Accurate identification of cable force in cable-stayed bridges is essential for ensuring the safe operation of the structure. The vibration method is a commonly used approach in the field of cable force identification. However, traditional force-frequency formulas are often derived based on the assumption of ideal boundaries, making it difficult to account for practical issues such as boundary disturbances at the pylon and girder ends, as well as interference from embedded pipes at both ends and internal and external dampers. To address these issues, this study proposes a cable force identification method that accounts for the coupled vibration of the tower-cable-beam-damper system, aiming to mitigate the effects of complex boundary changes and vibration-control devices on cable dynamics. First, the transmissibility functions among the tower-cable-damper responses are derived, effectively eliminating the interference of tower-beam-damper vibrations in cable spectrum analysis. Second, the partial coherence functions among the tower-cable-damper responses are formulated to quantify the contributions of different excitation sources to the cable vibration response. Third, a cable model updating method is developed, using frequency targets identified from the transmissibility function. Subsequently, by designing an optimization space for cable force and bending stiffness, time-varying cable forces are identified. Finally, field tests on a cable-stayed bridge validate the accuracy of the proposed method, the test results show that, for cables with an external damper, the calculation error of existing cable force identification formulas exceeds 40%, whereas the proposed method maintains a calculation error of less than 8% for both long and short cables, thereby providing an effective and practical method for tracking and identifying cable forces in cable-stayed bridges equipped with an external damper.

Cable force is a key indicator for evaluating the in-service state of stay cables. However, fluctuations in cable force caused by temperature variations can mask anomalous changes in cable force due to damage. To address this, this paper proposes a cable force prediction and anomaly assessment method based on nonlinear modeling of “multi-point temperature-cable force”. First, the nonlinear mapping mechanism between temperature changes in components (pylon, girder, cables) and changes in cable force was analyzed, with full consideration of the influence of multi-section and multi-point temperature variations. The basic form of the “multi-point temperature-cable force” prediction model was presented. Then, an objective function for evaluating the prediction effectiveness of the model was constructed. An intelligent optimization method for key parameters of the random forest (RF) algorithm, based on the particle swarm optimization algorithm, was proposed. A dimensionality reduction strategy for multi-point temperature monitoring data using principal component analysis was formulated, thereby establishing a method for constructing the cable force prediction model based on the optimized RF algorithm. Furthermore, a method for establishing cable force anomaly thresholds based on the probability distribution of modeling residuals from the cable force prediction model was proposed, along with a probabilistic assessment method for cable force anomalies. Finally, the accuracy of the cable force prediction model and the effectiveness of the cable force anomaly assessment method were verified using monitoring data from a single-pylon cable-stayed bridge. The results show that, compared to linear models, the proposed method more accurately characterizes the relationship between temperature and cable force in cable-stayed bridges. It provides an effective technical approach for identifying early-stage cable force anomalies under temperature variations. This study is expected to serve as a decision-making reference for the preventive maintenance and targeted inspection of cable-stayed bridges.

Due to multiple factors such as environmental disturbances and sensor drift, the actual monitoring data noise of bridges exhibits significant non constant characteristics, which limits the accuracy of traditional temperature-induced strain prediction models based on the assumption of constant noise. This paper is proposing a temperature-induced strain prediction method for bridges using variational heteroscedastic Gaussian process regression (VHGPR). A two-layer probabilistic framework that integrates structural response learning with noise-variance estimation was developed, and a heteroscedastic modeling strategy was introduced to explicitly capture non-constant noise. High-dimensional posterior inference was transformed into a low-dimensional optimization problem relying only on a small number of parameters by structural reconstruction of variational parameters. On this basis, a marginal variational free-energy-based modeling and optimization strategy was proposed, and a variational-inference-driven analytical method for approximating the joint posterior was developed. The non-constant time-varying law of strain noise level and amplitude of main girder under temperature effect was revealed. The effectiveness of the proposed method was verified by long-term monitoring data of an existing cable-stayed bridge. The results show that the VHGPR model enables probabilistic prediction of temperature-induced strain under non-constant noise, provides accurate estimates of the main-girder mean strain, and effectively quantifies predictive uncertainty. The predictions are reliable, and the model exhibits strong capability in uncertainty quantification. The time-varying noise characteristics are effectively captured through the integration of variational inference and a heteroscedastic modeling strategy. The proposed method outperforms conventional Gaussian process regression across multiple evaluation metrics. The root-mean-square error decreases by 14.4%, the mean absolute error decreases by 12.7%, and the coefficient of determination increases to above 0.96.

Structural damping is a key parameter for bridge dynamic analysis and vibration-mitigation design. However, under low damping, short free-decay records, and multimodal coupling, damping estimates are often scattered and highly noise-sensitive. Building on our previous Power Spectrum Decrement Method (PSDM), this study proposes an Enhanced Power Spectrum Decrement Method (EPSDM) by integrating the power-spectrum decrement concept with Frequency Domain Decomposition (FDD). A multi-channel power spectral density (PSD) matrix is first constructed and subjected to singular value decomposition, where the first singular value spectrum and associated singular vectors used to extract modal frequencies and mode shapes. An analytical relationship is then established between the temporal decay of the first singular value amplitude and the modal damping ratio, enabling multimodal damping estimation without time-domain modal separation or inverse transformation. Moreover, the Modal Assurance Criterion (MAC) is introduced for effective band selection, enhancing robustness against closely spaced modes and high noise levels. The eight-story shear frame example demonstrates that EPSDM achieves high identification accuracy for damping ratios across multiple modes, outperforming PSDM and the logarithmic decrement method overall. Compared with ERA, EPSDM exhibits better robustness under strong noise conditions and offers the advantage of eliminating the need for repeated model-order tuning. Further validation is conducted through a moving-vehicle excitation test on a long-span steel box-girder cable-stayed bridge, where short free-decay responses are used to identify the frequencies, mode shapes, and damping ratios of the first 12 vertical bending modes, yielding physically reasonable and highly repeatable results. The proposed EPSDM provides an accurate and efficient tool for multimodal damping identification from short-duration free vibration and multi-channel monitoring data, supporting bridge performance assessment and the calibration and evaluation of vibration-control parameters.

Constructing a lightweight structural health monitoring and safety assessment method for urban bridge groups is essential for ensuring efficient urban traffic operation and improving overall safety. However, existing studies have mainly focused on individual bridges, and a systematic framework applicable to bridge groups has not yet been established. This gap limits the effectiveness of network-level management and risk identification at the urban scale. Spaceborne Interferometric Synthetic Aperture Radar (InSAR) provides high-precision time-series observations with wide spatial coverage and offers a new paradigm for large-scale bridge safety inspection. This study proposed a multi-scale safety assessment method for urban bridge groups by integrating multi-source remote sensing data and Distributed Scatterer InSAR (DS-InSAR) time-series deformation analysis. Structural parameters, including bridge location, type, and geometric dimensions, were first extracted from high-resolution optical remote sensing data. The line-of-sight (LOS) time-series displacements derived from DS-InSAR were then used as the primary observations. The coupling between LOS deformation and air temperature was analyzed across different bridge types. Deformation rate and thermal expansion response were introduced to preliminarily identify bridge service deterioration. Subsequently, a comprehensive safety evaluation framework was developed based on the Analytic Hierarchy Process (AHP), enabling the quantitative mapping from macroscopic deformation to structural risk states. A case study was conducted on 48 representative bridges in Changsha. Based on the InSAR-derived time-series deformation, six bridges exhibiting potential abnormal responses were identified. Their structural risks were further quantified and classified using the proposed framework. The results show that the method enables efficient identification and grading of safety states at the bridge group scale and provides reliable support for urban bridge maintenance and risk early warning.

These incidents now represent a serious threat to bridge structures, public safety, and the sustainable development of maritime navigation. Vision-based ship detection and tracking serve as key approaches for the early identification of potentially hazardous ships. However, the collection and annotation of large-scale image datasets is time-consuming and labor-intensive, resulting in limited detection performance during the initial deployment phase due to sample scarcity. In addition, the diversity of ship types, significant scale variations, and complex background interference within waterways further hinder the accuracy of ship recognition. To address the aforementioned challenges, this study initially constructs a three-dimensional proxy virtual environment to simulate navigable bridge waterways. Using this environment, a synthetic image dataset is generated to enrich the training data, featuring six typical ship categories and diverse complex scenarios. Subsequently, Contextual Transformer modules are embedded into the You Only Look Once version 8 (YOLOv8) to enhance contextual awareness. The Wise Intersection over Union version 3 loss function is introduced to improve the model's focus on critical positive samples. Thereafter, the improved YOLOv8 is pretrained using the synthetic dataset and integrated with ByteTrack to build a baseline ship detection and tracking framework. Ultimately, a supervised domain adaptation strategy is employed to fine-tune the model using a mixture of synthetic and real-world images, enabling robust multi-object ship detection and tracking in real navigation scenarios. The results demonstrate that the improved ship detection and tracking framework outperforms the original model. When tested on real-world videos, the model adapted through supervised domain adaptation achieves a Multiple Object Tracking Accuracy of 82.9%, indicating strong tracking accuracy and continuity. Moreover, the use of virtual-to-real-world transfer techniques effectively reduces the reliance on large quantities of real-world samples for ship detection and tracking tasks.

For steel structures under long-term vibration loads, bolt connections are prone to loosening, and under complex working conditions, existing visual inspection techniques often fail to identify minor loosening due to texture blur and geometric distortion. By integrating super-resolution reconstruction and multi-scale perception, an intelligent bolt loosening detection method based on YOLOv26-CABU-Net was proposed. First, the YOLOv26 model was trained using a data augmentation strategy that simulated haze and motion blur, enabling accurate localization of bolt head regions of interest against complex backgrounds. Second, to overcome the limitation of existing methods that rely on regular arrangements, a RANSAC (Random Sample Consensus) perspective correction algorithm based on bolt head center point matching was proposed, effectively eliminating geometric distortion caused by irregular arrangements and large tilt angles. Then, the CABU-Net semantic segmentation network was constructed, which innovatively embedded a Real-ESRGAN super-resolution module and a convolutional block attention module (CBAM), achieving active detail restoration and high-precision contour segmentation for low-resolution bolt images. Finally, abandoning traditional straight-line fitting schemes, a corner positioning algorithm based on contour centroid distance spectrum peaks was designed, enabling loosening detection by quantifying the angular difference of corners. Test results show that the method achieved an object detection accuracy (AP) of 0.996 and a mean intersection over union (mIoU) for contour segmentation of 0.993. Under conditions including shooting distances within 2.0 meters, large tilt angles up to 60°, and varying lighting, the maximum error in loosening angle detection was strictly controlled within 4.0°. Compared to traditional algorithms, this method significantly improves robustness and detection accuracy under complex shooting conditions.

To address the challenges of scarce high-quality annotated data and poor generalization performance in vision-based bridge structural crack detection, this paper proposed a zero-shot interactive prompt segmentation method. Firstly, the visual foundation model SAM2 (Segment Anything Model 2) was introduced to construct a zero-shot segmentation framework for bridge cracks. Secondly, the relationship between different prompt strategies and model performance was systematically investigated to provide corresponding optimal prompt strategies for cracks of different morphologies, and the effect of iterative mask refinement on model performance improvement was explored. Finally, the proposed method was systematically compared with image processing and deep learning methods on multiple public crack datasets, and the model's robustness and generalization performance under noisy data and cross-domain crack scenarios of concrete road and steel bridge were tested. The results show that: model performance exhibits a trend of first increasing, then stabilizing, and finally decreasing as the number of prompt points increases; multi-point prompt strategies should be prioritized for crack segmentation tasks, and using prediction results as mask prompts for iterative refinement can further improve segmentation accuracy. Without relying on annotated data, the proposed method can achieve 87.8% of the performance of supervised deep learning methods, demonstrating excellent zero-shot segmentation capability. For typical noisy datasets, the SAM2-based model exhibits stronger robustness compared with deep learning methods. When migrated to the concrete road crack dataset, the performance of all methods declines to varying degrees, primarily due to the complex crack morphology and strong environmental interference in the images; when migrated to the steel bridge dataset, the performance of the proposed method increases by 22.45%, while deep learning methods decrease by an average of 38.02%, demonstrating superior generalization performance. The research results provide a new technical approach for structural health monitoring.

Bridge displacement can directly reflect bridge condition, and its accurate measurement provides crucial information for monitoring bridge condition. Existing UAV-based vision methods for displacement measurement usually rely on stationary references or artificial targets. For bridges spanning rivers or roads, insufficient stable references and inconvenient target installation limit their practical application. This paper proposed a target-free UAV-based approach for bridge displacement measurement aided by multi-laser projection. First, parallel laser beams were used to project multiple laser points onto the bridge surface to construct a stationary reference with a physical scale. By analyzing the motion of the target relative to this reference, the influence of UAV translation and rotation can be effectively reduced without camera pose information. Subsequently, a joint subpixel localization method and an improved Kanade-Lucas-Tomasi (KLT) algorithm were developed. These techniques improved the robustness of laser spot positioning and natural feature tracking under complex environmental conditions. Finally, the image scale factor is calculated in real time using the geometric features of the projected multi-laser points. This enabled the conversion from pixel displacement to physical displacement without requiring any additional known-size objects. Experimental results show that the method performs well under different laser projection conditions and environmental disturbances. The maximum Root Mean Square Errors (RMSEs) of the horizontal and vertical displacement are 0.031 mm and 0.028 mm, respectively. In a field test on a three-span continuous girder bridge, the proposed method effectively captured the bridge deflection response induced by vehicle loads, thereby verifying its practical applicability. The proposed method provides an effective technical solution for non-contact displacement measurement of bridge.

To study the corrosion evolution characteristics and corrosion resistance of the zinc based multi-element alloy anchoring system for bridge cables, based on copper salt ice acetate salt spray corrosion tests, combined with scanning electron microscope (SEM) microscopic morphology observation , energy dispersive spectroscopy (EDS) techniques and high resolution X-ray diffraction (XRD) corrosion product analysis, the corrosion macroscopic morphology, corrosion products, corrosion quality changing, corrosion microscopic morphology, and variation of the corrosion microcracks' number and width of the cable anchoring system specimens were studied. The Arrhenius equation was introduced to analyze the corrosion resistance. The results showed that at the interface between the anchor material and the steel wire, the corrosion distribution of the anchor body is radial with the steel wire as the center, and the corrosion of the anchor body is denser and faster near the steel wire, with flaky corrosion points appearing at 15 days, continuous corrosion points at 30 days, and the formation of corrosion pits at 60 days. The anchor material-anchorage interfaces are mainly uniformly corroded, and the accumulation of corrosion products in the gaps exacerbates the outer corrosion. The corrosion quality changing rate of ZnAl9Cu1MgRE alloy anchor specimens is lower than that of ZnAl6Cu1 alloy, the former has stronger corrosion resistance. Observation of surface microstructure (SEM) shows that the number of microcracks at the “anchor material-steel wire” interface and the “anchor material-anchorage” interface increases rapidly and then slows down with the corrosion process. The growth rate is the fastest at 15-60 days, and decreases after 60 days due to the widening and merging of microcracks. Energy dispersive spectroscopy (EDS) analysis shows that the corrosion morphology characteristics of the anchor material, anchorage surface, “anchor material-steel wire” interface, and “anchor material-anchorage” interface are basically the same. The corrosion mechanism is that Cl^- pitting corrosion causes honeycomb like depressions, tumor like protrusions under the electric couple effect, volume expansion and ion diffusion interact to generate sheet-like stacking morphology, and hydrogen embrittlement and internal stress coupling promote microcracks. The X-ray diffraction (XRD) indicates that the main corrosion products on the surface of the anchor material are ZnO, Zn(OH)₂, Cu₆Al₂ (OH)₁₆CO₃ et al. The main corrosion products of the anchorage are Fe₃O₄ and Fe₅CuO₈, and the corrosion of Zn is accelerated at the “anchor material-steel wire” and “anchor material-anchorage” interfaces due to electrode effects. Based on the Arrhenius equation calculation, the apparent activation energies of ZnAl9Cu1MgRE alloy and ZnAl6Cu1 alloy at the “anchor material-steel wire” interface are 114.32 and 89.37 kJ·mol^-1, respectively, both of which meet the A-level high corrosion resistance standard, and the former has relatively better corrosion resistance.

To address the inefficiency, heavy workload, and high omission rates of the traditional fatigue crack detection methods for orthotropic steel bridge decks, a novel tracked climbing robot is proposed for high-efficiency image acquisition and high-precision size measurement for fatigue cracks. The robot consists of a mechanical mobility system, an image measurement system, and an electronic control system. It demonstrates reliable magnetic adsorption and climbing maneuverability, enabling stable adsorption to the top plate, diaphragm plate, and U-rib plate, as well as transfer between plate surfaces. The robot system is equipped with a brushless motor-driven single-axis gimbal which can adjust the camera's tilt angle according to the crack location. Additionally, a laser rangefinder enables rapid image self-calibration and accurate crack dimension measurement. Static and dynamic analyses were conducted to evaluate the robot's mechanical performance under adverse motion conditions. The locomotion stability and measurement accuracy of the robot were further validated on a full-scale model of steel bridge decks. The results indicate that the tracked climbing robot can track along the pre-defined inspection paths and effectively collect the images of various fatigue cracks across all critical regions of the bridge deck. The crack dimension measurement exhibits high accuracy with a maximum error of only 1.36%, and a sub-millimeter resolution level is reached.

The anchorage system of external prestressed tendons operates under sustained high stress and is susceptible to tendon-anchor slippage damage, which may lead to failure of the prestressing system. To address this, this paper proposes an innovative Elasto-Magneto-Electric (EME) method with built-in excitation coil for monitoring tendon-anchor slippage of external prestressed steel tendons. Specifically, the excitation coil is embedded in the internal gap of the external tendons, allowing the tendons themselves to form the closed magnetic circuit and thus eliminating the bulky yoke structure required in conventional EME sensors. The sensor is installed in the stress-free zone outside the anchorage plate, enabling rapid installation during bridge operation. Based on magnetic circuit analysis and the EME method, a damage detection method for tendon-anchor slippage is established. The results of finite element simulation demonstrate that the designed sensor's excitation power consumption is reduced by 91.2% compared with conventional externally-mounted excitation coil schemes. Moreover, the proposed damage characterization index exhibits a quadratic relationship with the number of tendon-anchor slippage. Finally, zoned monitoring tests for tendon-anchor slippage damage were carried out. For single-region damages, the localization is realized through the synergistic variation of multi-sensing-element signals, and the coefficients of determination (R²) of the corresponding quantitative calibration curves all exceed 0.998. For multi-region damages, an identification method based on a Bayesian-optimized back-propagation (BP) neural network is proposed. A maximum mean absolute error of 0.083 tendon is achieved. The findings of this study provide a new and efficient method for long-term real time monitoring of the anchorage condition of in-service external prestressed tendons.

Carbon fiber-reinforced polymer (CFRP) materials have shown significant potential in fatigue performance enhancement of steel structures. Existing research predominantly focused on evolution of structural performance under constant amplitude fatigue loading. However, in engineering practice, steel structures in critical infrastructure are often subjected to variable amplitude fatigue loading induced by wind, waves, and vehicular traffic. Such loading may induce load interaction effects. An experimental study was conducted on the fatigue performance of CFRP-strengthened steel beams under variable amplitude fatigue loading. A total of eight specimens were tested, considering effects of high-low load sequences and overload ratios. Key parameters, including crack retardation distance, bond failure length, and crack growth rate were carefully recorded. Experimental results demonstrated that CFRP strengthening effectively retarded crack propagation and significantly enhanced fatigue performance under variable amplitude fatigue loading, extending fatigue life by a factor from 11.2 to 15.6. Under low-high load sequences, no significant load interaction phenomenon was observed. In contrast, under high-low load sequences, the crack growth rate significantly decreased after the load transition, with a more pronounced retardation effect corresponding to a larger overload ratio. The retardation effect in strengthened specimens was less pronounced in comparison with the unstrengthened specimens, with the retardation distance being only about 50% of that in the unstrengthened specimens.

The batter piles show obvious effects in wharves, transmission towers, foundation pit support, slopes and other projects, but the engineering characteristics of the batter pile with single lateral surcharge is yet unclear, the promotion and application for subgrade engineering are restricted. In this paper, indoor model test was carried out, the influence of inclination angles and pile end constraints of the negative batter pile on the settlement of foundation, lateral displacement, bending moment, soil pressure, earth resistance, and the characteristics of p-y curve were analyzed, experimental basis for in-depth research and development application of passive negative batter pile with embankment or single lateral surcharge is provided. The results show that the settlement of the post-pile foundation and horizontal displacement of the negative batter piles significantly decrease and the bending failure load obviously increases with the increases of inclination angle and the enhancement of pile end restraint compared with the vertical pile. The maximum of soil pressure or earth resistance of the negative batter pile decreases with the inclination angle increasing or the pile end constraint enhancing. The p-y curve for the passive negative batter pile can be represented by the equation p/p_ub=α(y/y₅₀)^β. The variation in the values of α or β indicates the sensitivity of the p-y curve as it changes with depth. This sensitivity is observed to increase as the inclination angle decreases and as the level of restraint increases. In subgrade engineering, employing negative batter piles at the toe of a slope, along with an increased angle of inclination or embedding the pile's tip into a firm bearing stratum, can significantly enhance the effect of lateral restraint and reduce roadbed settlement. The application prospect of the negative batter pile is good.

This study aims at improving the low strength of mud cakes filter pressed from engineering slurry flocculation. A process of flocculation-curing-pressure filtration to treat high-moisture engineering slurry is proposed by adding a curing system before pressure filtration. The study investigates the effects of flocculation and curing sequences on solid-water separation through slurry flocculation and sedimentation tests. The feasibility of the proposed flocculation-curing-pressure filtration process and its enhancement on strength of filter mud cake are examined via fluidity tests and laboratory pressure filtration tests. SEM microscopic observations are employed to analyze the treatment mechanism. Results demonstrate that flocculation and curing of slurry should be conducted separately by sequence. The proposed process achieves faster short-term drainage rate and significantly improved the strength of filter mud cake, particularly after 7-day curing. Implementing 3% cement curing after flocculation maintains slurry fluidity while enhancing the cohesion and internal friction angle of filter mud cake to 42 kPa and 29° respectively, representing 7 times and 2 time of that mud cake treated with conventional flocculation-pressure filtration. SEM analysis reveals that the strength improvement is mainly due to the hydration products which fills the macro-pores between flocculated aggregates and forming robust cementation. This novel process eliminates secondary treatment requirements when using filter cakes as engineering fillers, demonstrating higher efficiency, cost-effectiveness, and promising application prospects.

Culverts are critical underground structures in highway transportation infrastructure. However, compared with traditional small-diameter circular pipe culverts, the application and research of large-span pipe-arch steel corrugated plate culverts (PASCPCs), which serve both traffic and drainage functions, remain rare in subgrade engineering. Based on an expressway project, the mechanical characteristics of a 9.5 m(span)×6.5 m (height) PASCPC during the backfilling construction stage were tested and analyzed. A dynamic vehicle-pipe-soil finite element model was established. After validating the model through vehicle loading tests, the influences of different backfill heights, vehicle parameters, and pipe structural parameters on the dynamic response of PASCPCs were investigated. The mechanical behavior of PASCPCs under compaction loads, static vehicle loads, and dynamic vehicle loads was comparatively analyzed. The research results show that the most unfavorable stressed part of the PASCPC is the waist. The soil arching effect causes the earth pressure distribution in the soil layer above the pipe crown to exhibit a nearly V-shaped pattern, which becomes more pronounced with increasing backfill height. The vertical de-formation and deformation rate at the pipe crown are significantly greater than the horizontal deformation at the pipe side. The vertical acceleration, vertical deformation, and stress response at the pipe crown all exhibit clear three-stage evolution characteristics. During vehicle passage, the horizontal acceleration and horizontal deformation responses at the pipe side show some fluctuations, with peak values occurring at the onset of driving. Backfill height considerably influences the vertical dynamic response at the pipe crown. Under the premise of meeting highway transportation and culvert serviceability requirements, reasonably controlling vehicle speed and tonnage, while selecting a relatively small span, large corrugations, and an appropriately thin steel plate, can effectively reduce the dynamic response and construction cost of PASCPCs. The dynamic amplification effect makes the pipe response under dynamic vehicle loads significantly greater than that under static vehicle loads, but this effect gradually weakens as the backfill height above the pipe crown increases. When the backfill height exceeds 6 m, the design of PASCPCs in similar projects should take the construction backfill load as the primary calculation basis.

Instability and subsidence are the main diseases of silt subgrade in Yellow River flood area. In the process of embankment filling and under the long-term traffic load after construction, the principal stress axis of subgrade soil element is constantly rotating. At present, the mechanical response of silt subgrade in Yellow River flood area considering the effect of principal stress axis rotation is still unclear, and the constitutive model that can reasonably describe its mechanical state is relatively limited. In this paper, the effects of major principal stress direction angle α and intermediate principal stress coefficient b on the mechanical properties of silt in Yellow River flood area are analyzed. Based on the fractional plastic mechanics, the three-dimensional non-orthogonal plastic flow rule is established, and the explicit expressions of the loading and plastic flow direction in the principal stress space are derived. The major principal stress direction angle α is introduced into the initial state parameters, and the hardening law considering the principal stress axis rotation and the change of the initial state is further proposed, and the three-dimensional elastoplastic constitutive model of the silt in Yellow River flood area is established in the principal stress space. The effectiveness of the model is verified through the fixed axis HCA test results of the silt in Yellow River flood area. The experimental phenomenon is effectively explained by the constitutive mechanism. The results show that the constitutive model can calculate the stress-strain relationship and octahedral shear stress-strain relationship of silt in the Yellow River flooded area under different principal stress directions. The intermediate principal stress coefficient b has the opposite effect on the non-orthogonal plastic flow stability at α<45° and α>45°, and determined the strength by affecting the hardening modulus. The anisotropic mechanical characteristics of silt at the influence of principal stress axis rotation are reasonably reflected.

The climate in plateau regions is complex and variable. Combined with the frequent occurrence of heavy rainfall in recent years, the instability and sliding of highway talus slopes have been further induced. To improve the adaptability of highway talus slopes in plateau areas to intense rainfall, this paper takes a typical talus-bedrock slope in the plateau region as the research object. Based on indoor reduced-scale model tests, seven working conditions with different stone contents and protection forms are designed. The evolution laws of moisture content, pore water pressure, earth pressure, displacement and particle size distribution of the talus-bedrock slope during rainfall infiltration are systematically investigated. The progressive failure mechanism under multi-field coupling is analyzed, and the performance of three engineering protection structures is evaluated. The results show that talus slopes with low stone contents (30%, 40%) experience sharp variations in moisture content, pore water pressure and earth pressure, which present a four-stage evolution and trigger sudden instability in the late rainfall period. Talus slopes with high stone contents (50%, 60%) possess strong permeability and drainage capacity, presenting gentle responses of moisture content and pore water pressure as well as slight earth pressure fluctuations, with only slope deformation occurs without overall instability. The loss of fine particles further weakens the shear strength of rock and soil mass, and the critical position of fine particle loss varies with stone content. Among the three protection forms, retaining walls have the best anti-sliding performance but the weakest interception capacity for sliding mass; gabion cages exhibit excellent drainage performance and are effective in controlling shallow deformation. Although the composite system composed of gabion cage, mesh and steel frame advances the instability time, it achieves the optimum interception effect on coarse particles. In conclusion, this paper further reveals the multi-field coupling mechanism of seepage-stress-deformation of talus-bedrock slopes during rainfall infiltration. It provides theoretical reference and technical support for the research on multi-field coupled progressive failure mechanism, stability evaluation and protection design of highway talus slopes under rainfall conditions in plateau areas.

Truck platooning, as an intelligent connected driving technology, offers advantages including reduced aerodynamic drag and energy consumption, improved transportation efficiency, and decreased carbon emissions. It represents a key development direction for future freight transportation. To advance the application of this technology and comprehensively understand its critical challenges and future trends, this paper systematically reviews relevant research progress globally. Building on an overview of the current state of truck platooning, the analysis first addresses positive effects from road users' perspective. Platooning optimizes aerodynamic drag within fleets, significantly reducing vehicle energy consumption and carbon emissions. Subsequent examination from the infrastructure standpoint reveals negative impacts on road structures. The technology alters the spatiotemporal loading characteristics of traditional traffic loads, posing risks of exacerbated asphalt pavement fatigue cracking and rutting. It may also induce cracking or slab fracturing in cement concrete pavements, thereby increasing material consumption and carbon emissions for road maintenance. Meanwhile, the impacts of truck platooning on highway traffic are summarized from two perspectives: traffic efficiency and traffic safety. Finally, a multi-dimensional optimization framework is proposed from the “vehicle-road dual perspective+integrated traffic dimension.”The results show that achieving a comprehensive balance in environmental and economic benefits between vehicles and infrastructure-through dual perspectives of road infrastructure and its users-is identified as crucial for future development and widespread adoption of this technology.

The high coupling of multimodal transfer paths in integrated transport hubs leads to diverse and dynamic guidance needs among heterogeneous passengers. Existing systems rely on static signage, which cannot cover complex routes and time-varying states. Delayed information updates, weak content adaptability, and unreasonable node configuration limit efficient transfers. This study proposed a dynamic-static signage optimization method in terms of layout style, point configuration, and information content. At the static level, a global optimization model was developed based on questionnaire data. At the dynamic level, a demand breadth index and a comprehensive demand weight index were introduced to establish differentiated matching and updating mechanisms, thereby forming adaptive strategies. A virtual-real fusion experimental environment was built with Revit and Unity. Three scenarios were designed, including high-density, low-density, and holiday-night flows. Passenger wayfinding simulation experiments were conducted. Trajectory and behavioral data were collected and evaluated using statistical and fuzzy comprehensive evaluation methods. The results show that, across the three scenarios, the dynamic-static collaborative scheme improved the wayfinding success rate by an average of 44.5%, reduced hesitation stops by 2.57, and reduced wrong turns by 0.64. It significantly outperformed the static-only and current schemes. This study formed a dynamic-static signage optimization and utility evaluation framework driven by real passenger behavior and supported by a virtual-real fusion experimental environment, providing methodological support for the design and continuous optimization of guidance systems in complex transport hubs.

To achieve a stable and reliable operation of connected and automated vehicles (CAV) collaborative control systems and enhance multi-performance objectives under complex disturbance conditions, the paper proposed a data-model driven collaborative control of CAVs (DM-CCC) method oriented toward multi-performance objectives. Firstly, the paper analyzed the mode of multi-source information delays between vehicles in CAV collaborative control systems, and developed a delay-robust control model for collaborative control by designing the information synchronization and state predictor strategies. Based on the proposed control model, the paper conducted a stability analysis of the proposed control models using frequency domain methods and derived sufficient theoretical conditions of string stability. Moreover, the paper established the objective functions and constraints for enhancing string stability, and thus formulated the stability-oriented optimization model for CAV collaborative control. An improved Particle Swarm Optimization algorithm with the introduction of the nonlinear weight reduction was designed to efficiently find the optimal control parameters. Then, the acceleration trajectory generated under the optimal stability-oriented control was used as a reference. By constructing a reinforcement learning framework that incorporates acceleration differences into state and reward functions, the DM-CCC method was developed. Finally, numerical experiments were performed to verify the effectiveness of the proposed control model, solution algorithm, and the proposed DM-CCC method. The results reveal that the proposed CAV collaborative control method has a good robustness and stability under multi-source information delay conditions. The solution algorithm shows significant advantages in convergence accuracy and can quickly find the global optimal solution. In addition, by adopting the DM-CCC method, the traffic efficiency is improved by 0.4%, the emissions are reduced by 1.4%, and the comfort is improved by 22.8%, the proposed DM-CCC method can effectively suppresses disturbance propagation in CAV collaborative control systems under delayed information conditions and improve string stability of the systems.

To address the lack of systematic characterization methods and design indicator systems for environmental transition sections on expressways and freeways, a scientific safety-efficiency evaluation and control strategy was established. Using multi-source data collection methods, 19 typical environmental transition sections were selected from 5 provinces in China. A characterization system containing 15 core parameters across four dimensions-spatial transformation, information transformation, traffic flow transformation, and visual transformation-was constructed. A dual-indicator evaluation model centered on average minimum time-to-collision and travel time index was developed, with prediction models built using machine learning algorithms. Feature importance analysis and conditional effect analysis methods were employed to identify key design indicators and reveal their interaction mechanisms. Evaluation thresholds were determined through statistical quantile methods, interaction effect analysis, and engineering standards to establish classification criteria and hierarchical control strategies. Results show that safety and efficiency of environmental transition sections are primarily controlled by composite effects of traffic flow density and spatial geometry. The gradient boosting tree model achieved R²=0.921 for efficiency prediction, while the random forest model achieved R²=0.763 for safety prediction. Seven key design indicators, including average time headway, lane-changing convenience, and diverging taper length, exhibit significant interaction effects encompassing four typical patterns: spatial-density matching, lane-changing-headway coordination, ternary composite effect, and diverging-density matching. Three section types-tunnel-interchange, interchange-tunnel, and interchange-interchange-demonstrate differentiated characteristics in key controlling factors, necessitating stratified evaluation and differentiated control strategies. These findings provide theoretical foundation and technical support for parametric characterization, safety-efficiency evaluation, and control of key design indicators for environmental transition sections, offering significant value for improving design standards and operational safety of such special sections.

The thermal fade problem caused by long downhill continuous braking of heavy-duty trucks seriously affects the driving safety of vehicles, in order to investigate the mapping relationship between brake thermal fade and vehicle brake thermal fade during long downhill continuous braking of heavy-duty trucks, an experimental research method is proposed to study the evolution mechanism of vehicle brake thermal fade of heavy-duty trucks. Firstly, according to the “Commercial Vehicle Service Brake Technical Requirements and Bench Test Methods” (QC/T 239—2015), a bench test program for drum brake thermal fade is designed, and the brake efficiency factor-temperature characteristic curve is obtained; through the thermal-force coupling numerical simulation method of the drum brake, the distribution and change rule of the temperature field and stress field of the drum brake are obtained under the braking condition of long downhill braking. Secondly, the brake temperature-MFDD regression model based on Bayesian Ridge regression theory is established through the vehicle brake thermal recession road test method, which reveals the evolution process of brake and vehicle brake thermal recession. Finally, a graded evaluation method is proposed using the brake temperature-MFDD regression model and the brake drum stress field-temperature field characteristics, and the validity of the evaluation method is verified by the real-vehicle test on mountainous roads. The results show that the experimental research method proposed in the paper can effectively reveal the evolution process of brake thermal degradation of the whole vehicle of heavy-duty trucks, and the proposed grading evaluation method can accurately assess the safety of the whole vehicle's braking, which provides the basis for the research on the safety of the long downhill braking of heavy-duty trucks.

The intelligent vehicle utilizes the on-board perception system to acquire information about the road conditions ahead, which allows for feedforward control based on the road status, thereby enhancing the overall driving safety of the vehicle. When a potholed road covered by foreign objects occurs, the on-board perception system may fail to accurately identify the pothole, leading to a failure of the feedforward control, which may cause the wheels to become trapped in the pothole, resulting in a loss of stability. In such scenarios, feedback control strategies based on vehicle dynamic responses become the key to ensuring driving safety. Accordingly, a vehicle attitude control method based on centroid safety domain constraints for potholed roads covered by foreign objects was proposed and validated. First, a vehicle dynamics model incorporating a hydraulic active suspension system was established, and the model's accuracy was verified through real vehicle experiments. Second, a collaborative control process for the active suspension system when crossing potholes was designed, and the transformation mechanism of the centroid safety domain when the vehicle transitions from four-wheel support to three-wheel support was revealed. Third, based on a comprehensive consideration of the hydraulic active suspension system's response characteristics, the intervention timing for active suspension control was determined, and an active suspension feedback control-based vehicle attitude control strategy was designed. Finally, hydraulic active suspension displacement tracking bench tests and the vehicle attitude controller in-the-loop tests under the specified road conditions were conducted. The results show that the designed active suspension feedback controller can achieve rapid response of the hydraulic actuator and improve displacement tracking accuracy, thus meeting the response requirements of the active suspension for vehicle attitude control. The proposed vehicle attitude control strategy effectively prevents the vehicle from getting trapped in the pothole covered by foreign objects, ensuring the vehicle's driving safety.

To address the issues of imbalanced vibration-reduction and energy-regeneration performance, as well as frequent switching of controller parameters under varying-speed conditions in electromagnetic energy regeneration suspension systems. This study proposed an innovative dual-mode variable-parameter control strategy selectable by the driver. The strategy leveraged the independent configuration of the vibration-reduction and energy-regeneration coils in the four-phase winding electromagnetic linear energy-regenerative damper. Additionally, a dual-mode mass-control parameter mapping curve was also established. First, linear quadratic regulator (LQR), sliding mode control, and fuzzy control were integrated to develop an improved sliding mode-LQR controller, establishing a unified control framework for the system. Second, based on the modified controller, a hybrid genetic-sequential quadratic programming (SQP) optimization algorithm with lower-level regulation was employed.Under the constraint that dynamic tire deformation does not exceed 10% of that of the passive suspension, a hybrid vibration-reduction mode-primarily targeting ride-comfort improvement and a hybrid energy-regeneration mode-primarily targeting energy recovery were designed. To address the difficulty of fixed-parameter controllers in adapting to complex suspension operating conditions, a variable-parameter control strategy was introduced. The dual-mode mass-control parameter mapping curve was established using an interval speed optimal control parameter solver and an arithmetically increasing sprung mass approach. The curve provides the optimal solutions within the speed range for different sprung masses, thereby avoiding frequent parameter adjustments under varying-speed conditions and ensuring the controller's adaptability to mass variations. Finally, simulation and hardware-in-the-loop experiments were conducted to verify the effectiveness of the improved sliding mode-LQR dual-mode variable-parameter controller under varying operating conditions. Compared with the traditional fixed-parameter LQR controller, under varying operating conditions, the hybrid vibration-reduction mode improved ride comfort by at least 13.10%, and the hybrid energy-regeneration mode improved energy-regeneration performance by at least 19.20%. The dual-mode variable-parameter control strategy can effectively improve the vibration suppression and energy regeneration performance of the electromagnetic energy regeneration suspension system.

The corner module architecture electric vehicle, owing to its superior maneuverability, trafficability, and handling stability, serves as an ideal platform for investigating optimal vehicle dynamic performance. Aiming at the failure scenario where a single wheel steering/driving system was locked or a tire was damaged, this paper designed a chassis cooperative fault-tolerant control strategy based on three-wheel driving. By lifting the faulty wheel and coordinating the remaining three wheels to support and propel the vehicle, stable continued driving of the entire vehicle was achieved. First, the dynamic stability mechanism of three-wheel driving was analyzed and demonstrated, with the center of gravity stability region under suspension travel constraints clearly defined. Next, by integrating path information and vehicle parameters, the maximum allowable longitudinal speed was derived under the constraints of the center of gravity stability region boundary while considering the coupling effects of longitudinal and lateral acceleration. Subsequently, a fault-tolerant control method is developed, featuring the active suspension controller as the core, complemented by the integrated slip ratio controller and path tracking controller. Within this framework, the sliding mode control-based active suspension controller ensured high safety during three-wheel driving, the predictive sliding mode control-based slip ratio controller suppressed wheel slip induced by active suspension actions, and the model predictive control-based path following controller guaranteed path tracking accuracy. Finally, the effectiveness of the proposed chassis coordinated fault-tolerant control strategy was validated through co-simulation using MATLAB/Simulink and CarSim. Results show that the derived theoretical safe speed is 32.4 km·h^-1, and the designed controller achieves stable path tracking at 29.8 km·h^-1 with the maximum wheel slip ratio constrained below 0.3 and path tracking error within 68.6 mm. However, when the speed increases to 36.9 km·h^-1, the vehicle becomes unstable. This study provides theoretical foundation and technical reference for fault-tolerant control of corner module architecture electric vehicles experiencing single wheel steering/driving system lock or tire damage.

In response to the inadequacies of existing vehicle trajectory planning methods concerning environmental perception depth and risk foresight in dense urban interactive scenarios, a hierarchical trajectory planning framework based on risk perception and a spatial-temporal graph network (STGN) is proposed. Initially, multi-obstacle trajectory prediction model based on STGN is constructed to precisely capture the dynamic interaction features among traffic participants. This model transforms the predicted two-dimensional absolute coordinates of trajectories into coordinate increments for each obstacle relative to the current position, significantly improving the model's generalization ability and prediction accuracy. Subsequently, a dual-layer “rule-guided, optimization-solving” planning architecture is designed. The upper-level, rule-based decision module generates macroscopic detour strategies. The lower level then translates this decision into a soft constraint for sequential quadratic programming (SQP), which is co-optimized with cost functions such as path curvature and lateral displacement from a reference line to generate a kinematically feasible central path. Notably, this research introduces the concept of a proactive dynamic risk zone, created by the envelope expansion of a baseline elliptical risk zone along the planned path. This approach integrates the ego-vehicle's behavioral intentions into the risk assessment system. Based on the intelligent driver model (IDM), longitudinal, lateral and rearward comprehensive risk indicators are incorporated into the acceleration calculation to generate a risk-aware speed profile, enabling information feedback and hierarchical coordination between path and speed at the risk-perception level. Simulation results demonstrate that the proposed method generates planned trajectories within 34.7 ms in dense interactive scenarios. Compared with conventional static risk area, the safety score is increased by 13.3%;compared with the best-performing baseline method, the destination arrival rate is increased by 20.6%. Overall, the proposed method balances trajectory safety, task completion rate and real-time performance, proving an effective approach for safe and efficient trajectory planning of autonomous vehicles in dense interaction scenarios.