Regional high-speed railway networks are lifeline corridors after earthquakes. Line functionality is constrained by bridge safety and overhead catenary current collection. Most resilience studies model bridges only, which may systematically overestimate network recovery. This study presents a rapid resilience assessment framework for coupled bridge and catenary systems. Pier drift ratio and residual displacement at catenary registration points define damage states. Speed restriction rules map damage to a functional reduction factor, and state-dependent repair times generate recovery curves and an area-based resilience index. A Copula-driven regional adaptive random-forest surrogate (ACRF) reduces nonlinear time history analysis at the regional scale. ACRF builds a regional error field in the intensity and pier height space, generates targeted pseudo-samples within high error subdomains, and jointly optimizes the Copula family and surrogate hyperparameters. A case study shows that ACRF reduces out-of-fold errors with a controlled pseudo-sample budget. The results show that catenary repair can govern the recovery process for several segments. Under the assumed repair-priority rule, it can introduce an additional waiting-induced delay of 3.9-4.8 months, so ignoring it yields optimistic network resilience. The framework supports post-earthquake prioritization and resource allocation.

As a sudden and highly destructive natural disaster, earthquakes pose a severe threat to the safe and stable operation of critical infrastructure such as power systems. To enhance the efficiency of seismic resilience assessment for substation systems and quantify the cascading economic losses, this paper proposes a framework for seismic resilience and economic loss analysis based on a sample optimization strategy. Firstly, addressing the high computational cost of traditional Monte Carlo simulations, a Bayesian optimization algorithm is introduced to adaptively determine the optimal combination of sample sizes for seismic scenarios and repair sequences. This approach significantly improves computational efficiency while maintaining statistical precision. Secondly, to overcome the limitations of existing resilience metrics relying on empirical judgments, a comprehensive resilience index based on four attributes: robustness, redundancy, rapidity, and resourcefulness is constructed. Objective weighting using the information entropy method enables multidimensional scientific quantification of the system's seismic performance. Furthermore, a chained economic loss assessment model encompassing direct equipment damage, commercial/industrial output losses, social costs, and restoration expenses was established, revealing a loss distribution pattern dominated by social costs during major earthquakes.

To address the high computational cost and numerical stability constraints of conventional finite-difference solvers for the probability density evolution equation (PDEE) in train-bridge system (TBS) stochastic vibration analysis, this study proposes a physics-informed neural network (PINN)-based framework. A train-bridge coupled dynamic model is first established, and the governing equation of motion and the corresponding PDEE are derived. These two equations serve fundamentally different purposes, with the equation of motion governing the deterministic system response and the PDEE describing the probability evolution of that response. Then, the PDEE, together with initial and boundary conditions, is embedded into the PINN loss via automatic differentiation, enabling a mesh-free solution for the time-dependent probability density function (PDF). Taking the stochastic dynamic response of a high-speed railway bridge as the target, the proposed method is validated against conventional probability density evolution method (PDEM) benchmark solutions. Numerical investigations under varying train speeds and random train-parameter conditions demonstrate that the method achieves high accuracy and stability in solving the PDEE. The main contributions are fourfold: (i) a physics-informed framework embedding the generalized PDEE into the network loss for mesh-free PDF evolution; (ii) a 42.7 % reduction in PDEE-solving cost versus finite-difference PDEM with maintained statistical accuracy; (iii) robustness across train speeds of 260–340 km/h and load coefficient of variation (CV) of 0.08–0.16; and (iv) accurate capture of non-Gaussian features and time-varying evolution for infrastructure reliability assessment.

Dynamic analysis of large-scale high-speed train-track-bridge (TTB) interaction systems is often bottlenecked by the conflict between model fidelity and computational efficiency. Such analyses typically involve long-duration train operations and require fine temporal resolution, leading to a prohibitively large computational burden. For large-scale TTB systems modeled using the unified finite element (FE) framework, standard Newton-Raphson iterations necessitate frequent re-factorization of the global stiffness matrix, resulting in substantial computational expense. To address this challenge, this study proposes a novel Localized Nonlinear Stiffness Decomposition (LNSD) method, achieving a nearly order-of-magnitude improvement in computational efficiency. Taking advantage of the fact that strong nonlinearities are mainly localized within the wheel-rail interaction (WRI), whereas the large-scale infrastructures predominantly remain linear elastic, the proposed approach employs the Sherman-Morrison-Woodbury (SMW) formula to transform nonlinear stiffness variations into equivalent corrective force vectors. Consequently, the massive global stiffness matrix is factorized only once at the initial step, and nonlinear updates are strictly confined to the WRI subsystem, yielding substantial computational savings. Numerical examples demonstrate that a 30-span TTB system achieves accuracy identical to that of the conventional FE method while attaining a 9.8-fold increase in computational efficiency. Furthermore, a 50-span TTB system subjected to a 16-s Tabas earthquake validates the applicability of the proposed method for largescale seismic analysis, with efficiency improved by approximately 14.7 times. Results also indicate that the efficiency advantage of the LNSD method becomes increasingly pronounced as the system scale increases, highlighting its suitability for long-duration dynamic simulations of large-scale railway systems.

Conventional model testing of fixed offshore wind turbines (OWTs) suffers from scaling incompatibilities, which makes realistic aerodynamic loading difficult to reproduce in the laboratory. Real-time hybrid model (RTHM) test is therefore a practical alternative, but its fidelity can be degraded by real-time imperfections such as delay and noise, so a controllable baseline is needed for quantitative assessment. This paper develops a high-fidelity co-simulation-based RTHM framework primarily as an analysis and assessment tool for RTHM test. The aerodynamic and servo-control modules of FAST are compiled into a dynamic-link library (DLL) as the numerical substructure, and an OpenSees finite-element (FE) model is used as the simulated physical substructure (SPS). Real-time data exchange enables aero-servo-elastic coupling. For the National Renewable Energy Laboratory (NREL) 5 MW wind turbine, the proposed co-simulation reproduces key OpenFAST response time histories, confirming the validity of the coupling implementation. The baseline response is then used to evaluate a laboratory RTHM test, yielding thrust errors of 6% (standard deviation), 0.04% (mean), and 0.24% (peak). Parametric studies show that delay dominates over noise, increasing geometric scale mitigates delay-step effects for a given physical delay, and multi-hazard loading amplifies delay and noise impacts, providing quantitative guidance for RTHM test design.

The railway bedrock and the tunnels beneath are often subjected to train moving loads, and occasional seismic activity. In extreme conditions, they may experience the combined effects of seismic and train moving loads, easily resulting in their dynamic damage and instability. To investigate dynamic response of bedrock and tunnel subjected to seismic and train moving loads, large-scale physical model experiments are conducted using a train-rail-shaking table test system. The experiment results show that, under seismic load, the root mean square acceleration (arms) of the model increases as the input peak seismic acceleration (PSA) increases. As the PSA increases, the acceleration levels (ALs) evolve from exhibiting a single peak to three distinct peaks, with the peak ALs decrease as the center frequency increases. Under train moving load, there is no significant change in arms of the model as the train moving speed (vtr) changes. Under the coupling effect of seismic and train moving loads, the arms of the model increase as the PSA increases. The ALs are similar across different vtr and gradually evolve from a single peak to three distinct peaks as the PSA increases. Modal analysis indicates that the natural frequencies of the model remain essentially unchanged under different loading conditions. Furthermore, at low vtr, the vibration of the model under coupling effect is stronger than or equal to that under seismic load, and the latter is generally great than that subjected to train moving load. However, as the vtr increases, the vibration of bedrock and tunnel under train moving load is stronger than that under coupling effect, except at the peak and trough frequencies, and the latter is stronger than or equal to that solely under seismic load. The research findings are of great significance for maintaining the operation safety of bedrock and tunnels.

Shaking tables are critical facilities for simulating seismic effects via ground motion reproduction. However, single-table tests are often constrained by limited platform dimensions and load capacity. While multi-table synchronization partially addresses these limitations, traditional array control methods under rigid connections face challenges, including degraded precision from synchronization errors and experimental interruptions due to output forces exceeding safety limits. To address high-precision synchronization requirements for rigid-connected dual-shaking table arrays, this study proposes an impedance-based internal force coordination control strategy. This approach enhances synchronization accuracy and helps prevent failures from excessive coupling forces. Specifically, a global simulation model and a mechanical model of the dual-shaking table array under rigid connection were established. Through simulation and experimental validation, the impact of synchronization errors was evaluated and the strategy’s efficacy was verified. Results show the strategy significantly reduces peak-force discrepancy between platforms. The method effectively circumvents experimental bottlenecks, such as output force saturation, inherently associated with rigid connections.

Yaw system failure can induce misalignment relative to the wind direction, amplifying asymmetric loads and increasing the risk of offshore wind turbine (OWT) failure. Research on OWT response to yaw failure remains limited, and is primarily based on numerical simulations. This study employs real-time hybrid model testing (RTHM) to investigate yaw failure, providing new experimental insights into this topic. Meanwhile, a boundary coordination algorithm is proposed to synchronize the control rate with the computed frequency of the numerical model in RTHM. By developing the shaking table from displacement control to force control, and using it as the actuator for aerodynamic loads, force control tests demonstrate that the actuator effectively reproduces aerodynamic load components up to at least 2 Hz, capturing both peak values and overall trends. The actuator provides high load capacity with limited sensitivity to laboratory space constraints. Subsequently, RTHM is conducted to emulate aero-servo-elastic coupling under yaw misalignment. The results indicate that RTHM effectively replicates aerodynamic loads at the laboratory scale, with the peak error between reference and measured forces limited to a of 6.85 %. Furthermore, the RTHM system's high load capacity makes it suitable for model testing of large-scale, high-megawatt OWTs.

This paper proposes a model predictive control method based on active interference suppression to address the problems encountered in real-time hybrid simulation, including interference suppression, bandwidth expansion, and time delay compensation. A modified extended state observer is employed to detect disturbances, including external interference and modeling errors in the electro-hydraulic system at various frequencies. Then, the feedback controller eliminates these disturbances, enabling the established model to adapt to various working frequencies. By effectively rejecting this disturbance, the model predictive control algorithm can be designed based solely on the nominal model within a wide frequency range, thereby reducing design difficulty. Additionally, this paper introduces adaptive amplitude compensation to overcome the overshoot problem that occurs in traditional active polynomial extrapolation methods when compensating for high-frequency signals. While compensating for signal delays, the prediction control effect is fine-tuned, and the system's operating bandwidth is further expanded. Finally, the effectiveness of the proposed method is verified through simulation and experiments.

Shape memory alloys (SMAs) are increasingly used in seismic engineering due to their excellent self-restoring capability. However, an excessive elongation demand would deteriorate this self-restoring capability. Conventional SMA dampers lack the ability to adjust strength and deformation capacity according to different seismic demands. As such, this paper proposes an enhanced SMA friction self-restoring damper (eSMA-FSRD). First, the modular assembly configuration and working mechanism of the eSMA-FSRD are introduced, and analytical expressions for its force-displacement relationship are derived. And then, material property tests are conducted to identify the optimal heat treatment process for improving the self-restoring performance of the SMA, followed by low-cycle reversed loading tests on three eSMA-FSRD specimens. Finally, finite element simulations are employed to validate the working mechanism of the eSMA-FSRD, and a predictive model for the enhancement factor is proposed. The results indicate that the eSMA-FSRD exhibits a flag-shaped hysteretic behavior under both tensile and compressive loading. Increasing the inclination angle (theta) of the dowel bars or applying preload (F0) to the driving blocks effectively enhances its energy dissipation capacity. Both theoretical analysis and numerical simulation are in good agreement with the experimental results, confirming their reliability as methods for predicting the mechanical performance of the eSMA-FSRD. Based on the enhancement factor prediction model, the eSMA-FSRD is classified into three types: deformation-enhanced, globally-enhanced, and load-capacity enhanced. This model provides a theoretical basis and parameter selection method for the application of the eSMA-FSRD in seismic design.

Railway subgrade soils are subjected to intermittent cyclic loading characterized by train-induced dynamic stress and the intermittent periods between adjacent trains. Although soil deformation under such loading has been examined, the excess pore water pressure (EPWP) response remains unclear. This study performed triaxial tests on silty filler under continuous and intermittent cyclic loading, analyzed with an energy-based method across different dynamic deviator stress levels and intermittent times. Results show that higher dynamic deviator stress levels accelerate residual EPWP accumulation, whereas intermittent stages mitigate it, with longer intermittent time leading to lower EPWP accumulation in the subsequent loading stage. Normalized dissipated energy increased with loading cycles, and under continuous loading, a unique relationship was observed between residual EPWP ratio and normalized dissipated energy, independent of stress state. Load intermittency reduces the rate of cumulative dissipated energy development, indicating more constrained particle movement and rearrangement, which slows EPWP accumulation. Moreover, greater dissipated energy is required to generate residual EPWP after longer intermittent times, suggesting a denser soil skeleton structure after longer intermittent times. Two models were developed to describe residual EPWP generation under continuous and intermittent conditions by incorporating energy-based method. These findings demonstrate the critical role of intermittent time in EPWP evolution and provide valuable insights for evaluating the performance and long-term durability of railway subgrades under realistic train-induced cyclic loading.

This study systematically investigates the dynamic response and traffic safety performance of simply supported beam bridges on mountainous high-speed railways under rockfall hazards. By coupling rockfall trajectory simulation with refined finite element analysis, the study elucidates the damage evolution mechanism of bridge structures under rockfall impacts and the correlations among key parameters. The findings indicate that the dynamic response of superstructure components (including the track system, mortar layer, and sliding layer) is a strong correlation with pier displacement and exhibits higher damage sensitivity. Furthermore, an innovative bridge fragility model accounting for rockfall mass ranges is established, enabling quantitative assessment of the damage probabilities for piers, mortar layers, and sliding layers. Building upon this foundation, a method for evaluating train safety based on track geometric deformation after rockfall impact is proposed, identifying critical impact conditions at different operational speeds. This research provides theoretical support and technical references for the anti-impact design, safety protection, and operational decision-making of high-speed railway bridges in rockfall hazard environments.