Highway slope disasters pose a significant threat to the safe operation of road traffic. Clarifying the fundamental causes of these disasters and developing lightweight monitoring and early warning systems have become urgent priorities to ensure the safety of road networks. Based on a comprehensive review of the characteristics and spatial distribution of highway slope disasters, this study systematically summarized five typical failure modes-collapse, landslide, debris flow, subgrade subsidence, and surface slumping-and analyzed the specific features and failure mechanisms of each type. An in-depth analysis was also conducted on the contributing factors and triggering conditions. Building on these insights, the study proposed a framework for lightweight monitoring and early warning of highway slope disasters. This framework consisted of five main components: on-site inspection, selection of monitoring indicators, optimization of monitoring locations, refinement of warning models, and implementation of preventive measures. The study also outlined the essential conditions required for implementing such lightweight monitoring and early warning. Based on this, the paper focused on highway embankment slopes characterized by “long corridors and strong constraints”. A core framework of “pre-screening before measurement, low-cost sensing, and lightweight modeling” was summarized, and a feasible implementation scheme was proposed. This work serves as a reference for the development and application of lightweight monitoring and early warning technologies for highway subgrade slopes and holds important implications for improving the resilience and safety of road infrastructure.

To address the challenges of complex spatiotemporal evolution, diverse triggering mechanisms, and delayed responses of conventional treatments for highway-associated landslides, this study proposed an integrated full-process “identification-early warning-control” approach for landslide hazard management. In the hazard identification stage, a rapid landslide detection and evolution-tracking method was developed by integrating multi-source data including remote sensing imagery, Interferometric Synthetic Aperture Radar (InSAR) interferometry, Unmanned Aerial Vehicle (UAV) photogrammetry, and Light Detection and Ranging (LiDAR), combined with geophysical prospecting and borehole inclinometers to accurately delineate the geometric characteristics of sliding surfaces. In the monitoring and early warning stage, a multi-parameter monitoring system was established by integrating Global Navigation Satellite System (GNSS), inclinometers, stress meters, microseismic sensors, and meteorological instruments, and a hierarchical early warning model was constructed based on deformation-mechanical control-environmental responses, enabling dynamic monitoring and accurate early warning under complex geological conditions. In the control strategy stage, a phased rapid emergency treating strategy was proposed, integrating structural optimization with components such as micropiles and intelligent anchoring systems, to achieve highly targeted and efficiently deployable countermeasures. The proposed methodology was validated through its application to the Naliang landslide treatment project along the Duba Expressway. The results demonstrate its strong engineering adaptability and high potential for broader implementation in integrated “identification-early warning-control” treatment of landslides along highway.

The dynamic evolution of earthquake-induced high-level rock mass collapses is highly complex, characterized by significant energy release, and poses severe threats to the safety of transportation lifeline projects. To elucidate the mechanism of dynamic instability and the associated hazard effects of high-level rock masses under strong seismic excitation, this study focused on the Lanhuazhai high-level dangerous rock zone, located merely 2.5 km from the epicenter in the Hailuogou Scenic Area during the 2022 Luding earthquake. A dynamic model incorporating real terrain features was established. Using the finite-discrete element method, the catastrophic evolution of the Lanhuazhai rock mass was numerically reproduced, and the failure modes of the highway subgrade under collapse-induced impact loading were identified. In addition, a quantitative assessment of the post-seismic hazard characteristics of the unstable rock mass was conducted, clarifying rockfall trajectories and their implications for highway safety. The results demonstrate that: under the “9·5” Luding earthquake, the Lanhuazhai high-level rock zone was governed by three dominant structural planes. Collapse instability generated a loose accumulation body of 90 000-160 000 m³ aligned along the NW-SE direction, constituting a critical obstacle for emergency highway clearance and post-disaster reconstruction; the catastrophic process of high-level rock collapse can be categorized into four dynamic stages: seismic cracking and sliding initiation, projectile flight, collision-induced fragmentation, and frictional accumulation; the hazard effects of earthquake-induced collapse on the highway subgrade were driven by a dual mechanism, in which seismic shaking induced micro-damage within the subgrade structure, while impact loading from collapsed rock masses triggered compressive-shear failure; two rockfall trajectories observed on January 18 and January 27, 2023, posed direct threats to the highway subgrade. For emergency restoration, a combined strategy of ‘stepped slope excavation and clearance of accumulation body + passive net protection of unstable rock’ is recommended, whereas a tunnel bypass scheme is advised for long-term reconstruction. These findings provide critical scientific insights and practical guidance for disaster prevention and mitigation of transportation infrastructure in mountainous regions affected by strong earthquakes.

Aiming at the problems that there is a lack of a comprehensive assessment framework for landslide risks of mountainous expressways and existing studies only focus on landslide susceptibility, which makes it difficult to fully support risk management and control, this study intends to construct a scientific landslide risk assessment system for mountainous expressways. Based on Geographic Information System (GIS) technology, this study proposes a three-dimensional assessment framework integrating “landslide susceptibility of slope units-vulnerability of expressways-exposure”. Specifically, firstly, combined with the landslide triggering mechanism and engineering geological survey data, 15 landslide influencing factors including elevation, slope gradient and slope position were selected, and a LightGBM landslide susceptibility evaluation model was constructed using the semi-supervised learning method. Secondly, by quantifying the traffic volume after road completion and the economic losses that may be caused by the damage of different structures, 5 vulnerability indicators were determined, including life loss, vehicle loss, direct loss of expressway structures, repair costs and indirect loss caused by traffic interruption. The entropy weight method was used to calculate the weight of each indicator to complete the vulnerability evaluation. Finally, the exposure of the expressway was characterized by the area of the slope unit where the road affected by landslides is located and the average distance from the slope unit to the road. By multiplying susceptibility, vulnerability and exposure, this study realized the quantitative comprehensive assessment of landslide risks of mountainous expressways. This study breaks through the limitation of traditional studies that only focus on landslide susceptibility, and constructs a “susceptibility-vulnerability-exposure” trinity comprehensive assessment framework for landslide risks of mountainous expressways. Moreover, the selection and quantification methods of indicators in each dimension are more in line with practical engineering scenarios. The research results can provide methodological support for the comprehensive assessment of landslide risks, and at the same time offer a scientific basis for the mitigation, control and risk management decision-making of landslide risks of mountainous expressways.

To address the problem of random noise in micro-electro-mechanical system (MEMS)-based landslide deep deformation monitoring signals, a noise reduction method that integrates an improved Honey Badger algorithm (IHBA), variational mode decomposition (VMD), and an improved wavelet threshold is proposed. First, the Tent chaotic mapping, the Whale Optimization algorithm (WOA) spiral search mechanism, and the Levy flight strategy were introduced into the traditional Honey Badger Algorithm (HBA). The IHBA was used to optimize the two key parameters of VMD decomposition, K (number of modes) and α (penalty factor), to obtain an optimal parameter combination. The noisy signal was then decomposed into multiple intrinsic mode functions (IMFs) via VMD. The variance contribution rate of each IMF was calculated to filter out components with low contributions. Subsequently, the remaining components were subdivided into effective, noisy, and noise components based on their correlation coefficients. An improved wavelet threshold function algorithm was designed by incorporating a high-precision fourth-power modulus processing method to overcome the limitation of the constant deviation in traditional threshold functions. The noisy components were denoised using this improved wavelet threshold algorithm. Finally, the denoised and effective components were reconstructed into a denoised signal. Simulation experiments demonstrated that the proposed method achieved the highest SNR (28.7642) and lowest RMSE (0.0101) compared with conventional denoising methods. Furthermore, a comparative analysis of real-world landslide deep deformation monitoring signal denoising revealed that the new method yielded the smallest RMSE and RVR and the highest NM and SNR, with all SERs exceeding 99%. It can effectively reduce noise while preserving the detailed features of the original signals, significantly improving the accuracy and reliability of MEMS-based landslide deep deformation monitoring data, thereby demonstrating promising application prospects.

To enhance the rationality of the seismic stability assessment of soil slopes, this study integrates the pseudo-dynamic method with Spencer's limit equilibrium method to derive a formula for calculating the factor of safety under pseudo-dynamic conditions. The Karhunen-Loève expansion method was employed to generate random fields of the soil parameters, and a Monte Carlo simulation was used to compute the failure probability. A risk assessment model for soil slopes under pseudo-dynamic conditions was established to comprehensively account for the combined effects of seismic dynamic loading and spatial variability of soil parameters on slope stability. By analyzing the peak failure probability under pseudo-dynamic conditions and average failure probability over the seismic period, the influence patterns of seismic motion parameters, slope geometric characteristics, and spatial variability of soil parameters (including autocorrelation distance, coefficient of variation, and correlation coefficient) on failure probability and slope safety factor were revealed. The results indicate that as the horizontal seismic acceleration coefficient increases, the pseudo-dynamic factor of slope safety is reduced nonlinearly, while significantly increasing the peak failure probability within the seismic period. With an increase in seismic wavelength, the potential slip surface was distributed more dispersedly and extended deeper, leading to a higher peak failure probability. The failure probability of the slope was positively correlated with the slope height and slope ratio, and an increase in the spatial variability parameters led to a higher failure probability. Therefore, the spatial variability of soil parameters should not be overlooked in seismic slope risk assessment.

Highway slopes are prone to soil erosion, which poses a significant threat to slope stability and ecological security along road corridors. Vegetation is widely recognized as an effective measure for controlling slope erosion. Most existing studies have focused on qualitative analysis of the influence of vegetation on runoff velocity and erosion processes, while quantitative investigations into flow velocity variations and critical vegetation coverage for soil erosion control on highway-vegetated slopes remain limited. To address this gap, laboratory runoff scouring experiments were conducted under varying vegetation coverage levels, flow discharges, and slope angles. The flow regime characteristics and mean flow velocities were systematically analyzed. A predictive model for the mean flow velocity on vegetated slopes was developed, and a method for calculating the critical vegetation coverage was derived from the model. The results indicate that the surface runoff predominantly exhibits laminar and transitional flow regimes, with all flows operating in the supercritical regime. With increasing vegetation coverage, the Reynolds number increases, whereas the Froude number decreases. By contrast, higher flow discharges and steeper slope angles significantly increase both dimensionless numbers. The mean flow velocity decreases with increasing vegetation coverage, exhibiting a saturation effect at higher coverage levels. Both the flow discharge and slope angle have significant accelerating effects on flow velocity, following exponential growth patterns. Based on the principles of energy conservation and hydraulic theory, a predictive model for the mean flow velocity was established by incorporating modified Manning's roughness and local resistance coefficients under varying environmental conditions. The model was validated against 294 experimental datasets under diverse conditions, achieving high accuracy (R²>0.900) and demonstrating its robust applicability and stability. Based on the experimental results, the critical incipient velocity equation for soil particles and the Soil Conservation Service Curve Number model were integrated to establish a method for determining critical vegetation coverage on highway slopes. The effectiveness and reliability of this method in erosion control were evaluated using the cumulative sediment yield and particle size distribution data from runoff scouring experiments.

To investigate the long-term structural damage mechanism of asphalt mixtures under complex humidity conditions, this study focused on asphalt mixtures with 100% replacement of natural aggregates by steel slag. Constant humidity curing environments (60%RH, 80%RH, and 95% RH) were established. A multiscale coupling analysis was conducted by combining X-ray CT image analysis and long-term dynamic modulus testing. The results showed that the pore structure of steel slag asphalt mixtures under high humidity followed a three-stage evolution: micropore formation, development of small and medium pores, and coalescence into large pores. Valid pores dominated the volumetric expansion and performance degradation process. Significant differences were observed between mixtures with different gradations in pore evolution patterns and damage responses. In the SMA-13 gradation, strong pore coalescence formed large connected networks (average volume reached 47.12 mm³). In contrast, the AC-13 gradation showed micropore development (pore count increased by 91.4%), resulting in better structural stability. Dynamic modulus testing revealed that increased humidity and extended curing time significantly reduced the mixture stiffness. The hydration reactions of free calcium oxide (f-CaO) and free magnesium oxide (f-MgO) were the primary damage-inducing factors. In the micro-macro correlation analysis, the Mantel test was introduced to quantify the relationship between the pore structure parameter matrix and the dynamic modulus response matrix. The results confirmed that porosity and average coordination number were significantly negatively correlated with the dynamic modulus. The coupling relationship between microstructural parameters and macro performance varied with gradation. This study provides a theoretical basis and data support for optimizing the performance and promoting the efficient application of steel slag asphalt mixtures in road engineering.

The volume expansion characteristics and water-damage risks of steel slag restrict its engineering applications as a potential substitute aggregate for asphalt pavements. To address the challenge of predicting the volume expansion and water stability of steel slag asphalt mixtures, this study established a machine-learning prediction model that incorporated multiple factors. Based on immersion expansion tests and 300 water stability tests covering variables-such as asphalt type, steel slag content, f-CaO content, gradation, and environmental conditions, a backpropagation neural network model was developed based on water-induced volume expansion and a CatBoost prediction model was optimized using Bayesian optimization and cross-validation. SHapley Additive exPlanations (SHAP) theory was employed to analyze the feature importance and parameter sensitivity that affect water stability. The results indicate that the volume expansion of the steel slag asphalt mixtures was significantly correlated with the gradation composition, f-CaO content, and immersion time. The CatBoost model achieved the highest prediction accuracy for the residual stability and tensile strength ratio (TSR) and effectively reflected the prediction error, with R² >0.997 and MSE<0.344 5. Among the material factors influencing water stability, the f-CaO content of the steel slag coarse aggregate (mean SHAP values: 2.05, 1.21, 1.17, and 4.62, 1.44, and 0.77, respectively) was the most crucial, followed by the asphalt type (0.84 and 0.82), steel slag content (0.36 and 0.32), and asphalt content (0.12 and 0.38). There was an interactive effect between the feature combinations of steel slag f-CaO content-asphalt content and f-CaO content-steel slag content on water stability. To satisfy water stability requirements, the steel slag content in the surface layer of the asphalt pavement should not exceed 75%. Additionally, the f-CaO content thresholds for steel slag with particle sizes of 2.36, 4.75, and 9.5 mm should be controlled within 2.0%, 2.25%, and 2.0%, respectively. This study provides theoretical support for controlling steel slag expansion and predicting water stability, thereby promoting the resourceful utilization of steel slag in asphalt pavements.

The feet-lock pipes (bolts, cables) have the functions of stabilizing the feet of initial support of the tunnels, preventing arch falling, and suppressing the discrete deformation of the surrounding rock. They are widely emphasized in the design and construction of tunnel supports in weak rocks. The support mechanism, mechanical characteristics, design parameters, and construction techniques of the feet-lock pipes (bolts, cables) have always been the focus of research and attention. Based on relevant research and engineering practices regarding feet-lock pipes (bolts, cables) support in weak rock tunnels, this paper systematically reviewed and summarized the six aspects: development history of support, construction processes, support function principles, mechanical testing methods, stress characteristics and design methods, while analyzing achieved research progress. Development history of support: the development has progressed through stages of “early arch foot bolt → feet-lock bolt → small-diameter feet-lock pipe → large-diameter feet-lock pipe → small-diameter prestressed feet-lock cable → constant-resistance feet-lock cable”. Construction Processes: the support parameters, drilling, anchoring and connection processes for small-diameter feet-lock pipes (cables), large-diameter feet-lock pipes, and small-diameter prestressed feet-lock cables have been determined. Support function principles: it is revealed that feet-lock pipes (bolts) primarily function through an inclined pile to control settlement of the arch foot, while feet-lock cables anchor deeply into the surrounding rock and can apply high pre-tensioning forces, providing suspension and active restraint effects on the arch foot of initial support. Mechanical testing methods: A simulated loading test method for feet-lock pipes (bolts) and feet-lock Pipes for force measurement using Fiber Bragg Grating (FBG) suitable for on-site testing have been developed. Stress characteristics: the stress characteristics of the feet-lock pipes (bolts) and its sharing effect on the foot load were explored, and the distribution law of strain on the feet-lock pipes was revealed. Design methods: it was analyzed that the feet-lock pipes (bolts, cables) increase the constraint (or support) strength and stiffness of the feet of initial supports, and the support design methods for feet-lock pipes (bolts, cables) and combination structure of steel rib, shotcrete, steel mesh and feet-lock pipes (bolts, cables) were established. At the same time, the research development trends, design and construction technical specifications, as well as the promotion and application of feet-lock pipes (bolts, cables) are prospected.

During tunnel construction, complex geology, variable processes, and high external uncertainty make collapse risk a major threat to project safety, progress, and human life/property. Current assessments rely on subjective expert judgment, which is slow and inefficient, hindering timely emergency response. To address this, we propose a tunnel collapse risk decision-making intelligent framework based on large language model (LLM). The framework leverages the iS3 Tunnel Intelligent Construction Platform to integrate geological, construction, and deformation monitoring data into a comprehensive risk database for subsequent analysis. Using prompt engineering, the LLM automatically quantifies collapse likelihood and accident severity, achieving intelligent fusion and analysis of multidimensional data. An improved cloud model and fuzzy risk matrix then accurately characterize evaluation uncertainty and classify risk levels, providing scientific safety recommendations for construction sites. Validation on four typical construction sites in the Yanjiazhai Tunnel shows that the framework accurately identifies and quantifies potential collapse risk, delivers real-time risk feedback, and proposes targeted countermeasures, thereby effectively enhancing tunnel risk management. Overall, this intelligent framework overcomes the limitations of subjective expert judgment, offers an efficient, automated approach for tunnel safety management, and supports broader application of LLM-based risk assessment and decision-making technologies.

Field measurements of bridge vibrations provide crucial information for structural modal parameter identification, behavioral analysis, and load-bearing capacity evaluation. Long-span bridges often feature extensive spans and wide water areas at the bridge sites, making close-range monitoring difficult using traditional machine-vision measurement methods. Therefore, a bridge vibration displacement measurement method is proposed using integrated UAV close-range photography and adaptive Digital Image Correlation (DIC). This method first designs a novel remote laser projection device that allows precise projection of the laser spot onto the reserved area of the monitoring target by adjusting the screw of the instrument, compensating for and correcting changes in the UAV's hovering posture. Subsequently, breaking away from the traditional stationary setup of machine vision, a multirotor UAV was controlled to hover directly in front of the bridge monitoring target, capturing high-resolution, high-quality target images at a close range. Next, a UAV-based image processing method using adaptive DIC was developed by utilizing local feature matching to adjust and optimize the subset area of the measurement point frame-by-frame. Subpixel displacement was identified using the Fourier transform cross-correlation algorithm and the inverse compositional Gauss-Newton algorithm, thereby extracting the UAV measurement displacement sequence and laser correction value sequence. Finally, the UAV displacement measurement sequence is corrected using laser correction values and dynamic scaling factors to obtain high-precision bridge vibration displacements and frequencies. The verification experiments showed that the UAV method could accurately measure the vibration displacement and frequency of each model layer. Compared with the measurements by fixed cameras and laser displacement meters, the Normalized Root Mean Square Error (NRMSE) of the vibration displacement was not greater than 3.195%. The relative error of the vibration frequency was not greater than 0.96%. When applied to the vibration displacement measurement of the Hongshan Bridge in Changsha, the method successfully identified the vibration displacements and frequencies at the midpoint, 1/4 point, and 3/4 point of the main span, with a relative error in the vibration frequency not exceeding 0.77%, matching the results identified by the fixed camera and accelerometer. In conclusion, the UAV-based high-precision, high-efficiency, noncontact displacement measurement method proposed in this study provides a new approach for measuring the vibrations of long-span bridges and has significant scientific research and engineering application value.

To investigate the interface bond behavior between 600 MPa grade stainless-steel reinforcement and concrete, pull-out tests were conducted on 54 specimens (18 groups), with the effects of concrete strength, rebar diameter, cover thickness, relative anchorage length, split surface stirrup ratio, and reinforcement type on bond behavior were analyzed. The test results indicate that the failure patterns of the specimens include concrete splitting failure, rebar pull-out failure, and splitting-pull-out failure. For specimens with a relative anchorage length of 5, those specimens with a cover thickness of 3 times the rebar diameter exhibit a transition from concrete splitting failure to rebar pull-out failure. After configuring stirrups, specimens with a cover thickness of 2 times the rebar diameter shift from concrete splitting failure to splitting-pull-out failure. The interface ultimate interface bond strengths between 600 MPa grade stainless-steel reinforcement and concrete increase with the increase of concrete strength, initially rise and then stabilize as the cover thickness and split surface stirrup ratio increase, and show a phenomenon of first increasing and then decreasing as the increase of relative anchorage length, but is not significantly affected by the rebar diameter. The critical transition values are identified as 4.5 times rebar diameter for cover thickness, 2.0% for split surface stirrup ratio, and 5 for relative anchorage length. The bond behaviors of 600 MPa grade stainless-steel bars and ordinary steel bars before yielding are essentially equivalent in concrete with different strength grades, and the influence of elastic modulus on the bond behavior of steel bars can be completely ignored. After yielding, the bond stress-slip curves of stainless-steel reinforcement exhibit a relatively smooth transition, and the ultimate bond strength is slightly lower than that of ordinary reinforcement. Based on the test results of the bond behavior between 600 MPa grade stainless-steel reinforcement and concrete, a calculation formula for the ultimate bond strength with a certain guarantee rate and a bond stress-slip constitutive model are presented. Additionally, a formula for calculating the critical anchorage length of the steel bars expressing in terms of the design values of material strengths is proposed. It can serve as a reference for the bond anchorage design of 600 MPa grade stainless-steel reinforcement in concrete.

The entrances and exits of expressway tunnels often cause drastic changes in visual space and illuminance, which easily trigger “black-hole” and “white-hole” effects. The interior sections of tunnels are characterized by monotonous visual environments and a lack of reference objects, leading to “spatiotemporal tunnel” and “sidewall” effects. Tunnel groups and spiral tunnels are prone to inducing the “whipping effect” and “psychological rotation effect,” respectively. These negative psychological effects significantly interfere with drivers' perceptions and judgments, provoke inappropriate driving behaviors, and reduce traffic safety. Based on the logical chain of “tunnel visual environment-negative psychological effects-inappropriate driving behaviors-reconstruction of visual reference system-optimization strategies,” a framework for analyzing and regulating drivers' negative psychological effects was constructed. The relationship between the visual environment and negative psychological effects was explored, regulatory strategies were proposed, and methods for optimizing and evaluating visual reference systems were summarized. The results indicate that drivers' negative psychological effects mainly originate from drastic changes in the visual reference system at tunnel entrances and exits and from the lack of variation in the weak visual reference system within the interior section. Ordinary tunnels should upgrade their basic visual reference system to a safe or comfortable type; extra-long tunnels and connection sections of tunnel groups should adopt a rhythmic visual reference system; and spiral tunnels (and tunnel groups) require a constant, stable, continuous, and redundant visual reference system. Regulation strategies include clarifying the spatial right-of-way, aligning with drivers' perceptual needs, decomposing driving tasks, enhancing comfort and rhythm, and introducing linear visual guidance and curve-constant delineation systems. Shading facilities, lighting installations, visual guidance devices, landscaping, interior decorations, and pavement treatments can effectively optimize the visual environment. However, their design parameters require further systematic investigation. Currently, the evaluation of visual reference systems lacks a unified index system and methodology, and should integrate optical parameters with human factor indicators, including visual perception, visual characteristics, physiological responses, and driving behaviors. In the future, evaluation systems and optimization strategies should be improved through engineering case studies to support the development of safe, energy-efficient, comfortable, and aesthetically pleasing tunnel environments.

To ensure driving safety on highway segments with complex alignment combinations, this study proposed a data-driven prediction model for passenger car speed distribution. The model addressed the issues of insufficient quantitative representation of highway alignment information and the limited consideration of the coupling effect between highway geometric alignment and drivers' visual perception on running speed. The study conducted experiments using an eight-degree-of-freedom driving simulator and a high-precision eye-tracking device, collecting driving speed and visual perception data from 38 drivers across 106 typical mountainous highway segments. From the dual perspective of highway segments and drivers' visual field, highway alignment features were extracted, and driver visual alignment was reconstructed using a variational autoencoder model. The study introduced a “highway segment-visual field” graph to fuse and quantify alignment information, and a passenger car speed distribution prediction model was constructed based on spatio-temporal graph attention neural networks. The results indicate that the proposed model exhibits superior performance on the testing dataset, with a prediction accuracy of 96.3% and 98.1% for the mean and standard deviation of running speed, respectively. The root mean square errors are 4.6， 1.2 km·h^-1, while the mean absolute errors are 3.8， 0.8 km·h^-1. Additionally, ablation and comparative experiments further validate the model's effectiveness and applicability across different typical highway segments. The self-attention mechanism reveals that highway geometric alignment provides a fundamental temporal guide for passenger car speed, while drivers' visual alignment offers dynamic spatial adjustment and feedback. The findings could establish a theoretical foundation for intelligent highway safety evaluation and geometric design methods, while also contributing to the advancement of refined speed management in intelligent transportation systems.

Following the rapid development of vehicle-to-vehicle and vehicle-to-infrastructure technologies，Connected autonomous vehicles (CAVs ) have become a research hotspot. However，the widespread adoption of CAVs is still a long-term goal，and the coexistence of CAVs and human-driven vehicles (HDVs)will remain in a transitional state in the foreseeable future.When a CAV follows an HDV，its performance may degrade owing to the lack of intervehicle communication and coordination.To mitigate the adverse effects of performance degradation on mixed traffic，CAVs can use collaborative signals，such as honking，light signals，and heads-up displays to enhance HDV drivers'awareness of their acceleration and deceleration behaviors，thereby improving cooperation efficiency.This study investigates the effects of CAV performance degradation and collaborative signaling behaviors on the stability and road capacity of mixed traffic flow. By modifying various car-following models，this study quantifies the microscopic car-following behaviors of HDVs considering collaborative signaling，as well as those of well-performing CAVs and degraded-performance CAVs.A stability analysis method for mixed traffic flow is then developed，deriving the mixed traffic flow stability conditions related to CAV penetration rates.Additionally，this study develops a fundamental diagram model related to CAV penetration rates for mixed traffic，and analyzes the key factors affecting the fundamental diagram. Numerical simulations validate the reliability of the stability analysis and the constructed fundamental diagram.The results indicate that when CAV penetration is below 40%，the CAV advantages are almost negligible；only when the penetration rate reaches 60% or higher，the benefits of CAVs become evident.Moreover，collaborative signaling behaviors of CAVs enhance their cooperation with HDVs，mitigating the considerable performance decline caused by the lack of intervehicle communication，and improving the stability and road capacity regarding mixed-traffic flows.Therefore，while improving CAV technologies to enhance road capacity，it is equally important to design effective signaling mechanisms to promote cooperation between CAVs and HDVs，thereby improving CAV performance at lower penetration rates.

Intelligent connected technology is the future trend for intelligent transportationsystems.However，mixed traffic flows consisting of Connected and Automated Vehicles (CAVs) and Human-Driven Vehicles (HDVs)will persist for a long time.One of the key challenges is leveraging the connectivity and controllability of CAVs to optimize traffic operations and enhance the efficiency of road resource utilization.Currently，research on lane management in mixed traffic flows primarily focuses on the allocation of road rights under different traffic demands and CAV penetration rates，without fully considering the active adaptation capabilities of CAVs and their bidirectional interaction with traffic controlsystems.Therefore，the concept of Dedicated Lanes for Granule/Flow Cooperation (DEL4GFC) was proposed.The control method for DEL4GFC encompasses three components：the management zone，lane configuration，and granule/flow cooperativestrategies.Granule/flow cooperativestrategies are aimed at optimizing road management through centralized and distributed CAV control.For regular highway trafficscenarios，asingle management zone and DEL4GFC were established.Through threesets of distributed CAV granule controlstrategies，varying degrees of control were achieved in four aspects：CAV noncooperative lane changing，cooperative lane changing，dedicated lanespeed adjustment，and all-lanespeed adjustment，which enhanced the aggregation of vehicles on roads. The experimental resultsshow that the maximum improvement in traffic capacity can reach 17.0%.For highway accidentscenarios，three management zones-adjustment，lane changing，and recovery-were established，and the corresponding road rights werespecified. The management zones implemented cooperativestrategies for temporary lane closures through distributed CAV flow headway adjustments，centralized CAV granule traffic balancing，and distributed CAV granule lane recoverystrategies.The experimental resultsshow that the DEL4GFC control method can improve traffic capacity by up to 18.1% compared to the baseline，and the maximum reduction in the average vehicle delay time is 336s.Insummary，by enhancing the degree of physical coupling between CAV flows and mixed traffic flows and leveraging cooperativestrategies for targeted trafficscenarios，DEL4GFC cansignificantly enhance road capacity and effectively optimize traffic operationsimultaneously.

To improve the traffic efficiency and safety near freeway bottlenecks，this study proposes a two-stage dynamic speed limit control method based on CAV (Connected and Automated Vehicle)moving barrier for the Car-Truck and CAV-HDV (Human-driven Vehicle) mixed traffic flow.Firstly，an improved two-stage dynamic speed limit control framework including CAV lane change control and CAV dynamic speed control was developed，which is suitable for multi-lane，and a moving barrier that regulates the overall traffic flow is formed by optimizing the distribution and speed of CAV-Cs and CAV-Ts (CAV Cars and Trucks ). Secondly，considering the heterogeneous characteristics of cars and trucks，the dynamic speed control model was formulated for CAV-Cs and CAV-Ts to accurately determine the control speed.Meanwhile，the key evaluation index in the CAV lane-changing strategy-uniform distribution coefficientP，was also improved based on the difference between CAV-Cs and CAVTs.Under the condition of whether CAV-T can change to the fast lane，with the optimization objective beingP，four algorithms-enumeration，simulated annealing，genetic algorithm and reinforcement learning-were used to solve the optimal lane-changing plan for CAV-Cs and CAVTs.Finally，the control effect of the proposed method is discussed，and its effectiveness is verified by the SUMO simulation of the one-way two-lane and the one-way four-lane highways. The results indicate that the proposed CAV lane changing control strategy can significantly enhance the distribution uniformity of CAV-Cs and CAV-Ts.The uniform distribution effect is significantly influenced by the lane-changing constraints of trucks.When the proportion of CAV-Ts to CAVs is higher than 50%，the control strategy without vehicle lane-changing restrictions outperforms the strategy with restrictions. By combining with the lane-changing control strategy，the two-stage dynamic speed limit control strategy can significantly reduce driving risk and improve traffic flow stability.As the proportion of CAV-Ts increases，the control effect can generally be further improved，even though the truck's characteristics will disrupt the traffic flow.These research results provide effective theoretical support for improving the efficacy of active traffic control near freeway bottlenecks under intelligent and connected transportation scenarios.

Intelligent connected vehicles (ICVs)and human-driven vehicles (HDVs)coexist in mixed traffic flows，significantly affecting the merging behavior at highway on-ramps.This coexistence poses challenges，such as low throughput efficiency and high energy consumption，owing to the inadequate integration of merging control sequences and vehicle trajectory interactions in dense traffic scenarios. To address these issues，we introduced a dynamic cooperative merging control strategy tailored for highway vehicles. This strategy initially involved identifying potential gaps for ramp vehicles based on the real-time status within the control area.It then proceeded to collaboratively optimize the merging sequences and vehicle trajectories with the aim of minimizing energy consumption.This included proactive lane changes and speed adjustments for both ramp vehicles and those in anticipated merging gaps. Furthermore，considering the dynamics of HDVs，the strategy incorporated a dynamic control mechanism to regulate vehicles in the merging area，ensuring efficient integration under varying traffic densities.The effectiveness of the proposed strategy was validated using simulations in SUMO and Python，comparing it against scenarios with no control and a PID-based control strategy.The results demonstrated substantial improvements：① With high ramp traffic volumes，the strategy reduced the overall collision metrics by 38.84% to 94.25%，reduced total delays by 13.45%-80.82%，and increased average speeds by 36.01%-52.87%.② Compared with PID-based control，it further reduced collision rates by up to 81.38%，delays by up to 66.95%，and energy consumption by up to 8.17%，while ensuring smoother vehicle trajectories and more evenly distributed vehicle speeds.③ Although the preemptive lane-changing strategy results in higher energy consumption，it substantially enhances safety and throughput efficiency，underscoring the benefits of the proposed dynamic cooperative merging control strategy in managing mixed traffic on highways.

Mountainous cities face challenges owing to their complex terrain，leading to significant difficulties in multi-lane merging areas on arterial roads.These areas have complex geometric structures，varying traffic flow directions，and frequent conflict interactions.Under high traffic flow volumes，the merging of multiple traffic streams results in frequent vehicle behavior conflicts.This study proposes a cooperative control approach for multi-lane merging areas based on connected and automated vehicles (CAVs)to address the limitations of current control strategies such as isolated measures and single-objective designs.First，a cooperative control framework for multi-lane merging in an intelligently connected environment was developed，incorporating signal control and variable speed limit control agents.Second，the state space，action space，and reward mechanism for the agents were designed considering the specific traffic flow characteristics of mountainous cities. Finally，a multi-agent deep deterministic policy gradient algorithm was applied to achieve cooperative control among agents.Using a specific interchange in Chongqing as a case study，ten simulation experiments were conducted under high and medium traffic volume scenarios to evaluate the effectiveness of different control strategies. The results show that under conditions of low CAV penetration and high traffic volume，the proposed method reduces the average travel time on the mainline by 37.02% and the average vehicle delay by 69.57% compared with the fixed-timing control strategy.Compared with traditional feedback-based cooperative control methods，the proposed approach reduces the upstream queue length by 88.06%，increases the average speed in the bottleneck area by 8.77%，and improves the downstream discharge flow by 3.47%.These findings indicate that the proposed method can effectively mitigate traffic conflicts and congestion in multi-lane merging areas of mountainous cities，providing valuable theoretical support for traffic congestion management in these complex urban environments.

Analyzing the characteristics and laws of traffic phase change is the key to traffic flow state estimation and speed prediction.Due to the complex time-varying characteristics of traffic flow with both periodicity and chance，mathematical statistics and machine learning methods lack the analysis of the evolution mechanism of traffic flow state，and there is a shortage that the interpretability of the model decreases with the change of the scene.In order to reduce the negative impact of scene variation on the prediction effect，and to solve the problem of limited application environment，this paper analyzed the mechanism of phase transition of traffic flow，and proposed a traffic state estimation and speed prediction method considering the lag effect of phase transition.Firstly，by analyzing the traffic flow state change process of expressway exit ramps，it is found that the spatio-temporal transfer of motion differences between microscopic vehicles is the cause of macroscopic traffic phase change.This phenomenon was defined as traffic disorder，which was quantitatively expressed analogously to the Ising model.Subsequently，the spatio-temporal distribution of disorder before and after the traffic phase change was calculated，which revealed the law of traffic phase change，i.e.，the state of traffic flow has a time lag relative to the change of disorder.On this basis，an autoregressive distributed lag model of traffic flow speed with respect to the disorder was established，and the real-time speed and displacement of vehicles collected by the on-board and roadside connected detection equipment were used as inputs to obtain the time series of predicted values of traffic flow speed.The parameter analysis results of controlled experiments with different models show that ① the proposed lag model has higher prediction accuracy compared with Radial Basis Function Neural Network and Long Short-Term Memory.② In mixed traffic flow，the higher the Intelligent Connected Vehicle penetration rate，the higher the model prediction accuracy.In addition，the proposed method is applicable to traffic scenarios with varying levels of vehicle-road cooperative networking，which is conducive to timely traffic control measures，improving both the operational efficiency and safety of the transportation system from a prospective perspective.

Cut-in scenarios arehigh-risk situations for automatedvehicles（AVs），primarily because of unclear risk mechanisms and a lack of precise assessment models.In this study，30000 cut-invehicle trajectories and video segments were extracted from naturalisticdrivingdata collected under real road conditions.From these，489 cut-in scenarios with varying risk levels were selected using objective risk indicators such as the time to collision.A subjective riskassessment questionnaire tailored to the characteristics of AVs wasdeveloped，and engineers involved in thedevelopment and testing of AVs were recruited to conduct subjective risk evaluations and in-depth interviews.Subjective risk levels and contributing factors were identified from an AV perspective，and a comprehensive cut-in scenario riskdataset covering 24dynamic and static factors was established.Finally，a quantitative risk evaluation model for cut-in scenarios was constructed based on a random parameter-ordered logit（RPOL）framework.The results show the following：① road factors such as intersections and lane marking change indicator increase the probability of a cut-in scenario beingdangerous by 0.3% and 17.9%，respectively，② weather conditions like nighttime and rainy weather raise thedanger probability by 19.4% and 23.3%，respectively，③ in terms of traffic participant factors，the probability ofdanger increases by 10.6% for irregular cut-invehicles；the presence of non-motorizedvehicles or pedestriansduring cut-in raises the risk by 17.9% and 32.8%，respectively，while each additional nearbyvehicle increases the risk by 0.6%，and ④ amongvehicle kinematic features，each unitdecrease in initial longitudinaldistance and relative speed increases thedanger probability by 0.3% and 1.04%，respectively，while each unit increase in initial lateraldistance anddeceleration of the cut-invehicle raises the risk by 0.5% and 0.6%，respectively.These results indicate that vulnerable road users，irregular cut-invehicles，low visibility，lane marking changes，and lane obstructions arehigh risk factors for AVs in cut-in scenarios.Priority should begiven to optimizing the relevant perceptions anddecision-making algorithms.

Artificial intelligence (AI) technology has become a core component of national strategic science and technology. Its integration with bridge seismic engineering is emerging as a critical approach to enhancing the seismic resilience of infrastructure. Bridge seismic analysis has long faced challenges such as complex physical models and the difficulty of balancing efficiency with accuracy. Addressing these challenges, this study systematically reviews the application and innovation of traditional machine learning models, deep learning models, and next-generation AI fusion technologies in bridge seismic analysis, including: ① Intelligent synthesis and input of complex ground motions; ② Seismic capacity analysis and demand prediction; ③ Damage assessment and fragility analysis; ④ Resilience evaluation and recovery strategy optimization; ⑤ Seismic analysis of large-scale bridge networks. AI has significantly improved the efficiency and accuracy of bridge seismic analysis, opening new avenues for exploring problems involving multiple parameters and strong nonlinearity. However, existing AI models still face persistent challenges, including insufficient foundation in physical laws, weak model generalization capabilities, and difficulties in effectively integrating heterogeneous data sources. Looking ahead, AI technology will further advance the field of bridge seismic engineering through enhancing the interpretability of physical laws, developing multi-modal sensing technology, building high-fidelity databases, strengthening model generalization capabilities, and developing novel intelligent algorithms. This research facilitates a paradigm shift in bridge seismic studies, moving from reliance on manual expertise towards a deep integration of physical laws with artificial intelligence.

Rapid post-disaster assessment of bridges in near-fault regions plays an important role in saving time for earthquake relief and advancing post-disaster reconstruction. To predict the seismic response and structural damage fragility of near-fault bridges quickly and accurately, a deep learning-based rapid prediction method for the seismic response and fragility of near-fault bridges is proposed to rapidly predict the nonlinear seismic response and fragility curves of bridges under near-fault impulsive seismic effects. The method is based on a unidirectional multilayer stacked Long Short-Term Memory (LSTM) network with two parts of the seismic response time history and peak response as the output of the model. Furthermore, the techniques of sliding time window, residual connection, and constrained cyclic kernel, which effectively capture the pulse-like ground motion inputs and the bridge responses (e.g., pier bottom bending moments, pier bottom curvature, and top of pier displacements), are employed. The nonlinear mapping between the pulse-like ground motion inputs and bridge responses (e.g., pier bottom bending moment, pier bottom curvature, and pier top displacement) is accurately predicted based on the probabilistic seismic demand model. Considering the Miaoziping Bridge damaged by the Wenchuan earthquake as an actual engineering research object, based on the OpenSees numerical model, a database of 619 near-fault ground motions, corresponding nonlinear seismic responses, and bridge susceptibility is established to verify the accuracy and rapidity of the proposed method. The results show that the LSTM model can predict the response over a long period of time with multiple outputs and can accurately capture the seismic response demand of bridge structures under impulsive seismic effects. In the peak response prediction, the bending moment index has the best prediction effect, and the mean and standard deviation of the ratio of the predicted value to the actual value are less than 1.03 and 0.12, respectively, followed by the displacement index and curvature index. The results of the predicted fragility curve and actual results are very close to each other, with a coefficient of determination of 0.97 and a maximum difference of 1.92% in the probability of failure. The time required for this method and traditional method is approximately 1 s and 66 h, respectively. The fast fragility prediction method proposed in this study has high accuracy and rapidity, and it can provide strong theoretical support for post-disaster bridge assessment.

The seismic mitigation design of suspension bridges has many problems such as complex structure and multiple control optimization objectives, which is especially prominent under the ground motions in fault rupture zone with high seismic intensity. In order to realize the efficient multi-objective optimization of the seismic mitigation system of suspension bridge, adopting the safety level of multiple key components and damper costs as control objectives, the response surface theory, competitive multi-objective particle swarm optimization algorithm, and various solution select strategies were combined to construct a multi-strategy multi-objective intelligent damping optimizing method for suspension bridge. Taking a suspension bridge perpendicularly crossing strike-slip fault as an example, the proposed method was used to intelligently optimize the bidirectional viscous damper coefficient distribution and the rational damper distribution law for this bridge type was clarified. Finally, its rationality was verified. The results show that ① the proposed method can better realize the balance between multiple control objectives while it has high optimization accuracy, efficiency, and flexibility; ② the rational ratio of the total transverse damper coefficient at abutments to that at pylons locates at 1.5-2.0; ③ compared with the initial damping scheme, the damping scheme optimized by the proposed method increases the safety factors of the main truss, the suspender, the longitudinal and transverse damper by 3.42%, 18.13%, 30.11%, and 21.28%, respectively, while keeping the damping system cost basically unchanged. Meanwhile, the adaptability of the optimized damping system to large permanent fault displacement is improved.

Existing seismic resilience assessment frameworks for bridges often fail to adequately consider the repair sequence of components in functionality restoration models. Therefore, a post-earthquake functionality restoration model and seismic resilience assessment methodology considering component repair sequences were proposed. First, a segmented functionality restoration model that incorporates the influence of the component repair sequence during the bridge repair process was developed using typical analytical restoration models. Second, a graphical approach is employed to derive a simplified formula for the resilience indicator for the proposed restoration model, and a seismic resilience assessment framework considering the component repair sequence is proposed. Finally, a four-span continuous-girder bridge was selected as a case study, and various component damage scenarios were generated using fragility parameters and Monte Carlo sampling. The proposed resilience assessment framework was validated, and the impacts of different component repair sequences on the seismic resilience of an example bridge were explored. The results showed that the proposed model better reflects the repair sequence of bridge components and the contribution of the component repair process to functional restoration than typical restoration models. The derived simplified formula for the resilience indicator overcomes the computational difficulties associated with integration into the traditional resilience quantification method and improves computational efficiency without compromising accuracy. The resilience indicators of the bridges under various component repair sequences differ. The variation in the resilience indicator depends on the component damage scenario and functionality loss of the bridge. By reasonably optimizing the component repair sequence, the seismic resilience of small-to-medium-span continuous girder bridges can be effectively enhanced without changing their structural design.

Over 80% of bridges in China are of small to medium spans, making it critical important to rapidly predict and analyze their seismic performance using machine learning (ML). However, the “black-box” nature of many ML models, due to their algorithmic complexity, often raises concerns about reliability and applicability in real-world scenarios. Therefore, developing interpretable ML models has become an urgent necessity. This study focuses on continuous girder bridges with small to medium spans, aiming to explore high-interpretability ML methods for predicting longitudinal seismic responses. To achieve this, two popular ML algorithms-neural network (NN) and support vector regression (SVR)-were employed, using bridge structural parameters and ground motion intensity measures as input features. Predictive models for bridge seismic responses were developed. The interpretability of the predictive models was systematically analyzed using four machine learning interpretability methods, including SHapley Additive exPlanations (SHAP), Permutation Importance (PI), Partial Dependence Plot (PDP), and Local Interpretable Model-agnostic Explanations (LIME). The findings indicate that both NN and SVR can accurately predict bridge seismic responses, with coefficients of determination exceeding 0.9. Interpretability analyses using SHAP and LIME based on the NN models provide relatively stable and reliable explanations. Additionally, the high correlation among ground motion intensity features results in competitive contributions to the prediction outcome. Removing features with less contributions to predictions not only preserves predictive accuracy but also reduces model complexity, thereby enhancing interpretability. Based on the predictive performance and interpretability of ML models, the average spectral acceleration (AvgSa), Peak Ground Displacement (PGD), Peak Ground Velocity (PGV), and Housner Intensity (HI) are recommended as the most suitable seismic intensity measures for ML-based surrogate modeling of bridge seismic responses.

Segment self-centering bridges have attracted significant attention in bridge engineering research because of their good resilience and seismic performance. However, to date, their application in high-intensity areas is rare. Therefore, it is particularly important to examine the seismic response of self-centering bridges in high-intensity earthquake areas. However, self-centering bridge structures are highly nonlinear and uncertain, which pose significant challenges for the accurate prediction of their seismic response. In this study, a bridge seismic response prediction model integrating Variational Modal Decomposition (VMD), Convolutional Neural Network (CNN), and Bidirectional Long Short-Term Memory (BiLSTM) was proposed using measured data from a self-centering bridge shaking table test and simulated data from a finite element model. The model adopted VMD to capture the frequency feature information of the ground motion data. Furthermore, the model used a CNN to extract the spatial features of the data, and the long-term time dependence of the data was mined using BiLSTM to accurately predict the seismic response of the bridge. In this study, we first conducted prediction training on bridge superstructure response data under five types of seismic ground motions with different amplitudes measured from the shaking table test. We then conducted prediction training on the simulated data using the finite element model to understand the dynamic response of the bridge under larger-amplitude seismic ground motions. The results of the two predictions demonstrate that the proposed algorithm exhibits superior prediction accuracy and robustness when compared with four comparative algorithms of LSTM, RNN, SVR, and XGBoost. The model evaluation indices of R2 demonstrated improvements of approximately 11.9% and 3.2% when compared to SVR and LSTM models, respectively. Additionally, RMSE and MAE indices demonstrated reductions of approximately 52.4% and 32.5% and 49.6% and 30%, respectively, when compared with SVR and LSTM models.

Aiming at the difficulties in overcoming the pavement roughness, the significant influence of noise, and the pitching of single-axle vehicles when identifying the mode shape of bridges based on the indirect measurement method, a new method is proposed to indirectly identify the low-order frequency and mode shape of girder bridges based on the synchronized stationary acquisition by complex Morlet wavelet and symmetric dual-axle inspection vehicle with multiple degrees of freedom. Firstly, the dual-axle vehicle was parked at each synchronous measurement point of the bridge to collect the axle acceleration response for 30 seconds. The corresponding axle contact point response was obtained based on the inversion of the signal, and the wavelet denoising optimization algorithm was incorporated to reduce noise interference. The corresponding axle contact point response was obtained based on the inversion of the signal. Then, the wavelet denoising optimization algorithm was incorporated to reduce the noise interference. The frequency range of the bridge was obtained based on the wavelet coefficients of the denoised signal time spectra. Then, the continuous complex Morlet wavelet transform was performed on the axle contact point response. The optimal parameters were determined by combining the wavelet coefficients' Shannon entropy parameter selection criterion. The low-order modal shape of the girder bridge was constructed from the local to the whole by adopting the mean value processing method, so as to establish the framework of the process of the identification of the low-order frequency and modal shape of the girder bridge based on the axle contact point response of each synchronous measurement point. The effects of random vehicle excitation, bridge damping ratio and environmental noise were parametrically analyzed through numerical simulation. Finally, the method's applicability was verified by relying on a bridge in Chongqing. The results show that the proposed method has high accuracy in low-order mode shape identification, and can still identify the low-order frequencies and vibration shapes of girder bridges synchronously under the condition of random vehicle excitation, and is not affected by the roughness of the road surface, and has good robustness to the environmental noise and the damping ratio of the bridge. The method in this paper has certain advantages over the statistical moment, transfer rate and random subspace methods in terms of identification accuracy or computational efficiency, which provides a basis for the indirect measurement method to be popularized and applied in the actual field.

Research progress in parametric modeling, intelligent optimization, scheme generation, and intelligent details of bridge structures was systematically reviewed in this paper, and the development trend of the intelligent design of bridge structures was predicted. Current research on bridge modeling mainly includes parametric building information modeling (BIM) and parametric finite-element modeling. The BIM obtained from parametric modeling was mainly used for visualization, which is difficult to convert into a finite element model (FEM) for structural analysis and optimization. The FEM obtained from parametric modeling could be used for the analysis and optimization of structures. However, the layouts of the loads and boundary conditions for a relatively complex FEM were still manually implemented. Moreover, it was difficult to accurately convert FEM into a BIM model. In research on the intelligent optimization of bridge structures, heuristic algorithms remain the dominant approach for optimization, which mainly focuses on reinforced concrete structures. Few studies have been conducted on steel and steel-concrete composite structures. In addition, most studies have focused on single-load cases. The generative approach can be used for the rapid generation of bridge schemes. However, current studies have only considered span design and bridge-type selection. There have been few research results on the intelligent detailed design of bridge structures, and the integration of intelligent detailed design and digital manufacturing has not been considered. It is foreseeable that through intelligent design technology, the main direction of bridge structural design development will be the implementation of the intelligence of scheme generation, modeling, optimization, and detailed design of bridge structures, along with the integration of digital schemes with subsequent digital manufacturing and intelligent construction.

To review the latest technological advancements in the detection of underwater structural defects in bridges, this study focused on innovative applications of underwater robots to improve detection accuracy and efficiency. Underwater robots can carry both non-contact detection devices, such as optical and acoustic sensors, and contact detection devices, such as ultrasonic instruments and rebound hammers, demonstrating their potential for efficient detection in complex underwater environments. This study undertook a detailed analysis of the adaptability and improvement techniques of non-contact detection methods based on optical and acoustic principles, highlighting effective approaches for enhancing image quality and detection accuracy. It also clarified the current status of underwater contact detection research and proposed a solution for the collaborative operation of underwater robots with contact detection devices. This work emphasizes that the future direction of underwater detection lies in the use of underwater robots equipped with contact-detection devices. The challenges faced by current underwater bridge structure detection technologies are summarized, and new underwater detection methods based on intelligent algorithms and multisource data fusion are proposed, offering specific directions and technical paths for future research.

The conventional design process of bridge structures relies heavily on the experience of designers. The structural configuration proposal is repeatedly modified, resulting in low modelling efficiency. In addition, the optimal results are hard to achieve optimality. To this end, the intelligent modelling and optimization method of half-through steel box arch bridge with composite bridge deck system was proposed in this paper. First, the structural intelligent modelling method based on the human-computer collaboration was proposed. The automated layer classification method was adopted to extract the key information such as the arch rib, K brace, bridge deck system, and bridge pier abutment in the initial condition drawings. The cross-sectional information module of the bridge components was defined. The spatial information reasoning was conducted combined with the extracted information and the length and spatial coordinates of components were determined. The dividing criteria of elements and sectional fiber was established and the number of nodes and elements was determined, realizing the automatic definition of nodes and elements. The boundary condition module was defined. According to the structural parameter module, the load input module and structural output module were defined, realizing the intelligent modelling. Based on the parametric FE model, the structural intelligent optimization method was proposed. Parameters were input with combination of the intelligent modelling technology, forming the FE model to calculate the pseudo-objective function. The standard genetic algorithm, strengthen elitist genetic algorithm, and differential evolution algorithm were employed to optimize the structural cost. The proposed intelligent FE modelling and optimization method was verified with combination of the practical engineering case. The results indicate that the parametric FE model could be established by only marking the initial condition drawings using the proposed intelligent modelling method, which can significantly improve the modelling efficiency and quality. Compared with the manual structural optimization method, the cost of structural material is reduced by about 37.9% and the optimization period is reduced by about 74% using the proposed optimization method.

Lateral collaborative performance is a key indicator for evaluating the service condition of prefabricated beam bridges. The timeliness and applicability of traditional methods for obtaining lateral collaborative performance evaluation indicators can still be further improved. To address this issue, this paper proposed a dynamic characterization model for the lateral collaborative performance of prefabricated beam bridges based on the analysis of vehicle-induced response mapping relationships. This method can use the health monitoring data to realize the dynamic characterization of the changes in the lateral collaborative performance of bridges. It established a vehicle-induced response mapping model for prefabricated beam bridges using a Bayesian optimization Natural Gradient Boosting (Bo-NGBoost) model, and evaluated the location and extent of lateral connection damage based on the error distribution and magnitude of the response mapping models at different positions caused by lateral connection damage. The effectiveness of the proposed method was verified. The results showed that the Bo-NGBoost had certain advantages in terms of accuracy and robustness compared with standard NGBoost and Long Short-Term Memory (LSTM) networks. Under typical working conditions, its average coefficient of determination (R²) reached 0.986, representing improvements of 4.4% and 36.4% over standard NGBoost and LSTM, respectively. In particular, in the emergency lane region with sparse effective data, the model maintained a high R² of 0.953, whereas the R² of LSTM was only 0.374. Numerical simulations were conducted to consider various combined working conditions of different pavement roughness and traffic flow density. The R² of the deflection response was higher than 0.951, and the root mean square error (RMSE) was below 0.153 mm. In addition, by analyzing the error distribution and magnitude of the response mapping models at different positions, the location of lateral connection damage can be identified and its extent evaluated. In actual monitoring scenarios, the mapped strain responses of the model have an R² exceeding 0.981 with the true data, and an RMSE below 2.381×10^-6, with the true values generally falling within the 95% confidence interval. This indicates that the method not only accurately reflects the mapping relationship of the lateral vehicle-induced response of bridges, but also can locate and evaluate lateral connection damage, thereby providing strong support for the maintenance of prefabricated beam bridges.

Uncertainty analysis of structural performance indicators based on the statistical characteristics of measured influencing parameters is crucial for the accurate evaluation of prestressed concrete structures. However, due to the complex coupling of multidimensional uncertain influencing parameters in structures, existing evaluation methods have difficulty fully revealing the propagation law between measured random distributions of influencing parameters and the uncertainty of performance indicators. To address this issue, This paper proposes a probability density estimation method for concrete structures based on uncertainty propagation theory, incorporating the measured distribution characteristics of effective prestress. Firstly, the dimensionality reduction integration method is introduced to decouple the multidimensional prestress system into a linear superposition of single-variable subsystems. Combined with the measured prestress distribution characteristics, an analytical calculation method for the first four statistical moments of performance indicators is proposed. Secondly, using the statistical moments of structural response as constraints and based on the kernel density maximum entropy principle, the number of kernel functions is incorporated into the optimization framework. An objective function that comprehensively considers maximum entropy and statistical moment errors is proposed, establishing an improved kernel density maximum entropy method for estimating the probability density of prestressed concrete responses. The results show that, compared with traditional Monte Carlo simulations, the errors of the first to fourth-order moments of the statistical moment analytical solution are 0.01%, 0.11%, 0.29%, and 0.57%, respectively, which can accurately characterize the main stochastic features of the performance indicators. The improved kernel density maximum entropy method overcomes the limitation of traditional methods that focus solely on maximum entropy constraints, significantly enhancing the fitting accuracy of performance indicator probability density. Through engineering validation, the theoretical calculation error of the mid-span camber value of a prefabricated T-beam is controlled within 5%, and the measured camber values fall within the theoretical envelope region defined by the 95% confidence interval of elastic modulus. The results demonstrate that the proposed method exhibits good agreement with experimental data, effectively quantifies the impact of prestress uncertainty on structural performance, and provides theoretical and practical support for the performance evaluation of prestressed concrete structures.

Horizontal swivel construction, characterized by minimal disruption to existing railways and highways, has become a critical method for constructing long-span bridges over operational railways. However, swivel construction presents challenges, such as heavy loads, elevated overturning risks, and rapid state transitions. Therefore, effective monitoring and control are essential to ensure safety, structural stability, and precision. This study clarifies the primary and secondary objectives of swivel-monitoring feedback control and establishes their logical interconnections. The “traction-rotation” process was decomposed into five monitoring components: support system, traction system, swivel posture, structural stress, and surrounding environment. Both corresponding key and general monitoring indicators were proposed. Threshold calculation methods and state evaluation criteria were established for two typical swivel support configurations: swivel hinge-centered support and combined swivel hinge-support foot systems. A data linkage model was developed based on the physical interactions between the external environment, support system, traction system, swivel posture, and structural stress, and a dynamic feedback and control mechanism was established to ensure “mobility, stability, accuracy, and speed.” The ARIMA time-series model was incorporated to facilitate the dynamic prediction of monitoring data, allowing timely control of traction and balance systems. By integrating inertial sensing, image recognition, digital twins, and PLC wireless communication technologies, a lightweight visual monitoring feedback control system was implemented using WebGL, Python, and the B/S architecture. This system was successfully implemented during the swivel construction of a cable-stayed bridge. The results indicate that the primary objectives of monitoring and controlling rotating bridges are “mobility, stability, accuracy, and speed,” which follow a pyramid-shaped logical hierarchy, with mobility forming the foundational tier. The proposed monitoring-feedback control model for the swivel construction of bridges elucidates the interrelationships between control targets, monitoring indicators, state evaluation, and system responses. The developed lightweight, visualized swivel construction monitoring feedback control system enables real-time monitoring and regulation of traction, support, posture, and velocity states, effectively addressing challenges such as data redundancy, fragmented analysis, and state assessment and control.

The stability of high-fill embankments on mountain roads is affected by fill strength and hydrogeological conditions. In 2023, an expressway section between K83+400 m and K83+800 m encountered serious problems, such as roadbed settlement, pavement cracking, structural damage, stabilizing pile tilting, and retaining wall seepage. On-site investigations indicated that the presumed causes of the accident were insufficient fill material compaction and a significant increase in groundwater levels. The real-time monitoring of slope deformation in damaged sections using GNSS technology directly reflects the stability of the embankment. Based on finite element limit analysis, the following numerical simulations were conducted: plate elements were used to simulate stabilizing piles, and the equivalent strength and stiffness parameters of the sections were calculated according to the reinforcement diagram, which was validated by a classic slope case; the water head height on both sides of the slope was specified to calculate the steady-state seepage field of the slope, and a drainage line with zero water pressure was used to simulate deep drainage holes; the impacts of stabilizing pile reinforcement, groundwater seepage, and drainage measures were sequentially considered; and the stability of the high-fill slope at a typical section was calculated for five different cases. By comparing the simulation results with the actual groundwater level and slip surface, it was confirmed that groundwater seepage directly led to instability of the high-fill embankment and the effectiveness of the disease treatment measures was verified. Further research revealed that the safety factor for the actual embankment slope was negatively correlated with the relative groundwater level in front of the stabilizing pile. When it exceeded 0.4, the downward trend of the safety factor was aggravated. This study provides an important reference for the design, construction, operation, and emergency management of mountainous highways in China.

The current subgrade resilience modulus test method does not take into account the structural constraint effect of the overlying pavement and the influence of the real vehicle loading characteristics, resulting in the test resilience modulus is difficult to reflect the actual bearing capacity of the subgrade, which brings a lot of uncertainty to the construction of highways. In light of this, this paper proposes an in-situ testing method for the dynamic and static resilient modulus of subgrade. Based on simulating the self-weight and constraint effects of pavement structural layers using a ring, this method can investigate the influence of vehicle load characteristics, including load frequency and amplitude, on the structural modulus of the subgrade. The dynamic and static elastic modulus of the subgrade were tested by assembling different sizes of loading plates and different widths of rigid rings, to find out the best loading combination of loading plates and rigid rings, so as to optimize the structural system of testing methods for dynamic and static elastic modulus of the subgrade, and to provide an important means for the stiffness test of the subgrade in the actual use stage of the newly constructed or reconstructed subgrade. The test results show that the rebound modulus of the subgrade is positively correlated with the dynamic load frequency and dynamic load amplitude, and negatively correlated with the bearing plate size. For Φ30 cm load bearing plate without collar, the test value of resilient modulus increase by 119% when the dynamic load frequency is increased from 1 Hz to 10 Hz, and the test value of resilient modulus increase by 38.7% when the dynamic load amplitude is increased from 0.1 MPa to 0.35 MPa. In addition, for the same bearing plate size, the test value of the resilient modulus of the subgrade increases and then stabilizes with the increase of the constraint range of the overlying pavement structure simulated by the collar. For the 30 cm bearing plate, the modulus tends to be stable when the width of the ring reaches 20 cm, and the modulus increases by 13% compared with that of the unassembled ring; for the 20 cm bearing plate, the modulus tends to be stable when the width of the ring reaches 30 cm, and the modulus increases by 35% compared with that of the unassembled ring; for the 10 cm bearing plate, the modulus tends to be stable when the width of the ring reaches 35 cm, and the modulus increases by 31% compared with that of the unassembled ring. Meanwhile, based on the results of full-scale in-situ test, this paper establishes the rebound modulus conversion model of subgrade with different sizes of load bearing plates, the rebound modulus conversion model of subgrade with different rigid ring widths of the same load bearing plate, and the rebound modulus conversion model of subgrade with different loading frequencies. To realize the mutual conversion between the test results of various sizes of load bearing plates and various widths of collar combinations under different frequencies of dynamic loads, and the accuracy of the conversion model was verified through the site field test.

Strengthening using outer steel tubes and sandwiched concrete jackets is a new reinforcement technology that utilizes the performance advantages of composite structures. By using this technology to reinforce existing concrete-filled steel tube (CFST) columns, a new steel-concrete composite structure with excellent performance can be formed to quickly and controllably improve the mechanical performance of existing CFST columns. The bond behavior of the steel-concrete interface is the foundation for ensuring the coordinated work of the various components of the composite structure. In this study, interface push-out tests of 16 CFST columns strengthened with an outer square steel tube and sandwiched concrete jacket without initial stress were conducted, and bond behavior of the interface between the outer steel tube and sandwiched concrete (interface To) as well as that of the interface between the inner steel tube and sandwiched concrete (interface Ti) was studied. The study found that the load-slip curve characteristics of the interface To and interface Ti specimens are similar. Based on the characteristics of the curves, they can be divided into three stages: bonding, slip, and friction. During these stages, the slip transitions from shear deformation in the interface layer to micro-movement and sliding at the interface. The damage process transitions from the generation and development of micro-cracks to the extrusion and fracture of micro-protrusions to the wear and deposition of concrete. Chemical bonding, mechanical locking, and frictional resistance play predominant roles in this sequence. The study showed that for interface To, bond damage develops later, the energy dissipation capacity is greater, and the mechanical locking strength (τ_w) and friction strength (τ_f) are lower. The self-stress of sandwiched concrete can increase the elastic bond shear stiffness (K_{s, e}), accelerate the development of bond damage in interface Ti, enhance the energy dissipation capacity, and increase τ_f. When the strength of sandwiched concrete increases from C50 to C60 and C70, K_{s, e} increases by 23.6% and 69.9%, bond damage develops earlier, τ_w increases by 28.3% and 27.8%, and τ_f increases by 13.0% and 22.6%, respectively. When the thickness of the outer steel tube increases from 3.5 to 4.5 and 5.5 mm, K_{s, e} increases by 10.8% and 65.6%, the energy dissipation capacity enhances, and τ_f increases by 34.5% and 72.9%. Based on these results, a bond-slip constitutive relationship model was proposed. The calculated results of the model were in good agreement with the experimental results.

This study utilized field-testing methods to conduct explosion tests on concrete girder bridges to investigate the failure characteristics and dynamic response of concrete beam bridges subjected to above-deck explosion loads and provide a theoretical basis for the anti-explosion design of concrete bridges. Four cases were established in the tests, considering the integrity of the deck pavement layer, various explosion positions, and different explosive equivalents. Measurements of the dynamic strain, acceleration, and visible damage, such as cracks in the main girder, were performed before and after the tests. An analysis of the failure characteristics and dynamic responses was conducted. The results indicate that the shape of the breach on the bridge decking after a near-field explosion is closely related to the local distributions of the longitudinal and transverse stiffnesses in the section. When the longitudinal and transverse stiffnesses are similar, the breach is circular; by contrast, when the longitudinal and transverse stiffnesses are considerably different, rectangular breaches are formed. The variation in mechanical response of a bridge deck to near-field explosions is dependent on the explosive equivalents. When subjected to small equivalent explosive loads, the bridge deck is equivalent to be subjected to a concentrated force. In contrast, when subjected to large equivalent explosive loads, the bridge deck may be penetrated, and the force transmission effect of the bridge deck disappears. In the case of the same amount of explosives, a single explosion causes more damage to the structure than multiple explosions. Severe initial damage affects the dynamic response of a beam considerably when subjected to explosive loads. The dynamic response of the beam is not proportional to the amount of explosive but is influenced by the distance from the explosion.

In the structural health monitoring of tunnels, self-diagnosis and condition assessment are reliant on data. However, sensor-monitored data often contains considerable abnormalities. These abnormalities comprise both interference information caused by nonstructural factors and damaged information caused by structural factors. This issue is primarily resolved by extracting valuable damage information from the abnormal data and eliminating irrelevant interference. To this end, this study proposes a sliding slope anomaly detection and data reconstruction method that is based on the instability and discontinuity of abnormal data change rates. First, the data were segmented using a sliding window. Subsequently, a least squares linear fit was performed on the data within each window to obtain the slope and intercept. Thresholds were set based on the slope variance and goodness of fit, and slope and intercept values exhibiting high dispersion and poor fit were identified and eliminated. The data were then reconstructed using regression calculations and the median method. Finally, the effectiveness of the proposed method was compared with the traditional 3σ method, sliding median filtering, wavelet transform, and empirical mode decomposition method for analyzing the abnormal data of damage tests on reinforced concrete structures and field monitoring. The results indicate that the proposed method can effectively clean gain and bias data, which other methods struggle to handle. Furthermore, the proposed method can identify and remove abnormal data trends without compromising the original structural damage data, thus ensuring quality monitoring data and meeting practical application requirements.