence prediction and classification tasks. The results show that this physics-informed framework can accurately predict HSR safety indicators and classify over-limit conditions under earthquakes, providing a highly promising solution for assessing HSR operational safety in earthquake-prone areas.

The track-bridge system is a core component of high-speed railway infrastructure. However, assessing the reliability of the vehicle-track-bridge system (VTBS) under disaster effects poses numerous challenges. Existing studies have neglected the consideration of multi-source random parameters, lack exploration of time-varying reliability, and suffer from low solution efficiency of the probability density evolution equation (PDEE). To address these issues, this study proposes a precise and efficient framework for dynamic reliability assessment of the VTBS under extreme disasters (PEDRA-VTB). The specific contributions are as follows: (1) A vehicle-track-bridge dynamic model incorporating multi-source random factors is constructed, enabling accurate solution of random responses. (2) The differential grid division of PDEE is optimized, reducing the PDEE solution time for the VTBS by 834 times while controlling the computational error within 0.8%. (3) With earthquakes as the background, the temporal probability distribution evolution law of the response and the time-varying characteristics of dynamic reliability of the VTBS under disasters are revealed. The research results can provide reference for the design and optimization of key system parameters in railway disaster resistance.

Slab track systems have been extensively adopted in high-speed railways worldwide. Post-installed reinforcement bars are commonly employed at the interface between precast slabs and underlying concrete structures to mitigate thermal arching risks. However, field inspections have revealed localized void damage in slab tracks reinforced by post-installed reinforcement bars. This type of damage is often concealed, making identification and diagnosis challenging. Traditional detection methods are costly and entail complex operational procedures. This study presents a novel rapid identification model for void damage assessment in post-installed rebar-reinforced slab tracks, integrating impact hammer testing with deep learning algorithms. To systematically validate the model's feasibility and reliability, laboratory experiments were conducted to simulate varying degrees of post-repair void damage scenarios. Time-frequency analysis was performed to extract latent features from vibration acceleration profiles across different damage conditions, followed by damage diagnosis using specifically designed neural network architectures. The results show that the proposed model can achieve high-precision identification of void damage based on measured structural acceleration, with diagnostic accuracy exceeding 98 % when using one-dimensional convolutional neural networks and residual neural networks. The study represents the first successful identification of void damage in slab tracks post-repair with post-installed reinforcement bars, enhancing the assessment and diagnosis of track system service conditions and effectively guiding practical engineering maintenance.

In this article, we propose a novel switching strategy for maintaining constant output voltage in dynamic wireless power transfer (DWPT) systems, leveraging a positioning scheme based on mutual inductance estimation. First, we enhance the secondary coil positioning method by improving mutual inductance estimation to support various load types. Furthermore, a practical secondary coil positioning scheme is introduced to address real-world challenges in DWPT systems, factoring in parameter fluctuations and time delays. Following this, steady-state and transient models of the primary coil switching process are derived to thoroughly assess the switching dynamics and identify potential safety risks. Based on these insights, we develop a primary coil switching strategy that guarantees both system safety and consistent output voltage while facilitating practical implementation. Simulation and experimental results validate the design objectives of the secondary coil positioning scheme and primary coil switching strategy. Furthermore, we analyze the mutual interactions between the positioning scheme and switching strategy, confirming their compatibility and effectiveness under simultaneous operation.

Accurate online parameter estimation is crucial for effective system control, monitoring, and maintenance for wireless power transfer (WPT) systems. However, unavoidable and unpredictable deviations in compensation capacitances, caused by factors such as tolerance, environmental conditions, and aging, complicate the parameter estimation process. To address this issue, this article proposes a simple and fast online parameter estimation method based on a frequency adjustment strategy. This method alternates between secondary resonance and nonresonance states, allowing simultaneous estimation of mutual inductance, load resistance, and compensation capacitances using only additional two voltage sensors and three current sensors. Only dc and rms values are utilized, without the need for high-bandwidth communication, switches, or complex algorithms. Experiments demonstrated that the method could simultaneously estimate multiple parameters over a wide power range, with errors below 10% in 7.0 ms. The maximum average errors were 7.0% for mutual inductance, 6.7% for load resistance, and 2.8% for compensation capacitances when combined with efficiency optimization strategies. Furthermore, the frequency adjustment selection principles were outlined. Both analysis and experiments confirmed that the brief detuning duration and a 2% frequency deviation had minimal impacts on WPT systems, thus enabling a cost-effective and fast solution for practical applications.

The high penetration of distributed photovoltaic (DPV) systems in distribution networks (DNs) can lead to a series of issues such as reverse power flows and voltage violations, posing a significant threat to the safe operation of DNs. Achieving the minimum curtailment of DPV in DNs is one of the most direct and cost-effective means to ensure their safe operation and effectively utilize renewable energy sources under existing conditions. Introducing the concepts of PV moments and load moments, the moment difference analysis theory (MDAT) for DNs with DPV is proposed. This theory transforms the integration challenge of DPV into a problem of balancing the moment difference (MD) equations for power restoration and maintenance. For a given DN, when the highest node voltage reaches the specified voltage limit, the MD, defined as the difference between PV moments and load moments, approximates a constant known as the critical moment difference (CMD). This CMD is determined by the topological structure and line parameters of the DN, independent of load distribution and PV deployment. The theoretical derivations and case studies confirm this concept. The CMD represents the limit of a DN’s capacity to integrate DPV. The DPV moment is the quantity that the DN needs to accommodate, while the load moment serves as the resource for accommodating the PV moment. Based on MDAT, a method for the minimum curtailment of DPV in DNs is proposed and applied to the analysis and calculations of a 10 kV feeder line at Sichakou of the State Grid Shandong Electric Power Company and the 12.66 kV IEEE 33 bus and 69 systems. The case studies demonstrate that, compared to traditional particle swarm optimization (PSO) methods, the minimum PV curtailment strategy presented in this paper increases optimization speed by 4736.82 times under an error margin of 0.6 %. This validates the correctness and rapidity of the method, making it suitable for real-time optimization and scheduling for minimum PV curtailment in DNs.

Composite fibers have attracted significant attention due to their distinctive structural properties and performance characteristics. Currently, the focus of research lies in innovating composite spinning technology and optimizing mechanism parameters. Among these techniques, rotary jet spinning has emerged as a notable method for cost-effective and high-speed production of composite fibers. Rotary jet spinning utilizes centrifugal force to stretch two polymer solutions from a composite droplet into a composite jet, with this morphological transformation occurring within the micro-triangle. The continuous and stable stretching motion within the micro-triangle directly influences the velocity distribution and flow stability of the composite solution, ultimately impacting the morphology and quality of the resulting composite fibers. This article presents the utilization of a rotary jet spinning device for fabricating composite fibers. It provides an in-depth analysis of the cone formation mechanism in composite spinning solutions and establishes a theoretical model for micro-triangle motion. The flow field of the micro-triangle is simulated using finite element simulation software, investigating how rotation speed and solution concentration impact PEO/PVP composite fiber morphology. Experimental results validate both theoretical modeling accuracy and numerical simulation, demonstrating that adjusting relevant parameters can control composite fiber morphology and structure.

Wireless power transfer (WPT) technology can completely isolate the equipment from the power supply, and has been widely used. However, the transmission efficiency of WPT system is optimal only as the transmitting coil was completely aligned with the receiving coil. But in practical applications, the mobile devices are all dynamic system, and perfectly alignment of the coils is a very demanding condition. In order to solve this problem, the paper proposed a coil position adaptive adjustment control scheme of WPT system with lateral misalignment or angular misalignment. Firstly, the relationship between transmission efficiency and lateral misalignment and angular misalignment was analyzed. And the research results show that if there is a lateral misalignment between the coils, the transmission efficiency can be optimized by adjusting the angular misalignment. Secondly, based on the hill climbing algorithm, a coil position adaptive adjustment control method was proposed. Finally, a coil position adaptive adjustment WPT system experimental platform was built, and the experimental researches are carried on. The results of the experiment show that as the transfer distance is 0.10 m and the lateral misalignment is 0.12 m, the transfer efficiency can be increased from 38.38% to 55.61% through the proposed control scheme. And the results verify the feasibility and effectiveness of control scheme proposed in the paper. (c) 2024 Institute of Electrical Engineer of Japan and Wiley Periodicals LLC.

Vehicle-mounted detection methods have been widely applied in the maintenance of high-speed railways (HSRs), providing feasibility for diagnosing ballastless track arching. However, applying detection data faces several key limitations: (1) The threshold mostly requires manual setting, making recognition accuracy highly subjective; (2) the extensive workload of manual inspections makes it challenging to label detection data, hindering the application of supervised learning approaches. To address these problems, this paper utilizes the longitudinal level irregularity data obtained from vehicle-mounted detection, employing the concept of unsupervised learning for dimensionality reduction, combined with clustering algorithms and minimal label fine-tuning, to design two frameworks: the fully unsupervised framework (FUF) and the few-shot fine-tuned framework (FFF). Experiments on dynamic detection data from a Chinese HSR line were conducted, comparing the performance of data dimensionality reduction, clustering, and classification under different strategy combinations. The results show that the improved variational autoencoder significantly enhances the performance of the encoder in dimensionality reduction, facilitating better feature extraction; the FUF achieves effective clustering outcomes without any labeled samples and its adjusted rand index score exceeded 0.8, showcasing its robustness and applicability in scenarios with no prior annotations; the FFF requires only a small number of labeled samples (labeling ratio of 5%) and achieves excellent performance, with metrics such as accuracy exceeding 0.85, thus greatly reducing the reliance on labeled data. This study offers a novel method for solving engineering issues with limited labeled data, providing an efficient solution for identifying track arching defects and advancing railway infrastructure monitoring.

High-temperature heat pipes are often used as heat exchangers in nuclear reactors because of their remarkable advantages in terms of thermal conductivity, isothermal properties, and self-actuation. To ensure the safe operation and efficient heat transfer of a reactor system, it is necessary to analyze the transient heat transfer performances of heat pipes. The working temperature range of the high-temperature heat pipe is 800-1,300 K when sodium is selected as the working fluid. A network system consisting of the thermal resistances can be used to analyze heat pipe transients. For a high-temperature thermosiphon, the existence of a liquid pool and a liquid film will affect the overall thermal resistance and temperature distribution. In this paper, a thermal resistance network model including convection and phase changes was established, a linear differential equation was established based on the energy conservation equation, and code was developed based on this differential equation to solve the thermal resistance and temperature field equations of the heat pipe. It was found that the temperature of numerical calculation in each region was in good agreement with the experimental results. To simulate a complex reactor environment, a power wave was introduced to explore the influence of the heat source on the thermal resistance and temperature field. It is found that with the increase of heating power, the temperature of each region of the heat pipe also increases, and the time required for the heat pipe to reach equilibrium decreases. In addition, the variations of the thermal resistance and temperature field with different heat pipe sizes and cooling modes at the condensation section under variable-power conditions were explored. A thermal resistance network model including the liquid pool of the evaporation section and the liquid film of the condensation section was constructed.Numerical code was developed to solve the thermal resistance and temperature field equations of the heat pipe.A power wave was introduced to explore the influence of the heat source on the thermal resistance and temperature field.

Real-time monitoring and analysis of sensitive areas in high-speed railway (HSR) are crucial for ensuring the safe and smooth operation of high-speed trains. To address the problem of frequent missing and false alarms caused by anomaly data in HSR monitoring system, this study proposes an innovative network framework: Intelligent detection network of HSR anomaly monitoring data (HSRA-Net). The framework comprises of two modules: the data augmentation module and the anomaly detection module. The data augmentation module designs multiple alternative generative adversarial networks for sample augmentation. To achieve the end-to-end classification, the anomaly detection module improves the residual network by creating a deep residual shrinkage network with self-attention (DRSN-SA). An online monitoring system was installed and operated continuously for several years on a high-speed turnout of a continuous beam bridge to validate the proposed framework. The collected data includes displacement, stress, and temperature. The proposed framework has demonstrated excellent performance, generalizability, and deployability through sufficient model comparison. It can accurately and efficiently diagnose anomalies in the operation of the monitoring system. This study is of great significance for improving the anomaly detection task of the HSR monitoring system.

The integration of a high proportion of distributed PV into distribution networks can cause power backfeeding and voltage limit violations. This paper introduces moment difference analysis theory for distribution networks, reframing PV consumption as the balancing moment difference equations. When maximum node voltage approaches the upper limit, the moment difference between PV and load stabilizes to an approximately constant, termed the standard moment difference. The standard moment difference represents the limit of the network's capacity to consume distributed PV. Essentially, the PV moment is the target for integration, while the load moment serves as the resource to consume the PV moment. Based on this theory, a method for energy storage configuration is proposed. Simplifying a complex multi-branch distribution network into single-branch lines and solving linear equations determines the optimal storage configuration. This method was tested on the Jibei Power Grid and the IEEE 33 and 69 bus systems, increasing computational efficiency by 1611.47-4973.82 times compared to the PSO algorithm, while maintaining an error margin of less than 2.6 %. Furthermore, the discharging strategy reduces the ESS capacity to between 12.39 % and 31.69 % of the original estimate, thereby enhancing both the economic and practical appeal of the approach.

During the regular service life of high‐speed railway (HSR), there might be serious defects in the concrete slabs of the infrastructure systems, which may further significantly affect public transportation safety. To address these serious issues and fulfill the regular functions of HSR, the traditional methods for railway engineers involve carrying out regular on‐site inspections manually or by semi‐automatic inspection vehicles, and conducting timely corresponding repairing approaches and maintenance, where these methods are time‐consuming and dangerous. In recent years, machine learning methods have been widely applied to the intelligent and automatic detection of severe defects in HSR. Currently, one of the most serious problems is the lack of sufficient high‐quality data for model training, resulting in low recognition accuracy in HSR defects. To solve this problem, this paper proposed an intelligent recognition of defects in concrete slabs of HSR based on a few‐shot learning model, that is, an artificial intelligence model based on limited data size, which recognizes three service conditions of concrete slabs in HSR: cracks, track board gaps, and unbroken state. Lightweight few‐shot learning models specifically designed for HSR detection were proposed. Experiments were conducted to compare the performances of different lightweight‐designed models, including accuracy, parameter quantity, and testing time. Results showed that the optimum model can fast and satisfactorily recognize the defects in HSR with a very limited data size of 10 samples for each training category, with a satisfactory accuracy of 73.9% in the test dataset with 20 samples for each category, parameter amounts of 2.8 million, and a testing time of 2.2 s per image. This study provides a reference for the automatic recognition of defects in HSR by railway engineers with insufficient samples.

Dynamic wireless power transfer (DWPT) can solve the limitation of battery capacity of electric vehicles, and has a good application prospect. In order to solve the problem of weak coupling power supply in a DWPT system with segmented coil mode, an improved real-time strong coupling coil arrangement mode is proposed in this paper. First, the characteristics of the typical coil arrangement mode is analyzed by establishing a DWPT model, in which the mutual inductance decreases sharply in a certain range. In order to ensure the stability of mutual inductance, an improved long receiving coil arrangement mode is proposed, and its advantage of the low coil cost is analyzed theoretically. Further, the optimal process of the improved coil mode is designed based on the genetic algorithm, and an application case of the electric bus is given. Finally, the feasibility of the coil design process is verified by simulation, and the performance advantages of the proposed scheme are verified by a 30 kW WPT platform.

Compensation capacitors are naturally susceptible to manufacturing defects and aging effects, leading to the degraded performance of a wireless power transfer (WPT) system. This article focuses on the compensation parameters optimization during the design stage and control strategy during the operation phase to improve the inherent capacitor error tolerance of the WPT system. The Sobol sensitivity method is applied to rank the importance of deviations of three capacitors on the transfer characteristics, and then the method of tracking the secondary resonance frequency is proposed. The numerical method is applied to find the optimal compensation parameters, with the constraint that the output voltage change caused by the shift of the designed compensation condition is limited to be less than ±5%. Experimental results show that with the proposed frequency tracking method and compensation parameter optimization, the deviation tolerance index is decreased from 0.485 to 0.363, showing an improvement of 25.2%, and the minimum power factor is increased from 0.78 to 0.89. Besides, the characteristics of constant primary coil current and voltage gain are almost not affected.

In order to increase distributed photovoltaic (PV) utilization, help energy green low-carbon transformation, and ensure safe and stable distribution network operation, based on the analysis of moment difference, this article firstly proposes the distributed PV consumption index of the distribution network. Secondly, it makes full use of the optimized interaction of the source-load-storage resources, and establishes a bi-level planning model of the distribution network that includes a high percentage of distributed PV. The planning layer takes into account the expansion of wire diameter and energy storage planning, and establishes a multi-objective optimization model with the minimum annual comprehensive cost, the minimum carbon emission, and the maximum PV consumption. The operation layer optimizes the distribution network operation state with the aim of minimizing network losses and realizing source-load-storage interaction by taking advantage of the regulation capability of flexible loads. The proposed model is solved using the improved multi-objective beluga whale optimization (MOBWO) algorithm and with CPLEX solver, and the final planning scheme is determined based on the fuzzy set theory for the resulting non-inferior solution set. Simulation and analysis on the IEEE33 node distribution network system verifies the correctness and validity of the proposed bi-level planning model.