Fire incidents, especially caused by tanker trucks, pose serious risks to suspension bridges and greatly endanger public safety. Current approaches for heat transfer between tanker truck fires and the structural elements of suspension bridges are time-consuming and require expertise. Therefore, a relatively simpler for application, yet reliable heat transfer method is required for fire performance analysis. This paper presents a new heat transfer method called the ‘Curtain Method’. The method was formed based on Adiabatic Surface Temperature (AST) and gas temperatures obtained by validated Fire Dynamics Simulator (FDS). A total of eight tanker fire cases with two different spill areas and heat release rates have been modeled in a windless environment. The AST and gas temperatures have been determined from a large number of locations in planes to create contour plots, ‘curtains’, at several distances away from tanker truck fires to be used in subsequent thermal analysis of suspenders and main cables. The application of the presented method is demonstrated through a suspension bridge. The temperature distributions in the suspenders and main cables obtained by the presented method are compared with the conventional method using radiative heat flux and gas temperatures by modeling the suspension bridge in FDS. The temperature distributions obtained by the Curtain Method present a good correlation with those obtained by the conventional method. The Curtain Method is promising to standardize the heat transfer independent of bridge geometry without the need for complicated computational fluid dynamics models.

Tuned mass dampers (TMDs) are extensively employed in vibration control of long-span bridges to suppress excessive vibrations. However, their performance may gradually degrade over time in harsh working environments, and thus their dynamic behaviour should be re-evaluated timely. This study focuses on the evaluation of TMDs installed in the Chongqi Bridge, which connects Chongming Island and Qidong City in eastern China. Modal parameters of the main girder are identified by analysing the ambient vibration response of the main girder acquired from the structural health monitoring (SHM) system. An innovative evaluation index of TMD working performance is proposed, and its variation pattern relative to wind speed is investigated. On-site free vibration tests are conducted to obtain frequencies and damping ratios of TMDs, and theoretical performance indices of TMDs are numerically studied by the finite element method (FEM) for comparative analysis. It is demonstrated that the damping ratio of the main girder is significantly higher than the design requirement, confirming the effectiveness of TMDs. Working performance of TMDs can reach the theoretical level at high wind speed, with their evaluation indices consistent with FEM results. Natural frequencies of TMDs could be adjusted to achieve better TMD working performance.

This study presents a high-performance, fully parallelized computational framework for stationary and non-stationary stochastic wind field simulation that reconciles the conflict between computational efficiency and spectral fidelity. The framework reformulates the classical Shinozuka spectral representation method into a dual-level parallel algorithm that (i) eliminates nested spatial loops for computing pairwise spectral densities with array broadcasting and (ii) parallelizes frequency-wise synthesis via vectorized mapping, transforming the original sequential complexity into fully parallelizable operations. Unlike existing computational acceleration techniques that rely on dimension reduction or frequency interpolation, this reformulation maintains complete spectral fidelity without introducing approximation errors. To bound peak memory use while preserving full spatial coherence, the framework integrates an adaptive frequency batching strategy for large-scale problems. A multi-backend implementation (JAX, PyTorch, NumPy) enables efficient deployment on CPUs and GPUs. Extensive benchmarks show significant performance improvements, achieving speedups of up to several thousand-fold compared with conventional methods and demonstrating superiority over other mainstream algorithms. Practical applications to several engineering case studies, with comparisons to field measurements and theoretical models, validate the accuracy and reliability of the simulated wind fields. This software is released open-source on GitHub to support reproducible research and inspire further innovations in wind engineering.

Physics-informed neural networks (PINNs) represent a promising paradigm for solving partial differential equations (PDEs) by integrating physical laws into the learning process of neural networks. However, ensuring that such frameworks fully reflect the physical structure embedded in the governing equations remains an open challenge. This stems from a fundamental limitation: standard neural networks lack an inherent notion of physical admissibility, leaving them vulnerable to producing mathematical artifacts that satisfy the training loss but lack physical meaning. In this work, we address this issue by introducing a simple, generalized, yet robust irreversibility-regularized strategy that enforces hidden physical laws as soft constraints during training, thereby recovering the missing physics associated with irreversible processes in the conventional PINN. This approach ensures that the learned solutions consistently respect the intrinsic one-way nature of irreversible physical processes. Across a wide range of benchmarks spanning traveling wave propagation, steady combustion, ice melting, corrosion evolution, and crack growth, we observe substantial performance improvements over state-of-the-art PINN baseline equipped with advanced training techniques, demonstrating that our regularization scheme reduces predictive errors by up to more than one order of magnitude, while requiring only minimal modification to existing PINN frameworks. Physics-informed neural networks (PINNs) are a promising paradigm for solving partial differential equations. In this study, an irreversibility-regularized physics-informed neural network enforces hidden thermodynamic irreversibility as a soft training constraint; across five benchmarks spanning combustion, melting, corrosion, fracture, and wave propagation, the method reduces predictive errors by more than an order of magnitude with negligible additional computational cost.

As the primary joining method, welding inevitably introduces residual stresses, which may accelerate localized pitting corrosion and thereby compromise structural durability and safety. To investigate the influence mechanism of welding residual stress on pitting corrosion evolution, this study develops a thermo-mechanical-electrochemical multiphysics-coupled phase-field model. First, a thermo-mechanical simulation of V-groove butt welding was conducted using a 6200 W moving heat source traveling at 1.0 m/min to generate the initial residual stress distribution. This field is then adopted as the initial mechanical state in the phase-field corrosion simulation. Subsequently, the effects of residual stress level, initial pit morphology, and multiple-pit coexistence on corrosion evolution are systematically investigated. The results show that tensile residual stress promotes pitting corrosion, exhibiting a threshold effect. When the residual stress exceeds 15% of the material’s yield strength, the corrosion depth increases by more than 40% compared to cases below this threshold. Under a static residual stress field, the corrosion morphology tends to become smoother. The initial defect geometry has a limited effect on the final corrosion damage. In the presence of multiple pits, the stress and ion concentration fields from adjacent pits overlap. It primarily promotes lateral expansion and area growth, with only a minor effect on vertical depth. The proposed phase-field framework reveals the underlying mechano-corrosive synergy between welding residual stresses and pitting corrosion. It thereby offers both a theoretical foundation and a numerical tool for assessing the long-term performance of welded steel structures in corrosive environments.

This study investigates the wind field characteristics of long-span suspension bridges in mountain valleys terrain, with a particular focus on the disturbance effects caused by bridge structure on wind measurements. Field data are collected using the Wind3D 6000 LiDAR installed near the bridge. By comparing wind field characteristics before and after bridge completion, this study evaluates the influence of the bridge structure on both mean and turbulent wind characteristics. The findings show that the presence of the bridge tower and deck reduces the measured mean wind speed and modifies its probability distribution. The bridge tower increases the effective ground roughness coefficient, thereby attenuating the vertical wind speed gradient. In addition, the bridge tower raises the measured turbulence intensity, alters its probability distribution, and decreases the agreement between the turbulent wind power spectrum and the von Kármán spectrum. It is necessary to correct the data affected by these disturbances to improve the accuracy of wind load assessments for long-span bridges, thus enhancing the reliability of bridge structural operation.

Predicting the microstructural and morphological evolution of materials through phase-field modelling is computationally intensive, particularly for high-throughput parametric studies. While neural operators such as the Fourier neural operator (FNO) show promise in accelerating the solution of parametric partial differential equations (PDEs), the lack of explicit physical constraints, may limit generalisation and long-term accuracy for complex phase-field dynamics. Here, we develop a physics-informed neural operator framework to learn parametric phase-field PDEs, namely PF-PINO. By embedding the residuals of phase-field governing equations into the data-fidelity loss function, our framework effectively enforces physical constraints during training. We validate PF-PINO against benchmark phase-field problems, including electrochemical corrosion, dendritic crystal solidification, and spinodal decomposition. Our results demonstrate that PF-PINO significantly outperforms conventional FNO in accuracy, generalisation capability, and long-term stability. This work provides a robust and efficient computational tool for phase-field modelling and highlights the potential of physics-informed neural operators to advance scientific machine learning for complex interfacial evolution problems.

Physics-informed neural operators (PINOs) aim to learn solution operators for partial differential equations by using the governing physics as supervision, rather than relying solely on paired input-output simulation data. By incorporating physical constraints into the training objective, PINOs combine the cross-instance generalization of neural operators with the data efficiency of physics-informed learning. Despite this promise, how to train PINOs efficiently and robustly remains less well-understood than the training of either data-driven neural operators or physics-informed neural networks (PINNs). To bridge this gap, we examine key components of the PINO training pipeline, including architecture design, optimizer choice, loss balancing, and collocation-point sampling strategy. We study three representative operator backbones, Deep Operator Network (DeepONet), Fourier Neural Operator (FNO), and Continuous Vision Transformer (CViT), across five diverse parametric PDE systems. Our results show that CViT provides consistently strong and stable performance across the considered benchmarks. Beyond architecture, we find that several optimization pathologies previously identified in PINN training naturally arise in PINOs, including gradient conflicts and causal violation. We also find that mitigation algorithms developed for PINNs remain effective in the PINO setting. We further compare physics-informed and data-driven training under different data regimes, revealing that a carefully designed physics-informed training pipeline can match, and in some cases, outperform purely data-driven neural operators. Taken together, these findings provide a systematic empirical understanding of the optimization challenges in PINO training and inform a practical pipeline for efficient and robust physics-informed operator learning. Code and data are available at https://github.com/NanxiiChen/PI-CViT.

Bridge vibration responses are typically the superposition of wind-induced and vehicle-induced vibration responses. However, the presence of vehicle-induced components in the vibration measurements poses a significant challenge to the accurate analysis of wind-induced vibration. To address this issue, this study innovatively introduces a hybrid deep learning model that combines Fourier feature encoding (FFE) and a convolutional neural network (CNN), and two vibration components can be separated by the model based on their distinct power spectral densities (PSDs). The model is built as a mapping from the high-frequency part to the low-frequency part of the vehicle-induced vibration PSD, and then it is trained and tested on datasets dominated by vehicle-induced vibration, exhibiting high prediction accuracy and excellent generalization capability. Subsequently, the trained model is employed on datasets of the entire monitoring period to obtain vehicle-induced vibration PSD in the low-frequency range, thereby separating out wind-induced components. Finally, wind load effects on bridge vibration at different wind speeds are investigated. Results demonstrate that the wind-induced vibration increases markedly with the mean wind speed at high wind speeds, while the wind loads primarily exert a damping effect at low wind speeds. Compared with conventional approaches, the proposed method provides more authentic and reliable wind-induced bridge vibration responses, enabling more accurate performance analysis and response prediction based on monitoring data.

With the growing span and flexibility of modern bridges, wind-induced buffeting has received increasing attention. Data-driven predictions of bridge buffeting responses are generally conducted using samples at high wind speeds, thereby mitigating interference from vehicle-induced vibration responses. Nevertheless, high wind speed samples could be scarce for some bridges, which adversely affects the performance of prediction models. To this end, this study proposes a novel deep learning model that can utilize the vehicle-interfered vibration data and predict the vibration response in the frequency domain. The model employs two encoders to extract and process features corresponding to wind-induced and vehicle-induced vibrations, and then fuses them to predict vibration responses in the low-frequency range, where the dominant part of buffeting energy is contained and two vibration components are superposed. Using monitoring data of a long-span suspension bridge, the proposed model is trained and tested on datasets with numerous low wind speed samples, and then it is evaluated on unseen strong wind samples and compared with a baseline model that is incapable of capturing vehicle-induced components. Results demonstrate that the proposed model achieves satisfactory prediction performance, and it shows greater generalization and extrapolation ability than the baseline model. The proposed method enables effective use of abundant monitoring data at low wind speeds and improves model robustness, which supports accurate predictions of buffeting responses and enhances early warning capability for bridges with insufficient strong wind observations.

Accurate characterization of wind fields is essential for effectively analyzing wind-induced responses of long-span bridges. However, field measurements often suffer from severe anomalies, compromising their reliability. One significant cause of these anomalies is the disturbances induced by the bridge deck. This study aims to eliminate the effects of such disturbances on the measured data and wind field characteristics. We begin with a comparative analysis of disturbed versus undisturbed wind measurements. To remove the disturbance effects and clean the data, we introduce a novel skip-connected variational autoencoder (SVAE) neural network that can effectively reconstruct undisturbed data from disturbed measurements. The SVAE is trained on a dataset comprising both disturbed and undisturbed measurements collected during strong wind events and is validated against other generative models based on reconstruction accuracy. We then employ the trained SVAE to clean measurements from another event and analyze wind field characteristics such as mean wind speed, turbulence intensity, integral scale of turbulence, and power spectral density from the reconstructed data. Our findings confirm the effectiveness of this approach, suggesting its broader applicability in ensuring the accuracy and reliability of wind field characterizations and subsequent analyses of wind-induced responses on long-span bridges.

Tanker trucks carrying flammable and explosive materials on highways pose risks to infrastructure, including bridges, particularly concerning parallel steel wire cables used in cable-supported bridges. To investigate the dynamic response of these cables under combined fire and explosion effects and assess their safety, a numerical simulation study was performed. The simulation’s accuracy was validated by comparing it with existing cable explosion tests. The temperature-time history of the cable was acquired after the heat transfer analysis. With a simplified constitutive model considering high-temperature and high-strain rate effects, simulations for the cable under explosions at different maximum temperatures were conducted. Finally, a damage calculation method based on critical fracture strain and effective plastic strain was developed for safety assessment. Results indicate that during a tanker truck fire, steel wire heating accelerates initially but slows near maximum temperature. High-temperature softening affects the cable's response during explosions. As temperature rises, the maximum cable force decreases, and the effective plastic strain increases. When maximum temperatures exceed 400 ℃, transverse displacement of the cable is apparently affected and the frequency decreases. The proposed damage calculation method effectively reflects damage, making the safety assessment process reasonable and effective.

With the development of hydrogen fuel cell vehicles, the possible problem of hydrogen safety is gathering increasing concerns for transportation infrastructures. This study focuses on the possible hydrogen safety issue on a bridge suspender, by assuming a scenario of hydrogen leakage and combustion when a hydrogen delivery truck is situated on a bridge. A CFD simulation is performed with respect to the process of pressure relief - diffusion - combustion of hydrogen. The results indicate that the ignition of high-pressure hydrogen leads to a jet fire, and the high-temperature airflow may be influenced by the suspender, which leads to different heat inputs at different positions of the suspender. Based on the equivalent heat fluxes near the suspender, the thermal analysis is conducted. The results show the temperature histories on different points of the suspender with significant differences. Particularly, the temperature on the front side of the jet fire may reach approximately 1050 K in around 300 s, and gradually decrease due to the less heat input caused by lower hydrogen leakage flow rate over time. For the central part of the suspender, the temperature rises in a lower rate, reaching 800 K in about 600 s. It hence describes the characteristics of hydrogen-induce jet fire on structures, which can be beneficial to the enhancement of the resilience of bridges and other infrastructures in the transportation system with respect to hydrogen safety issues.

Understanding and mitigating concrete fracture deterioration under non-uniform corrosion is essential to extending the lifespan of structural materials. This paper introduces an innovative electrochemical–chemical–mechanical phase field model to simulate and predict precipitation-induced cracking in concrete. The model uniquely integrates multiple coupled processes: electrochemical reactions, precipitation-induced pressure, transformation of distinct precipitations, concrete cracking, and porosity variations due to both precipitation and cracking. Through an automatic coupling of corrosion-precipitation-damage-diffusion mechanisms, the model captures the complex interactions where porosity changes dynamically with rust precipitation and cracking, impacting the diffusion coefficient and thus influencing fracture behavior. Rust-induced eigenstrain is correlated with precipitation volume fractions, and a phase-field-regularized cohesive zone model effectively describes concrete fracture evolution. Validated against experimental data, the model shows strong accuracy and provides insights through parametric studies on the impact of reaction rate constants, filling coefficients, and oxygen diffusion coefficients on cracking patterns. This model developed is made freely available at https://github.com/XuruiFang96 to facilitate further research and practical applications.

Cable clamp systems play an essential role in force transferring and stability of suspension bridges by linking various key elements such as main cables and vertical hangers. Cable clamps consist of multiple high-strength bolts, steel clamps, main cable, hanger rope, and a bolted pin connection linking the clamps and hanger. Failure in one of these may cause the stability of suspension bridges to deteriorate, especially in fire accidents. This study numerically investigates the performance of cable clamp systems adopted from a real suspension bridge under vehicle fire. The structural and heat transfer analysis is performed in Abaqus considering ductile fracture of the bolts at elevated temperatures. The effects of fire intensity and tension load in the cables on the fire performance are investigated by adopting different fire locations in the longitudinal direction of the suspension bridge. The findings are discussed from the points of local performance of the connection and the overall behavior of the suspension bridge.

Concrete slabs in composite bridges are inevitably subjected to heavy vehicular loads during their service life. To evaluate the fatigue performance of the prestressed concrete slabs in composite bridges, two full-scaled models of prestressed concrete slabs were first designed and tested, with the load amplitude was selected as the variable. To simulate the damage caused by the initial passage of heavy vehicles, this was simplified into the form of a static cyclic load. The mechanical deformation state and crack distribution of the slab were analyzed. Further, a finite-element model was established, and a parametric analysis based on the variation in loading form, such as monotonic displacement loading, static cyclic loading followed by monotonic displacement loading, and cyclic displacement loading, was conducted to discuss the performance-enhancement mechanism of prestressed concrete slabs. Finally, in consideration of the influence of static cyclic damage on the fatigue performance of prestressed concrete slabs, evaluation parameters were proposed to account for static cyclic damage by considering the effects of stresses in concrete, tensile rebar, prestressed tendons, and external loading. A comprehensive fatigue performance evaluation method for prestressed concrete slabs, which neglects the tensile hardening behavior of cracked concrete in the tension zone, was established and verified by test results. The results indicate that the damage caused by static cyclic loading has a significant influence on the fatigue performance of the slab. Applying prestress can significantly mitigate the influence of initial damage on the mechanical and deformation behavior of the slab, which benefits from the prestress compensating for the cracking stress at the bottom of the slab. The proposed fatigue performance-evaluation method for prestressed concrete slabs, which considers static cyclic damage, can predict fatigue deformation behavior with an error of less than 10%, while reasonably determining the fatigue life and failure modes of prestressed concrete slabs. The parametric analysis reveals that when the prestress value exceeds 9 MPa, the failure mode of the prestressed concrete slab transfers from rebar fracture to concrete failure.