Accurate prediction of structural dynamic responses is critical for seismic analysis and decision-making throughout the structural life cycle. While model-driven and data-driven approaches have advanced practice, reliable prediction under limited data remains challenging due to the high cost of acquisition and simulation. This study proposes a Self-Attention-Enhanced Physics-Informed Gated Recurrent Unit network, SA-PhyGRU, for efficient and accurate seismic response prediction. The proposed network integrates GRU dynamics with a self-attention mechanism to capture long-range temporal dependencies and improve computational efficiency, while embedding physical constraints to enhance fidelity and generalization. Numerical and experimental validations on a three-story frame and a California hotel building show that SA-PhyGRU consistently outperforms conventional baselines in both accuracy and runtime, achieving improvements of up to 11.6% in R2, with pronounced gains in small-sample regimes. These results highlight SA-PhyGRU as an effective and generalizable approach for structural seismic response prediction and performance evaluation.

Long-span self-anchored suspension bridges (SAS) with rigid cable towers are intricately designed to satisfy the aesthetic standards of structural design. Unlike the traditional arrangement of bridge towers and main girder, the ellipse-shaped bridge tower runs parallel to the longitudinal girder and exhibits significant variations in stiffness in the longitudinal and transverse directions. There is an internal force redistribution of the girder between the main and side spans during seismic excitations relative to static equilibrium. These unique bridge designs induce striking spatial impacts; earthquakes can potentially result in intricate and unforeseeable reactions within the entire structure. To understand this better, a shaking table test was performed on a 1/50 scale model of a long-span self-anchored suspension bridge with rigid cable towers to analyze its seismic response. The full-bridge model was tested on a dual-shaking-table system to investigate its seismic response under consistent excitation in various input directions. The test results confirmed the accuracy of the test model based on its comparison with a finite element model (FEM). The main tower exhibited the most significant transverse displacement change at the top relative to the base, with the largest strain responses observed near its waist. The vertical acceleration peak of the bridge was substantially greater than that in the horizontal direction owing to the large main span and the involvement of the third mode in the bridge model. Furthermore, the bridge model experienced a higher average variation in the suspender force under longitudinal and vertical excitations than under transverse and vertical excitations. These study findings provide insights into improving the structural designs of modern SAS, thereby ensuring better resistance to earthquake excitations.

Accurate assessment of the stator-winding insulation condition in high-voltage machines is crucial for safe operation. Conventional evaluation methods treat the insulation as spatially uniform and provide only a holistic assessment; in practice, however, degradation within the insulation wall is non-uniform, with the inner layers adjacent to the copper conductor typically deteriorating more severely. As a result, existing diagnostic approaches cannot accurately evaluate non-uniform degradation. To address this gap, this paper proposes a diagnosis method for non-uniform aging of stator-winding insulation based on multi-indicator fusion of dielectric spectroscopy. A Frequency Domain Spectroscopy (FDS) simulation model that explicitly incorporates non-uniform aging is established and experimentally validated. The validated model is then used to generate FDS responses under various non-uniform-aging scenarios, from which multiple characteristic indicators that quantify the degree of non-uniformity are extracted. Building on these indicators, a non-uniform aging evaluation method using Multi-Anchor TOPSIS is constructed for diagnosing non-uniform aging.

Quantitative evaluation of fatigue-induced microcracks in metal structures remains challenging. The nonlinear acoustic parameters with high sensitivity to structural cracks are the basis for this study. In this article, a novel Vold–Kalman filter-guided nonlinear vibro-acoustic modulation technique is proposed for the quantitative evaluation of microcrack damage. A new nonlinear acoustic parameter γ is introduced as the damage index, which reduces dependence on complex sideband analysis while incorporating the effect of the filter weight factor on the filter performance. The optimal weight factor is determined through application to specimens containing microcracks. Using low-frequency pumping and high-frequency probing ultrasonic excitations in a VAM framework, the proposed method quantitatively evaluates specimens with microcracks of different widths and depths. Both numerical and experimental results confirm that the proposed parameter γ decreases with the increase of microcrack width, and increases with the increase of microcrack depths. This behavior stands in sharp contrast to the fast Fourier transform-based index β and the Hilbert transform-based index α , as the proposed parameter γ effectively eliminates frequency modulation interference while providing significantly higher reliability and sensitivity for the quantitative assessment of microcracks. The proposed Vold–Kalman filter-based approach provides a more accurate and robust tool for microcrack evaluation and holds strong potential to advance quantitative structural integrity assessment.

Rapid portfolio-level post-earthquake bridge damage assessment is essential for restoring transportation networks and supporting emergency decisions, yet the current approaches face two persistent barriers. High-fidelity nonlinear analyses are computationally prohibitive at portfolio scale, while data-driven classifiers have been hindered by scarce and severely imbalanced damage-state labels, especially for severe damage. To address these gaps, this study presents a computational model consisting of two coupled components: a recurrent neural network surrogate for efficient response prediction and a gradient boosting-based classifier for damage-state identification. The surrogate learns the mapping from ground-motion features and bridge parameters to curvature-related response quantities, which are then used as physically meaningful features for classification. The classifier prioritizes the informative samples based on the predictive uncertainty and incorporates imbalance-mitigation strategies during training. Across five representative bridge types, the proposed model achieves test accuracies of 87.5%-96.5% for type-specific training and 83.1%-89.8% for joint cross-type training, demonstrating strong cross-type generalization.

Crack detection is an important measure in the field of structural health monitoring. However, visual crack detection is labor-intensive, time-consuming, inefficient, and expensive. Although image-based detection and processing provides an efficient way for structural crack detection, its accuracy depends on image quality. For engineering structures, especially bridges, the change of light conditions and the difference of surface characteristics of structural components pose a major challenge to traditional crack detection methods. In this paper, a novel crack detection method based on convolutional neural networks is proposed. The development of this method is divided into the following stages. The initial automated crack classification is carried out by using MobileNetV3, and then the improved DeepLabv3+ network is used to segment the classified crack image semantically accurately. Finally, the real crack image is used for verification. To verify the proposed method, several conventional deep learning networks are trained and compared. The improved DeepLabV3+ integrates MobileNetV3 as its feature extraction backbone and incorporates the convolutional block attention module, which achieves 87.79% average intersection and 93.87% average pixel accuracy on public and real data sets. Compared with traditional models such as VGG16, the proposed method shortens the training time by more than 80% while maintaining high detection accuracy. In addition, the compact parameter configuration and moderate model size make it particularly suitable for deployment on mobile detection devices.

Expansion joints (EJs) are critical components of a bridge to accommodate the temperature-induced movements and prevent structural damage. Predicting the EJ displacements and providing early warnings are crucial to the maintenance and safety of bridges. This paper presents a novel probabilistic framework to predict the EJ displacements, integrating a recurrent mixture density network and Bayesian linear regression. This approach addresses the inherent uncertainties of the measured structural temperatures and linear regression parameters through robust simulations. The Monte Carlo simulation can effectively evaluate the marginal posterior distribution of the EJ displacements. This framework not only derives the critical parameters from the simulations, but also provides the probability distributions associated with the random forecasting errors under significant temperature variations. The recurrent mixture density network, Bayesian linear regression and the combined models, upon examination with different evaluation indicators, prove that the models work well in predicting the probability distributions. The reliability and anomaly indices obtained show that this innovative methodology can provide precise and probabilistic estimation of the factors governing the EJ displacements for steering the early warning systems.

Traditional techniques for detecting internal defects in concrete are limited by the weak directivity of ultrasonic waves, significant signal attenuation, and low imaging contrast. This paper presents an improved synthetic aperture focusing technique (SAFT) enhanced by the Delay Multiply and Sum (DMAS) algorithm to address these limitations and improve both the resolution and signal-to-noise ratio. The proposed method sequentially transmits and receives ultrasonic waves through an array of transducers, and applies DMAS-based nonlinear beam-forming to enhance image sharpness and contrast. Its effectiveness was validated through finite element simulations and experimental tests using three precast concrete specimens with artificial defects (specimen size: 240 mm × 300 mm × 100 mm). Compared with the conventional SAFT, the proposed method improves image contrast by approximately 40%, with clearer defect boundaries and a vertical positioning error of less than ±5 mm. This demonstrates the method’s promising potential for practical applications in internal defect visualization of concrete structures.

Fatigue crack detection is an important issue to ensure the safety of steel strands. The key to solve this problem is to extract the nonlinear response generated by fatigue cracks. In this paper, the nonlinear VAM method is used to detect the structural cracks. The structure with fatigue cracks is excited using low-frequency pumping and high- frequency probing, and the power spectrum analysis of the modulated signal is carried out. In view of the shortcomings of spectral analysis, a new method combining S-transform and bispectrum is proposed, which is called S-transform bispectrum. S transform contains the phase factor, which can retain the absolute phase characteristics of each frequency, and has good time-frequency multi-scale focusing performance. The bispectrum can suppress Gaussian noise, retain phase information, and quantitatively describe the quadratic phase coupling in the signal. Then the simulation and experiment of damaged straight rod, helical rod, and steel strands are carried out. The results show that the proposed method can effectively detect the nonlinear features by using the sideband peaks in the S-transform bispectrum three-dimensional plot, and the nonlinear features are important to identify the structure with damage. At the same time, in order to prove the ability of S-transform bispectrum, an S-transform bispectrum detector is used to verify it, which is superior to the spectrum in terms of its ability to localize modulation sidebands. The proposed S-transform bispectrum has a good application prospect and provides a new tool for structural damage detection.

Post-earthquake damage assessment of bridge portfolios faces challenges in balancing efficiency and accuracy. While machine learning offers a promising solution, it often faces difficulties due to complex damage scenarios and imbalanced data, where safe samples vastly outnumber those requiring inspection or classified as dangerous. This imbalance limits the performance of machine learning models. To address these issues, this study proposes a rapid seismic damage assessment method for bridge portfolios using a SMOTE-enhanced XGBoost model. By utilizing bridge parameters and seismic data, this method eliminates the need for time-consuming on-site inspections, enabling quick and accurate predictions of bridge safety and traffic capacity. SMOTE is used to generate synthetic samples for all categories, with sampling ratios carefully adjusted to determine the optimal configuration that maximizes accuracy and improves data balance. The enhanced XGBoost model is then trained on the balanced dataset to classify bridge conditions, effectively mitigating the impact of class imbalance. The study also compares the performance of XGBoost, Random Forest, and AdaBoost on both original and SMOTEbalanced datasets. Results indicate that the SMOTE-enhanced XGBoost model achieves a best accuracy of 83.3 %-87.5 %. This study integrates oversampling techniques with machine learning and proposes an automated modeling framework for imbalanced data. Its key innovation lies in the automatic determination of the optimal oversampling ratio, which is incorporated into the XGBoost training process to significantly improve classification performance and generalization ability, providing reliable support for emergency response and functional recovery.

The Bayesian framework in structural health monitoring includes both modal identification and model exploration. Probabilistic model exploration, also named as model updating, can effectively estimate the structural parameters and quantify their uncertainties. However, it can be computationally intensive on application to real-world large-scale structures. Meta-models, e.g. Kriging models, can help tackle this challenge but they also introduce more uncertainties. In this paper, a novel Bayesian framework combining the active learning Kriging approach is proposed. The framework comprises three major components: the improved fast Bayesian spectral density approach for modal identification, the active learning Kriging method for meta-modelling, and the Bayesian structural model exploration. The Transitional Markov Chain Monte Carlo algorithm is implemented throughout the framework to sample the posterior distributions. The uncertainties from three aspects, i.e., (1) measurements, (2) meta-model construction and (3) finite element modelling, are considered in definition of the likelihood function adopted in both the active learning and model exploration processes. Compared with the ordinary Kriging model and adaptive Kriging approach using U function, the proposed active learning method significantly reduces the uncertainties of the Kriging predictor and improves its local prediction performance with fewer samples. The proposed framework is validated by a continuous test beam in the laboratory and applied to a real-world cable-stayed bridge using structural health monitoring data. A mode-matching criterion is used to overcome the difficulty of closely spaced modes in model exploration of the cable-stayed bridge. As the proposed framework is data-driven, no weighting hyperparameters are required. The active learning Kriging-based Bayesian framework can directly process structural dynamic time history response and conduct probabilistic model exploration with multiple uncertainties included, and therefore is promising in application to major structures.

Due to material degradation and inherent structural weaknesses, pagodas often suffer severe damage in earthquakes. Conducting an accurate and efficient pre-earthquake fragility assessment of these pagodas is a key task within the performance-based seismic engineering framework. Therefore, this paper proposes a Width-tunable Basis Spline-Partial Least Squares-Response Surface Model (WiBS-PLS-RSM)-based framework for seismic fragility analysis of ancient masonry pagodas. In this framework, WiBS-PLS-RSM introduces the width parameter and jointly optimizes it with the number of segments to solve the overfitting or underfitting issues of traditional response surface model (RSM) in seismic fragility analysis under high nonlinearity, high dimensionality, and strong parameter correlations. The workflow is detailed and validated through a case study of the Guang Pagoda, including surrogate modeling of the means and standard deviations of damage indicators, derivation of fragility curves, and SHAP-based sensitivity analysis of collapse fragility. The results show that the prediction accuracy of the WiBS-PLS-RSM for the mean and standard deviation of the structural damage indicators is superior to that of the traditional RSM; determination coefficients R2 increased by 3.8%-5.8% and 5.6%-6.4% respectively, thereby constructing an accurate seismic fragility curve. The proposed framework accurately resolves the magnitude and direction of parameter influences on collapse probability across different damage indicators.

Long-span suspension bridges are widely adopted in seismic-prone regions due to their excellent seismic performance. However, their extensive spans often necessitate crossing active fault zones, where cross-fault earthquakes impose unique ground motions on structures, causing responses that differ significantly from those caused by typical earthquakes. This study investigates the seismic performance of a long-span self-anchored suspension bridge under cross-fault ground motions through shake table tests. A 1/50 scale model of a 1,150-meter prototype bridge was constructed and tested using synthetically generated uniform excitation waves and recorded crossfault ground motions at two amplitude levels. The experiments examined the effects of two distinct fault types and the vertical ground motion reduction. The results demonstrate that cross-fault ground motions, particularly pulse-type motions, induce heightened localized responses at critical bridge sections compared to uniform artificial waves. Furthermore, the bridge exhibited significant differences under the reduced and non-attenuated vertical ground motions. These findings improve the understanding of the seismic behaviours of long-span selfanchored suspension bridges subjected to cross-fault earthquakes and highlight the importance of considering cross-fault earthquakes in the seismic design of long-span bridges traversing active fault zones.

Under continuous environmental erosion and long-time high tension, fatigue micro-cracks of bridge cables appear frequently to challenge the reliability of cables. The vibro-acoustic modulation analysis is adopted because of its sensitivity to contact-type defects, especially fatigue cracks. In this paper, using multiple sets of low-frequency vibration and high-frequency ultrasonic excitation, the propagation of the ultrasonic wave in the steel strands is studied. Then, the modulated signal bispectrum is proposed to improve the detection reliability of nonlinear modulation components in the nonlinear vibro-acoustic modulation signal of the contact-type defect. The numerical simulations and laboratory tests of straight wire, helical wire, and multi-wire strands are carried out for micro-crack detection to verify the robustness of the proposed method. Compared with the traditional method, the modulated signal bispectrum can still produce clearer peak detection under high noise conditions. The results show that it provides a promising, dynamic and rapid cable defect detection technique.

This study introduces an active learning-guided online cable force monitoring system based on a modified S-transform approach, addressing key challenges in real-time system identification: complex non-stationary excitations, computational efficiency, and robustness. The framework identifies cable tension during non-stationary wind loads, significantly improving accuracy and efficiency via an extended active learning Kriging method. It effectively detects potential outliers, identifies the cable's fundamental frequency using a data fusion technique, and calculates real-time cable force and tensile stress with empirical formulae. A comprehensive analysis, including numerical and sensitivity studies, shows an error rate of less than 4 % in all cases, proving the proposed framework's superior accuracy, efficiency, and robustness compared to traditional methods. Laboratory validations using cable test data and Jiu Zhou Bridge data demonstrate the system's stability, even under extreme conditions, such as during Super Typhoon Mangkhut, providing a reliable solution for real-world cable force online monitoring.

The deterioration of bridge networks poses a major threat to the availability and function of transportation systems and ultimately affects social development. Deep reinforcement learning is expected to provide intelligent decision support for bridge network maintenance. However, existing studies have neglected to explicitly consider the impact of maintenance behavior on the cost-effectiveness of bridge networks. The complex traffic environment and the interconnection of bridge networks also pose unique challenges in balancing maintenance costs and benefits. It is necessary to explore how to use the specific traffic data of each bridge in the bridge network to effectively balance cost-effectiveness and rationalize maintenance decisions. Aiming at the maintenance requirements of the bridge network, a multi-agent ranking proximal policy optimization framework is proposed. The performance of the proposed framework is rigorously evaluated using a real bridge network example. The results show that the maintenance policy based on the proposed framework can maximize the cost-effectiveness of the bridge network in its life cycle, effectively reduce the excessive risk cost and achieve a harmonious balance between different costs. In addition, the proposed framework is superior to the traditional maintenance policy and provides higher performance and efficiency.

On the basis of the steel-concrete composite orthotropic bridge deck, corrugated steel plate is introduced as the longitudinal bottom plate, onto which shear connectors are welded to form a new type of composite bridge deck, namely corrugated steel-concrete composite bridge deck. The integration of the steel sheeting and concrete together can make full use of the performance of steel in carrying the load. This paper describes an investigation of a real bridge in Anhui Province, China, focusing on the load-carrying capacity of the corrugated steel-concrete composite deck and the failure mode of the shear connectors. First, the mechanical properties of a shear connector welded onto a corrugated steel-concrete composite bridge deck are simulated and explored experimentally. A set of comparative tests on three groups of push-out specimens are designed and carried out. The test results show that the ultimate load-carrying capacity of the corrugated steel-concrete composite deck is increased by about 45% as compared with the standard specimens and the slip at the interface is reduced by about 34%. A comprehensive parametric study of the three sets of push-out tests is then conducted using the nonlinear finite element method. The parametric analysis shows that the ultimate shear capacity of the shear connector increases with the strengths of the concrete and corrugated steel plate as well as with the diameter of the stud, albeit with diminishing return.

Rapid post-earthquake structural assessments are essential, as they significantly contribute to community recovery and enhancing seismic resilience. This research introduces a novel vision-based approach for rapid damage inspection of building facades affected by earthquakes. It employs structural point cloud models, derived from unmanned aerial vehicles (UAVs) imagery, to detect surface damages like spalling and cracks on facades. The developed algorithm effectively identifies these damages through a strategic consideration of three critical features: depth, grayscale, and local Principal Component Analysis (PCA). The study explores the impact of integrating these features on the accuracy and efficiency of damage segmentation. By adopting two real-world seismic-damaged buildings, the algorithm achieved a relatively high segmentation precision for wall cracking and spalling. Furthermore, the research provides a detailed assessment of wall damage, including quantifications of damage distribution, damaged wall areas, crack skeletons, and overall dimensions.

The ultrasonic guided wave technique is extensively used for nondestructive structural testing, and one of the key steps is to extract a single mode with certain purity from multi-order mixed modes. In this paper, the propagation of ultrasonic guided waves in the cylindrical rod is simulated first; the appropriate broadband excitation signal is selected to excite the multi-order modes in a specific frequency range; and the time–space signal containing multi-order modes is converted to the frequency-wavenumber domain signal by two-dimensional Fourier transform. In the frequency-wavenumber domain, the frequency-wavenumber ridge is extracted from the multi-mode frequency-wavenumber domain based on the dynamic programming method, and then the time-domain signal corresponding to a single mode can be reconstructed. By comparing the excited multi-order mode and the separated single mode with the theoretical results, it is observed that the two results are consistent. Thus, the employed mode-excitation method can accurately excite the multi-order modes in rod structures. Furthermore, the proposed method enables the separation of a single-mode wave with high purity, providing a foundation for future utilization of isolated modes.