This study investigates the magneto-mechanical relationship in eight 26-year-old naturally corroded reinforced concrete (RC) girders through self-magnetic flux leakage (SMFL) tests, finite element simulations, and systematic correlation analysis. The results indicate that rebar tensile stress leads to a proportional increase in the magnetization intensity change ratio (MR) of the corroded rebar. The MR variation introduces errors into the SMFL-based assessment of the rebar cross-sectional corrosion degree (eta). When MR <= 0.2, the stress-induced mean error (EM) and standard deviation error (ESD) for the probabilistic assessment results of eta under the local corrosion mode do not exceed 6.53% and 5.70%, respectively, while those of the uniform corrosion mode are as high as 30.20% and 19.41%, respectively. By integrating SMFL data, Bayesian-based probabilistic corrosion assessment, and Monte-Carlo simulation, a novel SMFL-based reliability assessment framework that accounts for the effect of stress was proposed. This framework enables accurate calculation of the failure probability pf and reliability index beta for RC girders with varying corrosion modes and MR. The stress-induced assessment error in pf reaches up to-203.5% for locally corroded girders and-2168.3% for uniformly corroded girders, highlighting the necessity of stress modification based on actual service conditions.

Cables are essential load-bearing components of cable-supported bridges, and cable tension serves as a critical parameter for assessing bridge safety. Vision-based cable tension measurement has gained popularity for its efficiency and cost-effectiveness. However, existing vision-based cable tension measurement methods remain constrained by substantial prior information requirements and the challenges associated with target-free full-field measurement of cables exhibiting low-amplitude vibrations. To overcome these limitations, this study proposes a cable tension estimation method using dual matching tracking and half-wavelength of mode shape. The effectiveness of the proposed method was verified through laboratory experiments and field tests, demonstrating a strong correlation with reference tension values. This method requires only one piece of prior information (mass density), allowing for full-field modal analysis and tension estimation without the need for markers or distinct textures, thereby facilitating the measurement of dynamic parameters under environmental excitation.

Strengthening existing reinforced concrete (RC) beams to enhance their mechanical performance is a current research focus in civil engineering. Fiber-reinforced polymer (FRP), renowned for its outstanding strength-todensity ratio, is frequently employed to upgrade the load-bearing performance of RC beams. Notwithstanding its benefits, a common drawback in these applications is premature failure triggered by abrupt debonding at the FRP-concrete interface-typically happening before the FRP can attain its full tensile capacity. Accurately identifying failure modes is thus critical for safe and efficient design. This study aims to develop an intelligent system capable of accurately distinguishing between the two primary debonding failure modes-intermediate crack-induced and plate-end debonding-in RC beams strengthened with FRP. A dataset comprising 229 experimentally tested specimens was curated from the literature and augmented using Generative Adversarial Networks (GANs). Several machine learning classifiers were trained using both original dataset and its augmented counterpart. To interpret model predictions, SHapley Additive exPlanations (SHAP) method was applied, offering quantitative measure of each input feature contribution to identified failure mode. Results show that GAN-based data augmentation improves model accuracy by at least 5%, and support vector machine outperforms other algorithms in classification performance. SHAP-based interpretability analysis highlights that the yield strength of the tensile reinforcement and the FRP stiffness are the dominant factors governing the failure mode. These insights offer practical guidance for optimizing FRP strengthening strategies in RC structures, helping engineers make more informed design decisions.

Metal Magnetic Memory (MMM) technology is an effective non-destructive method for detecting the early fatigue damage in steel structures. Since loading conditions strongly influence magnetic signal, this study investigates how three parameters (load amplitude, load ratio, and load peak value) affect magnetic field evolution during fatigue. Magnetic measurements are conducted on welded specimens subjected to three-point bending. The results show that magnetic intensity changes significantly as the load rises from low to high levels, while the spatial distribution remains nearly unchanged. In contrast, load reduction produces smaller variations in the magnetic response. A Bayesian hierarchical model was established to quantify the influence of each parameter. The results indicate that load peak and load amplitude exert more pronounced and consistent effects, whereas load ratio has a comparatively minor impact. A probability-based ranking analysis further confirms the following order of influence: load peak, load amplitude, and load ratio.

Accurate parameter identification and response reconstruction in nonlinear structural systems subjected to seismic excitations are critical for understanding and predicting their dynamic behavior. However, traditional approaches often face challenges in handling highly nonlinear systems, particularly those exhibiting stiffness degradation, strength degradation, and pinching effects. This paper proposes an improved Square Root- Spherical Simplex Radial Cubature Quadrature Kalman Filter (SR-SSRCQKF) for parameter identification and response reconstruction in nonlinear structural systems subjected to seismic excitations. The algorithm introduced QR decomposition for solving the square root of the error covariance matrix ensured numerical stability during the recursive estimation process. Numerical simulations and experimental validations demonstrate that the improved SR-SSRCQKF exhibits exceptional robustness under highly nonlinear and degraded dynamic conditions, effectively overcoming the challenges associated with stiffness degradation, strength deterioration, and pinching effects, outperforming other existing methods. Despite the challenges of system complexity and limited observables, the improved SR-SSRCQKF effectively reconstructs displacement time histories and hysteresis loops, showing robustness under highly nonlinear conditions. This study highlights the potential of the proposed method for real-time structural health monitoring and seismic response analysis.

The accurate detection of stress states in bridge steel welds is of critical significance for ensuring structural safety and durability. In this study, metal magnetic memory testing (MMMT) technology is innovatively introduced to detect and evaluate stress states in steel welds. By conducting magnetic detection experiments on two categories of welds (comprising eight specimens) under three-point bending, the effects of variables such as detection paths and specimen differences (structure type and initial magnetic state) on the distribution of the magnetic field were analyzed, and the evolutionary patterns of magnetic field under varying stress conditions were elucidated. Drawing upon the shared characteristics observed in the magnetic field variations of both weld categories, the relative entropy and the gradient Euclidean norm of magnetic memory signals were extracted as key magnetic characteristic parameters in the study. The results show that relative entropy changes gradually during the elastic stage but increases sharply upon transition into plastic deformation, with its rate of rise markedly exceeding that in the elastic phase. This finding indicates that relative entropy may serve as a reliable indicator for accurately identifying the stress states of the structure. Furthermore, there is a high linear correlation between the Euclidean norm of the magnetic signal gradient and elastic stress. By employing optimized feature parameters, structural stress can be effectively quantified, with most measurement errors maintained within 20

To address the technical challenges associated with complex connection configurations and excessively long lap zones in the industrialized and prefabricated construction of reinforcements, this study proposes a novel parallel-lap splice that incorporates a third overlapping reinforcement. This innovative design offers several advantages, including neat ends, ease of construction, and enhanced economic efficiency. An experimental investigation was conducted to evaluate the effects of this new splice on the flexural behavior of reinforced concrete (RC) beams, with lap length (ll) as the key variable (ll = 64d, 40d, and 25d). A total of nine simply supported RC beams (three groups of three specimens each), all incorporating parallel-lap splices, were tested under four-point bending. The key mechanical properties were analyzed, including the mechanical characteristics, failure modes, flexural capacity, bending stiffness, and maximum flexural crack width. The experimental and analytical results reveal that RC beams with the new parallel-lap splice exhibit a distinctive “one primary + two secondary” crack pattern, characterized by a dominant flexural crack at midspan and secondary cracks at the ends of the lap zone. At the ultimate limit state, specimens with ll = 64d experienced concrete crushing at the top surface of the midspan while those with ll = 40d and ll = 25d did not. Additionally, the ll = 64d and ll = 40d beams showed slight strength hardening, whereas the ll = 25d beams exhibited rapid strength degradation. In terms of load-bearing capacity, both the ll = 64d and ll = 40d beams met the requirements specified in current design codes, while the ll = 25d specimens showed a reduction in capacity exceeding 20%. Under serviceability limit states, midspan deflections and maximum crack widths for the ll = 64d, ll = 40d, and ll = 25d specimens were found to fully comply with, marginally satisfy, and fail to meet the requirements of the design code, respectively. Based on these findings, as well as regression analysis of the relationship between peak load and lap length, it is recommended that a reasonable lap length for the proposed parallel-lap splice be taken as 60d, with a lap length correction factor of 1.5.

Accurate localization of circumferential corrosion and quantitative evaluation of corrosion severity remain challenging for parallel steel wire bundles used in bridge stay cables. To address these issues, a self-magnetic flux leakage (SMFL)-based method for corrosion localization and quantitative assessment is proposed. First, a three-dimensional magnetic dipole model considering hyperboloid corrosion morphology is established based on magnetic dipole theory to investigate the relationship between corrosion parameters and magnetic signal characteristics, thereby revealing the influence of corrosion damage on the SMFL response. Subsequently, a circumferential corrosion localization method based on the centroid of a peak-to-trough magnetic signal radar chart is developed. By extracting the spatial distribution characteristics of peak-to-trough magnetic signals and calculating the centroid of the equivalent radar polygon, the circumferential corrosion center can be quantitatively determined. The proposed method is validated through numerical simulations and electrochemical corrosion experiments under various corrosion configurations. The results demonstrate that the method can effectively identify circumferential corrosion centers under different corrosion distributions, with a maximum localization error of 1.7° in the numerical simulations and an average localization error ranging from 1.1° to 4.3° in the experiments. Furthermore, the repeatability and robustness of the proposed method are evaluated using the mean absolute error (MAE) and root mean square error (RMSE), and the results confirm its satisfactory repeatability and localization stability. Finally, a quantitative relationship between corrosion mass and the magnetic characteristic parameter, ΔB, is established. Comparative fitting analyses using the Boltzmann nonlinear model and the linear model indicate that the Boltzmann model provides a superior fit and more accurately characterizes the nonlinear evolution of magnetic signals during the corrosion process. The proposed method provides an effective and reliable non-destructive approach for quantitative circumferential corrosion localization and corrosion severity assessment of parallel steel wire bundles in bridge stay cables.

This study systematically evaluates the mechanical performance of waterborne epoxy resin (WER) systems using a mortar-based testing method. The effects of WER content and mineral filler dosage on key volumetric parameters and mechanical strengths were investigated. Results show that increasing WER content reduces specimen density while significantly increasing void ratio and mineral aggregate (VMA). Conversely, higher mineral filler content enhances density, reduces void content and VMA, and achieves a peak compressive strength of 12.47 MPa at 12 % filler content. Optimal preparation parameters were determined as follows: mixing time of 2 x 30 s, compaction effort of 2 x 30 blows, and demolding at 1 d. The K12B2 mix (12 % mineral filler) was identified as the benchmark formulation for mechanical performance. Strength development analysis under the adopted 40 degrees C curing regime revealed that both flexural and compressive strengths stabilize by 14 d, reaching 99.3 % and 97.4 % of their 28 d values, respectively. It suggests that under this accelerated curing condition, the 14 d strength can serve as a representative index for performance evaluation and comparative screening of different WER systems. The procedure clearly differentiates between resin-hardener formulations and provides a practical basis for formulation selection and optimization. Given the measured strength and relatively fast curing, the developed mixtures are suitable for moderate-load applications that require rapid readiness, such as pavement surface repairs, bonding or surface treatment layers on concrete or asphalt, and non-structural repair or bonding tasks.

In practical engineering, such as bridge engineering, parallel steel wires are subjected to the action of axial tension and also to the action of transverse force. This situation can result in difficulties in accurately determining the effective tension of steel wires. This study examines the identification of effective tension in parallel steel wires subjected to transverse forces on the basis of the principle of the resonance-enhanced magnetoelastic effect. A theoretical relationship among transverse force, effective tension, and induced voltage is established through the principles of the magnetoelastic effect, electromagnetic induction, and structural mechanics. The identification of the effective tension of steel wires under transverse force is experimentally studied using a custom-made double-coil sensor. The frequency characteristic value of the voltage signal of the induction coil under varying axial tension and transverse force is analyzed in detail via fast Fourier transform. An effective method for identifying the tensile force of parallel steel wires, along with a correction approach that considers the influence of transverse forces, is proposed. Results indicate that the relative error of the effective tension recognition method is less than 13% after considering transverse force correction, representing a 22% reduction in relative error compared with the method before correction.

Virtual trial assembly (VTA) is emerging as a digital alternative to conventional physical trial assembly by leveraging 3D laser scanning and computer‐aided modeling to verify the fit‐up of large‐scale structures. This paper comprehensively reviews the VTA pipeline across three core stages: (1) multisource data acquisition combining terrestrial laser scanning, handheld scanning, and industrial photogrammetry for global coordinate system establishment; (2) point cloud processing involving preprocessing, deep learning–based semantic segmentation, and assembly‐critical feature extraction; and (3) virtual assembly including constrained registration and kinematic chain modeling for multisegment alignment, together with multiobjective optimization for rectification and physics‐informed compensation as an emerging direction not yet realized in the cases reviewed. We propose a novel four‐level capability maturity model charting VTA’s evolution from offline geometric inspection toward real‐time closed‐loop digital fabrication; the reviewed cases populate the lower levels, while the upper two levels are defined as forward‐looking targets not yet fully demonstrated in practice. The review further identifies technical bottlenecks, including precision‐utility trade‐offs, real‐time computational constraints, geometry‐mechanics decoupling, and the absence of standardized validation protocols. Future research directions are outlined to provide a clear roadmap for advancing intelligent and automated construction.

Although ultrahigh-performance concrete (UHPC) has shown exceptional promise for strengthening engineering structures, critical gaps remain in understanding the failure behavior of UHPC-strengthened flexural reinforced concrete (RC) beams. Existing approaches lack the precision required to reliably distinguish between ductile and brittle failure modes, which represents a key barrier to safe and efficient design. To address this gap, this study introduces a novel data-driven framework that integrates advanced machine learning with feature engineering to develop a highly accurate and interpretable model for identifying failure modes. First, the experimental data of 163 sets of strengthened beams with brittle and ductile failure modes were collected, followed by data augmentation using Tabular Generating Adversarial Networks (TGAN). Subsequently, classification models were constructed using six machine learning models. Finally, the Shapley Additive Explanation provides clear insights into the optimal machine learning model. Findings indicate that TGAN can improve the performance of machine learning models. Random forests exhibit the most balanced performance. The thickness (hu) and tensile strength (ft) of UHPC had the greatest influence on the failure modes, with an increase in hu leading to ductile failure and an increase in ft leading to brittle failure. The conclusions of this study can serve as a reference for the design of damaged RC beams strengthened by UHPC.

Bridge cracks, which are vital indicators of structural health, pose challenges in detection due to complex backgrounds and edge feature extraction difficulties. This study introduces CMC-Net, a dual-encoder network combining convolutional neural network (CNN) and Mamba for precise bridge crack segmentation. The CNN encoder, built with depthwise over-parameterized convolution, enhances local feature extraction, while the Mamba encoder captures global context information. A new feature fusion module (FFM) integrates these features, and the edge enhancement module (EEM) addresses edge feature extraction challenges. Experiments on CQBCD, DeepCrack, and CrackTree260 datasets show CMC-Net achieves MIoU of 89.47 https://github.com/DJ-hjk/CMCNet .

The SMFL-based non-destructive assessment of rebar corrosion characteristics provides a novel approach for assessing the load-carrying capacity and reliability of corroded RC structures. However, existing models for evaluating rebar corrosion characteristics based on self-magnetic flux leakage (SMFL) indices Is, NHx, and dH have not accounted for the adverse effects of corrosion products and the concrete cover. In this study, experimental tests revealed that the relative magnetic permeability of corrosion products can reach up to 2.218, significantly differing from that of air. In contrast, the relative magnetic permeability of concrete does not exceed 1.099, showing minimal difference from air. The analysis revealed that corrosion products and concrete cover induce a negligible error (<= 1.2 %) in the SMFL-based assessment of the rebar's cross-sectional corrosion degree, but a substantial error of up to 35.2 % in the assessment of corrosion unevenness. This study provides the crucial insight that existing SMFL models can be directly used for corrosion degree evaluation, yet must be corrected for unevenness assessment, thereby enabling more reliable capacity and reliability predictions for corroded RC structures.

The varying environmental conditions inevitably affect the effectiveness of damage detection methods applied to actual bridges. To achieve reliable damage detection results, it is vital to remove environmental influences from damage features affected by both environmental fluctuations and actual damage. In the earlier publications, the concept of Characteristics of the Narrow Dimension (CND) was introduced to describe certain damage features, along with tailored detection methods for features possessing CND, which demonstrated promising results in mitigating environmental effects. The present paper focuses on a further extension of the CND, specifically enhancing the CND of damage features through reconstruction methods to better accommodate the proposed damage detection method for CND. Firstly, the explicit mapping of a polynomial kernel within of kernel method is introduced to investigate the feasibility of reconstructing damage features. Then, the theoretical feasibility of improving the correlation coefficient of the reconstructed damage features is demonstrated through formula derivation. Moreover, the influence of the polynomial kernel on the mathematical relationship following the high-dimensional mapping of damage features is elucidated. Following this analysis, in conjunction with the calculation method for the CND evaluation index and the increase in correlation coefficients among the reconstructed damage features, a reconstruction method is proposed to enhance CND. Finally, numerical simulations and a real bridge case study are subsequently presented to verify the significance and effectiveness of the proposed reconstruction method.

By introducing the spectral decomposition-based explicit time-domain method, this paper presents an innovative computational framework for analyzing the non-stationary random vibration problem in a three-dimensional train-bridge system involving multi-variate random track irregularities. First, the time-dependent train-bridge model is formulated by coupling the train and bridge dynamics via a spatial wheel-rail interaction model. Next, the random track irregularities are decomposed through the spectral representation technique, enabling their time-domain discrete characterization in terms of three orthogonal random vectors. Then, an explicit mapping between the system responses and the orthogonal random vectors is constructed by integrating the precise integration method with a finite difference approach, leading to a recursive formulation that facilitates efficient computation of the response coefficient matrices. The obtained explicit response formulation allows straightforward computation of time-frequency response statistics of the train-bridge system, eliminating the need for repetitive time-domain simulations or extensive numerical integrations commonly associated with conventional non-stationary random vibration techniques. Lastly, the pseudo-excitation method and Monte Carlo simulation are adopted to verify the applicability of the proposed method, a comprehensive parametric investigation is also conducted to examine the individual contributions from different track irregularity components on the stochastic dynamic behavior of the train-bridge system.

Long-span arch bridges are widely adopted in mountainous areas of western China, yet the rapid construction requirements of main arch segments under complex working conditions are barely satisfied by conventional manual welding. To address this technical gap, the dedicated ring-welding robot for main arch segments and its collaborative obstacle-avoidance welding path planning method are proposed. Point cloud, kinematic and planning offset models of the robot and target arch segments are established, and multi-objective collaborative path planning under multi-obstacle scenarios is realized by integrating preset ring-welding paths and rib plate obstacle-avoidance paths. Numerical simulations and experimental validation results show that the proposed method satisfies all collision safety and welding posture requirements, and guarantees smooth motion in both joint and task spaces, with a 35 mm minimum safe clearance and torch inclination controlled between -7.156° and 28.075°. It also achieves the sub-millimeter Grade E end-effector positioning accuracy and uniform defect-free coating formation.

ABSTRACT Pavement roughness measurement constitutes a fundamental basis for evaluating road serviceability, ride comfort, and maintenance decision‐making. Existing vehicle‐mounted roughness measurement systems generally rely on dedicated survey vehicles, presenting limitations in mounting adaptability, installation complexity, and versatility, which hinder flexible deployment in grassroots maintenance scenarios, particularly for low‐volume and rural roads with growing maintenance demands. To address these issues, this study develops a lightweight pavement roughness measurement system suitable for ordinary vehicles, based on the combined sensing principle of laser displacement, acceleration, and rotary encoder. A portable mounting scheme is proposed, along with a data processing framework encompassing multi‐sensor synchronized acquisition, signal denoising, vibration displacement compensation, longitudinal profile reconstruction, resampling, and International Roughness Index (IRI) computation. The IRI is computed from the reconstructed longitudinal profile using the standardized quarter‐car simulation. Under the tested pavement conditions, experimental results indicate that both the 0.1 and 0.25 m sampling intervals adequately characterize the macroscopic longitudinal profile, while the finer 0.1 m interval is more favorable for capturing localized surface details; shorter computation intervals (10–20 m) exhibit higher sensitivity to localized anomalies than longer ones. Within the speed range of 30–50 km/h, the coefficient of variation of segmental IRI values remains below 5%, demonstrating satisfactory repeatability. Compared with a commercial comprehensive inspection vehicle, the CICS Road Condition Rapid Detection System (CICS), operated at the same speeds, the system yields generally consistent results, with optimal overall performance at 40 km/h, where the correlation coefficient reaches R = 0.989, and the root‐mean‐square error is 0.098 m/km.

Long-term performance of reinforced concrete (RC) beams poses significant challenges in civil engineering due to environmental degradation and material aging. Accurate assessment of the residual strength of corroded RC beams is crucial for ensuring structural safety and developing scientifically informed repair and rehabilitation strategies. In this study, an experimental database comprising 292 test specimens of corroded RC beams was established to evaluate the reliability of four existing empirical models. Machine learning algorithms were employed to develop a predictive model for the residual strength of corroded RC beams, and SHapley Additive exPlanations analysis was conducted to assess parameter importance and sensitivity. A graphical user interface was developed based on the optimal machine learning model, and the model’s reliability was validated against experimental data. The results indicate that while existing empirical models achieve moderate predictive accuracy, their stability is poor. The eXtreme Gradient Boosting demonstrates superior performance and is identified as the most suitable approach for predicting the residual strength of corroded RC beams. Parametric analysis reveals that beam width, tensile reinforcement corrosion ratio, and tensile reinforcement ratio are the most influential factors affecting residual capacity and should therefore be prioritized in structural assessment and modeling. Furthermore, the proposed model shows high predictive reliability when the reinforcement corrosion ratio is below 20%.