Existing bridge bolt disease detection models based on deep learning suffer from high computational complexity and slow inference speed, making it difficult to meet the real-time monitoring needs of edge devices. To address this issue, a lightweight bridge bolt defect detection model based on an improved YOLO11n was proposed, and its improvement mainly consisted of frequency-domain feature enhancement module (FFEM) and cross-scale feature fusion network (CFFN). FEEM combined wavelet convolution and attention mechanism, and the model could extract the detailed features of bolts more accurately, enhance the perception ability of global information, and improve the identification accuracy of bolt diseases. CFFN optimized the feature fusion process at different scales, improving inference speed. Moreover, a lightweight downsampling module ADown was introduced in the backbone network to reduce the number of model parameters while retaining more useful feature information. Shape IoU considering the shape and scale of the target box was used as the bounding box loss function to improve the regression accuracy of bolt bounding boxes. The model was trained and tested on a dataset of 5 050 self-made bolt images covering three typical types of defective bolts such as rust, looseness and detachment, as well as normal bolts. Experimental results show that, compared with YOLO11n, the improved model has reduced the number of parameters and FLOPs by 47% and 34%, respectively. The accuracy and mAP50 have reached 94% and 91.7%, respectively, surpassing YOLO11n by 1 % and 0.8 %. The proposed model not only achieves high precision, but also has excellent lightweight characteristics, making it suitable for deployment on edge devices such as smartphones.

To achieve active reinforcement of fatigue cracks in the U-rib webs of orthotropic steel decks (OSDs), an arrayed Fe-based shape memory alloy (Fe-SMA) strip reinforcement technique with active strengthening characteristics was proposed, which was based on the arrayed steel strip reinforcement technology. This technique realized active strengthening by introducing prestress into damaged regions. To verify its effectiveness, a finite element (FE) model with damaged steel plate reinforcement was established based on ABAQUS. Based on the proposed model, the influence pattern of key parameters such as steel strip thickness, arrangement spacing, Fe-SMA recovery stress, spacing between bolt holes and cracks, as well as symmetrical arrangement of Fe-SMA steel strips on the reinforcement performance of this technology was evaluated. The research results show the arrayed Fe-SMA strip reinforcement technique significantly reduces the stress intensity factor (SIF) at the crack tip by applying prestress to the damaged steel plate, with reduction ranging from 16.0% to 91.5% compared with the arrayed steel strip reinforcement technique, thereby effectively alleviating stress concentration at the crack tip. Steel strip thickness, arrangement spacing and Fe-SMA recovery stress affect the strengthening performance by influencing the prestress level of the reinforcement system. In contrast, the spacing between bolt holes and cracks as well as the symmetrically arranged Fe-SMA strips have a significant working condition dependence on the reinforcement performance. Under tensile and bending conditions, reducing the spacing between bolt holes and cracks will cause a decrease in the reinforcement performance of this technology, while the opposite is true under shear conditions. The symmetrical arrangement of Fe-SMA strips is beneficial for tensile and shear resistance but performs poorly under bending conditions.

Bridge bearings exhibit significant temperature dependence in their mechanical behavior; however, existing structural health monitoring (SHM) technologies overlook the effects of temperature fluctuations on their mechanical performance. By conducting 6 independent temperature gradient tests (-20~30 ℃, ΔT=10 ℃), the base strain of the plate rubber bearing was accurately measured, and a monitoring method for rubber bearings based on temperature-deformation coupling perception using button sensor was proposed. The research indicates that under a compressive stress condition of 10 MPa, decreasing temperature alters the mechanical properties of rubber bearings, such as the compressive elastic modulus. The vertical compression decreases by 5.64 % per 10 ℃ temperature reduction, resulting in changes of the mechanical behavior of the rubber bearings. A real-time inverse model of temperature-strain-stress was developed based on the mechanical behavior of the bearing. By reading sensor data, the bearing load can be calculated with an error of less than 5%. The proposed technology provides a non-destructive, high-precision online monitoring solution for the monitoring of bearing, particularly suitable for swing bridges and other structures with bearing monitoring requirements in regions experiencing significant seasonal temperature fluctuations.

In response to the problem of insufficient feature learning and generalization performance of deep learning models as well as difficulty in meeting the practical needs of engineering, caused by severe road surface damage and a lack of training samples, the improved residual and multi-attention deep convolutional generative adversarial network (RM-DCGAN) was proposed in the research context of road surface diseases in Nigeria (with scarce samples and high degree of damage), to generate high-quality and diverse road surface disease images to expand the training dataset. The proposed model was based on the generator and discriminator foundation structure of DCGAN, introducing residual modules to alleviate the problem of gradient vanishing in deep networks, and integrating self-attention mechanism and channel attention mechanism to construct a hybrid attention module to enhance disease feature capture. Meanwhile, the Wasserstein distance with gradient penalty term served as the loss function was employed to effectively solve the problems of unstable loss function and poor generated image quality in traditional DCGAN. Experimental verification based on the self-constructed Nigerian pavement dataset demonstrates that the IS and FID values of the images generated by RM-DCGAN are 5.44 and 183.68, respectively, which are respectively 110.9% higher and 13.1% lower than those of DCGAN. The augmented dataset of the proposed model was used to train the YOLOv5 detection model, and a mean average precision (mAP) for typical road surface diseases such as cracks and potholes reaches 92.31%, which is 6.68% higher than traditional data augmentation methods. Research shows that the RM-DCGAN data augmentation method can effectively solve the problem of insufficient training data in small sample road surface disease detection, providing reliable technical support for road surface disease detection in small sample scenarios.

To compare the effects of two RAP pretreatment techniques, namely fine separation and conventional crushing, on the performance of recycled ultra-thin overlay mixtures, the road performance of these mixtures under these two pretreatment techniques was first evaluated. Furthermore, based on SEM and molecular simulation, interface models of new-old asphalt and recycled asphalt-aggregate in the mixture were established, and the interface performance was comparatively evaluated. The research results show that compared with the conventional crushing technique, the ultra-thin overlay mixtures prepared by fine separation technique exhibit better moisture stability, low-temperature crack resistance (at 50% RAP content, the residual stability and flexural tensile strain increase by 5.7% and 5.3%, respectively) and interface performance (the bonding performance of new-old asphalt interface and asphalt-aggregate interface increases by 21.0% and 29.3%, respectively), but slightly worse high-temperature performance (at 50% RAP content, the dynamic stability decreases by 5.5%). Considering meeting the road performance standards, the RAP content of recycled ultra-thin overlay mixtures under conventional crushing technique is 30%, while that under fine separation technique can reach 50%.

Water conveyance pipelines are a critical component of urban water supply systems, and their operational safety is directly related to the guarantee of water resources and the normal functioning of society, therefore, it is of great significance to carry out efficient and reliable leakage monitoring. Based on fiber Bragg grating (FBG) sensing technology, a large-scale experimental platform of water conveyance pipelines leakage was constructed by deploying FBG strain sensors on the outer surface of the pipeline, to monitor the hoop strain responses induced by different leakage conditions. Experimental results show that the hoop strain caused by leakage exhibits a clear spatially asymmetric attenuation characteristics, and the response amplitude of the upstream sensor is generally greater than that of the downstream sensor. As the leakage aperture increases and water pressure rises, the mutual coupling of stress concentration and local stiffness weakening effects significantly enhances the hoop strain response of pipeline structures and its fluctuation characteristics. It is verified that FBG sensors have relatively high sensitivity and response efficiency for monitoring the leakage of water conveyance pipelines, providing experimental basis and technical support for building an intelligent leakage early warning system based on FBG in water supply networks.

Underwater riprap projects serve as a key measure to ensure the stable operation of waterway regulation projects, however, current underwater riprap projects face challenges such as difficulty in precise positioning and real-time monitoring. To address these challenges, the real-time monitoring technology for underwater riprap based on multibeam bathymetry system was investigated. A “wet-end/dry-end” detection equipment architecture was established and a real-time monitoring system for underwater riprap was developed, which achieved positioning accuracy of ±0.1 m and data processing latency within 200 ms, enabling real-time, dynamic, and point-specific data collection and visualization for underwater riprap projects. Through multi-level trials at terminals and navigation channels of Yangtze River, the high-precision monitoring and 3D real-time visualization capabilities of the proposed system were verified. The research findings significantly enhance the construction quality and efficiency of underwater riprap, providing reliable technical support for navigation channel improvement projects.

In response to the green transition of Yangtze River shipping, and to address decision-making challenges in the planning and layout of green energy replenishment facilities, a comprehensive evaluation system that integrated multi-source information and combined weighting was constructed. First, an improved CRITIC-entropy weight method and fuzzy C-means clustering were employed to screen and construct an evaluation system consisting of 20 core indicators, including policy support, economic feasibility, technological infrastructure and operational support capabilities, from the initial selection indicators. A game theory combination weighting model was further applied to integrating AHP subjective weights and entropy objective weights to form a balanced comprehensive weight. On this basis, an improved TOPSIS-grey relational projection method was proposed, which comprehensively considered both the positional proximity and distributional similarity of alternatives and ideal solutions, thereby improving evaluation reliability. An empirical study of 12 major ports in the Yangtze River Basin shows that the proposed model can effectively classify ports, with Wuhan Port and Nanjing Port identified as priority development ports. Most ports are in the primary coordination stage, with the weak coordination between the economic and technological subsystems. Obstacle degree analysis identifies differentiated barrier factors, such as the investment payback period in Yueyang Port and the energy conversion efficiency in Wuhan Port. This study provides a reliable decision-making tool for the layout of green energy replenishment facilities along the Yangtze River and offers referential insights for the planning of similar infrastructures.

Aiming at the uneven distribution of spatiotemporal demand for urban taxis and the strong temporal variability in multiple hotspot areas, a taxi demand prediction method based on improved deep extreme learning machine (DELM) in multiple hotspot areas was proposed. Firstly, the variational mode decomposition method was applied to decompose the original demand series into multiple sets of modal components with clear physical meanings, effectively separating demand fluctuation features at different time scales. Then, based on the construction of the deep extreme learning machine model, a population-based metaheuristic optimization algorithm was employed to optimize the weights and bias parameters of the DELM. Finally, the effectiveness of the proposed method was validated through taxi demand data from six hotspots in Manhattan, New York. The results demonstrate that the improved DELM achieves accurate taxi demand predictions across different hotspot areas, with statistically significant reductions in EMA, ERMS and EMAP of order volume by 2.43~8.54 orders per hour, 3.01~11.77 orders per hour, and 0.54%~2.91%, respectively, compared to other baseline models and single models.

To address the optimization of emergency medical facilities in expressway service areas, a two-stage stochastic programming location model that simultaneously considered stochastic demand, stochastic travel time, road conditions and casualty triage characteristics on expressways was constructed. To validate the performance of the proposed model, Monte Carlo simulation was first employed to generate scenarios incorporating multiple random factors, and then the K-means clustering method was used to select representative scenario clusters as the final scenario set. Secondly, a hybrid-greedy simulated annealing-adaptive large neighborhood search (H-GSA-ALNS) algorithm, integrating greedy strategies, simulated annealing and adaptive large neighborhood search, was proposed and solved. In numerical analysis, the effectiveness of the proposed H-GSA-ALNS algorithm was demonstrated through comparison, and the site selection and ambulance vehicle allocation under different demand levels were analyzed. The research results indicate that the proposed H-GSA-ALNS algorithm outperforms single simulated annealing and conventional ALNS algorithm in both solution quality and convergence stability. Sensitivity analysis further reveals the impacts of different demand levels, travel-time uncertainty and maximum service distance on site selection results and ambulance allocation strategies. Under various stochastic scenarios, the proposed model achieves high demand coverage and a reasonable response-time distribution, exhibiting overall good robustness and adaptability. The proposed model and solution approach can effectively support decision-making for site selection and ambulance allocation of expressway emergency medical service system, providing quantitative decision-making basis and management insights for improving expressway emergency medical service capabilities.

Vehicle trajectory prediction is one of the key technologies in intelligent transportation systems and autonomous driving. To address the issues of insufficient attention allocation precision and inadequate spatiotemporal feature fusion in existing vehicle trajectory prediction models under complex vehicle interaction scenarios, a multimodal trajectory prediction model incorporating dynamic perception masking and cross-attention was proposed. Firstly, a spatial interaction feature extraction module based on graph attention network (GAT) was constructed, and a novel dynamic perception masking mechanism was introduced, which dynamically adjusted attention weights according to the vehicle’s own speed and the relative distance to surrounding objects, thereby enabling adaptive focusing on critical interactive objects. Secondly, to facilitate deep coupling of spatiotemporal features, a spatiotemporal feature fusion module based on bidirectional cross-attention was designed, which effectively captured the deep nonlinear dependencies between the two through bidirectional querying and enhancement of temporal and spatial feature streams. Finally, an intention-aware trajectory decoder utilized a gated recurrent unit (GRU) to generate the parameters of bivariate Gaussian distribution describing the trajectories, thereby achieving multimodal trajectory prediction. Experimental results on the public NGSIM and HighD datasets demonstrate that, compared to baseline models, the proposed model exhibits superior performance across all evaluation metrics. The root mean square error (ERMS) in the 5s prediction time domain is reduced by 11.7% and 15.8% on the NGSIM and HighD datasets, respectively, confirming the effectiveness and robustness of the proposed model.

Agricultural product transportation has significant timeliness characteristics. The ship lock scheduling department needs to grasp the future scale of agricultural product freight in advance in order to formulate a reasonable gate clearance dispatch plan. Therefore, accurate prediction of agricultural product freight volume is of great significance for improving gate clearance efficiency. The agricultural freight volume in the Ganjiang River exhibits characteristics such as intense fluctuations and complex patterns, and current prediction methods are difficult to capture its deep nonlinearity and complex temporal patterns. To address this issue, a hybrid model integrating attention mechanism, sparrow search algorithm (SSA), convolutional neural network (CNN), and long short-term memory network (LSTM), namely SSA-CNN-Attention-LSTM, was proposed. First, CNN was employed to capture local spatial features from the data, while LSTM was combined to extract temporal dependency relationships. An attention mechanism was further introduced to assign weights to the hidden states of LSTM, thereby enhancing the proposed model’s ability to capture key information in complex fluctuation scenarios. In addition, SSA was adopted to adaptively optimize the hyperparameters of the proposed model, improving its adaptability and convergence efficiency in complex forecasting scenarios. In the performance evaluation, multiple representative methods based on CNN, RNN, MLP, and Transformer architectures were compared. The results show that the proposed model outperforms other models in three indicators, that is EMA, EMAP and ERMS, which are 471.26, 0.114, 30, and 628.81, respectively. The prediction results indicate that the proposed hybrid model can effectively improve the prediction accuracy of waterway agricultural product freight volume, which is beneficial for the ship lock to make advance scheduling plans and provide reliable data support for port and navigation management and development decisions in the Ganjiang River Basin.

As the continuous growth of aviation carbon emissions, guiding passengers to participate in aviation carbon offsetting has become an important pathway for the aviation industry to achieve net-zero emissions. Based on the theory of planned behavior, a theoretical model of passengers’ willingness to participate in aviation carbon offsetting incorporating fairness perception was constructed. Then, through a questionnaire survey of 502 air passengers and combining partial least squares structural equation model (PLS-SEM) and necessary condition analysis (NCA) method, the formation mechanism and its necessary conditions of carbon offset participation willingness were explored. The research results show that: ①except for the insignificant effects of benefit visibility on fairness perception and information transparency on participation willingness, all other hypotheses are supported; ②fairness perception plays a mediating role among certification credibility, information transparency, environmental efficacy, price acceptance, and participation willingness; ③ information transparency is a necessary condition for the medium effect of fairness perception. The research conclusions contribute to passengers’ deeper understanding of the factors influencing their willingness to participate in carbon offsetting and offer theoretical reference and practical basis for the design of carbon offset projects and the development of participation promotion strategies.

To address the reduction in the driving range of pure electric vehicles caused by battery performance degradation and surging energy consumption for cabin heating under low-temperature conditions, an integrated thermal management heating system based on the diesel liquid heater was proposed. Using the diesel liquid heater as an independent heat source, the passenger cabin and battery circuits were coupled via an electronic three-way valve, enabling centralized supply and dynamic distribution of thermal energy. For the dual-heating mode, a multi-level fuzzy control strategy framework was designed for the electronic three-way valve and the battery water pump. The valve opening and battery water pump speed were dynamically adjusted according to the heating demands of the passenger cabin and battery circuit, as well as changes in battery temperature. The system model was built and co-simulated by use of AMEsim and MATLAB/Simulink. The results demonstrate that, compared with the traditional independent circuit system, the proposed integrated system saves approximately 8 % of battery power and enhances the driving range by about 27 % in a -15 ℃ environment. Furthermore, the multi-level fuzzy control strategy shows superior dynamic response and energy efficiency performance in the coordinated control of the electronic three-way valve and the battery water pump, reducing energy consumption of battery water pumps by about 52%.

Due to the lack of degree-based global scaling, the current trajectory prediction models partially loss the effectiveness of spatial feature representation, and the direct concatenation and decoding of spatial-temporal features results in insufficient attention to primary features. Therefore, a vehicle trajectory prediction model based on PNA spatial encoding and SE-MoE fusion of spatiotemporal features was proposed. Temporal features were extracted by a residual-based gated recurrent unit, then spatial features were extracted by spatial encoder, and finally an improved hybrid expert strategy was adopted to fuse and decode spatiotemporal features, realizing the vehicle trajectory prediction. Specifically, based on the extraction of vehicle trajectory spatial features based on PNAConv and GATv2Conv, the degree-based global scaling was employed to enhance the consistency of spatial feature representation, while the attention of key information was enhanced by dynamic attention mechanism. Based on SE-MoE for spatiotemporal interaction feature fusion, the input feature space was divided into three regions with different expression patterns, and the main feature extraction was carried out jointly through routing and shared expert networks. In contrast to traditional concatenation decoding, the SE-MoE-based fusion enhances the expression of trajectory motion trends. Tests on the NGSIM dataset show that compared with models such as Transformer-GAT, S-LSTM and ST-LSTM, the proposed model performs the best in long-term prediction tasks, reducing 5-second root mean square errors by 23.8%, 62.8% and 56.5% respectively, and performs well in short-term prediction tasks. Tests on the Bahe West section of the Xi’an Ring Expressway demonstrate that the proposed model has good generalization capability in real-world scenarios. The 1-second and 5-second root mean square errors for straight-driving samples are 0.78 m and 6.92 m, respectively, while those for lane-changing samples are 0.85 m and 3.29 m, respectively.