[Objective] Fixed-speed green wave strategies on arterial roads often fail to match real-time traffic demand, which reduces the traffic efficiency. This study proposes a speed-guidance and trajectory-optimization method for arterial traffic in connected-vehicle environments. [Method] First, a speed-guidance model was developed according to vehicle speed, signal timing, and preceding-vehicle behavior. The scenario-specific guidance enabled vehicles to pass multiple intersections with few or no stops. Subsequently, vehicle trajectories were optimized by using a signal-constrained optimal-control method to avoid abrupt acceleration and deceleration. Finally, a simulation platform was built with SUMO, Traci, and Python. The proposed method was tested on three consecutive signalized intersections on Yejin Avenue in Wuhan. The performance of speed-guidance method was compared with those using Webster signal timing and MAXBAND coordination methods. [Result] The speed-guidance method reduces average travel time, number of stops, and stop duration. Its effectiveness increases with higher connected-vehicle penetration. Trajectory optimization further reduces acceleration disturbances by 58.70%, 31.68%, and 76.72% in acceleration-through, deceleration-through, and deceleration-to-stop scenarios respectively, compared with that using the original speed-guidance method and MAXBAND method. [Conclusion] The proposed method improves arterial traffic efficiency and driving comfort. It demonstrates strong potential for connected and intelligent transportation systems.
[Objective] The closed-set classification adaptability and prediction accuracy are both insufficient in compact and complex parking scenarios. This study aims to adapt to the diversity and uncertainty of driving intentions in unstructured parking scenarios, and achieve accurate prediction on target vehicle's future driving intentions, thereby providing critical technical support for parking decision-making of autonomous vehicles. [Method] This study proposes a hybrid CNN-GRU-based vehicle driving intention prediction method optimized with whale optimization algorithm (WOA) for parking scenarios. First, the historical trajectory and map information of target vehicles are encoded into grid-based bird's-eye view images. The spatial features are extracted via convolutional neural networks, which are then fused with interaction features, e.g., relative time, azimuth, and path features based on A* algorithm. Second, these fused features are input into the gated recurrent unit network to further capture temporal dependencies. The fully connected layer outputs the intended parking spaces and drivable road segments that the target vehicles may enter in the future. Furthermore, rules for intention category classification and probability assignment within target vehicles' decision scope are established to derive all intention points and their corresponding probability distributions. Finally, a real-world parking lot dataset is used to train and validate the model. WOA is adopted to globally optimize the key hyperparameters of the proposed hybrid model during the training process, improving model's prediction accuracy and generalization ability. [Result] The prediction accuracy Top-1 of the proposed method reaches 79.43%, and Top-3 reaches 100%. The proposed method exhibits superior performance across all evaluation indices compared with baseline models. Specifically, the proposed method improves Top-1 prediction accuracy by 7.27%, and reduces the model loss value by 46.13% compared with the optimal baseline model. [Conclusion] The proposed method can effectively improve the recognition accuracy of driving intentions in parking scenarios, as well as provide a reference for trajectory planning and safety decision-making of autonomous vehicles in parking scenarios.
[Objective] Long single-lane two-way roads in rural tourist areas suffer severe congestion during peak periods. Head-on conflicts further threaten safety. This study proposes a real-time signal control method that balances efficiency and safety. The method pursues two goals in saturated traffic, i.e., reducing vehicle delays, and zero violations of opposing-conflict and passing-bay capacity constraints. [Method] A distributed safe multi-agent deep reinforcement learning framework was developed. Each agent generated signal control decisions according to local traffic states. A dynamic action-masking mechanism was introduced as well to guarantee engineering feasibility and phase-switching legality. Safety constraints were handled via Lagrangian multiplier method. Both the passing-bay capacity constraint and the opposing-conflict constraint were embedded directly into the reward function. That ensured safety requirements systematically satisfied during policy training. Comparative experiments were conducted on a typical segment of Yiwu-Longquan-Qingyuan expressway. SUMO platform was used for simulation. [Result] The proposed method achieves a global average delay of 246.95 s, representing 78.5% reduction over fixed-cycle control. It also outperforms the best safe multi-agent baseline by 15.6%. Zero conflict violations and zero capacity overflows are observed throughout the simulation. The decision response time is 13.35 s, which is within 1% of the time required by conventional mathematical programming. The method therefore combines safety with real-time performance. [Conclusion] The findings provide a methodological framework and technical support, enabling smart mobility deployment in single-lane two-way rural scenarios. The proposed method also supports transportation-tourism integration. It advances the intelligent management of rural road traffic.
[Objective] This study proposes an AI-based daily traffic demand prediction model to accurately predict the daily traffic demand at the entrance of construction site for building projects; as well as solve the traditional prediction fails to comprehensively consider process differences, road network congestion and dynamic disturbances. [Method] First, the average daily traffic demand was obtained based on the construction schedule. Second, various processes were classified based on deterministic factors at planning level, e.g., relation between construction duration and progress, material characteristics, mutual influence of adjacent processes, personnel allocation and facility configuration. Third, the fitting function model was used to analyze the correction coefficient of daily average traffic demand for each type of process, considering the above-mentioned deterministic factors and uncertain factors, e.g., weather and management efficiency. Finally, Markov chain was used to predict the congestion degree of traffic network around the construction site in prediction period; and the system planning transfer caused by congestion was analyzed based on this basis. [Result] The classification accuracy with the proposed model is higher than that with SVM; and the mean square error is significantly reduced after correction. The predicted value considering road network congestion has a smaller error with the actual traffic volume; and the prediction accuracy meets the requirements of engineering application. [Conclusion] The proposed model can accurately predict the daily traffic demand for construction units in the prediction period. It provides a reliable basis for transportation planning decision-making and saves the cost of transportation.
[Objective] For calibration of road traffic LiDAR, the absence of traceability chain for small-angle reflectance in the corresponding reflection conditions makes it impossible to directly calibrate LiDAR using small-angle reflectance.This study establishes an absolute measurement device with small-angle reflectance. A calibration method is proposed based on an optimal reflectance estimation algorithm. A reflectance traceability chain is formed in small-angle geometry conditions, solving the difficulty of direct calibration traceability using small-angle reflectance for LiDAR. [Method] First, the spatial occlusion between light source and detector was solved by using precise optical path design in small-angle conditions. This enabled small-angle reflectance measurement with both incident and observation angles smaller than 0.1°. Second, LiDAR calibration method was proposed based on an optimal reflectance estimation algorithm. This method was based on the calibration value of reflective standard plate in small-angle conditions. Finally, LiDAR was used to measure the pedestrian target. The differences between target's normal reflectance and small-angle reflectance were obtained. [Result] The relative deviation of the measured reflectance of standard plate from the theoretically calculated value based on the national reference calibration value is less than 1%. It verifies the accuracy and reliability of the established device. The small-angle reflectance of AEB adult and child pedestrian targets is 0.711 and 0.697 respectively using the proposed LiDAR calibration method. The obtained small-angle reflectance of pedestrian targets has a maximum deviation of 31.1%, compared with the reflectance values specified in ISO 19206-2. This discrepancy may cause intelligent vehicles to misidentify pedestrians as concrete pillars or similar objects. [Conclusion] The obtained data and analysis are positive. They improve the accuracy and reliability of quantitative environmental perception testing for intelligent vehicles.
[Objective] This study aims to address the limitations (e.g., low efficiency, and insufficient generalization across diverse environmental conditions) of existing detection methods for traffic marking retroreflection coefficient. An automatic assessment method for traffic marking quality is proposed, which is rapid, cost-effective, and applicable to multiple scenarios. It proposes the use of mobile lidar to replace conventional handheld retroreflectometer, achieving full-coverage and non-intrusive inspection. [Method] Traffic marking point clouds were acquired via mobile lidar, and processed through ground segmentation, maximum entropy thresholding, and region growing. The laser incidence angle and sensor distance were calculated, forming a multi-dimensional feature vector. The domain adaptation model was then constructed with semi-supervised learning, based on substructural optimal transport. It achieved grading of retroreflection coefficients by integrating limited labeled retroreflectometer measurements with large-scale unlabeled lidar data. [Result] The proposed method effectively identified white and yellow traffic markings at varying degrees of wear, demonstrating significantly improved generalization performance over traditional supervised learning methods across day and night conditions. The detection accuracy improved by an average of 3.17% for both marking colors in combined diurnal conditions. The detection accuracy reached the maximum improvement of 21.91% in specific scenarios, e.g., clear daytime environments. [Conclusion] The proposed lidar-based semi-supervised learning method effectively resolves the challenge of cross-scene generalization with scarce labeled data, providing a feasible technical pathway for surveys of large-scale traffic marking wear. The method offers both safety and economic advantages, supporting informed road maintenance decisions, extended pavement service life, and enhanced driving safety.
[Objective] This study investigates the underlying logics of intermodal transfer mode choice, as well as the factors influencing intermodal transfer mode choice at comprehensive passenger transport hubs. [Method] A set of influencing factors for transfer mode choice was established with individual socio-economic attributes, passenger hub travel attributes, and passenger travel scenario attributes collected through questionnaire surveys. A random parameters Logit model was applied to investigate the influence mechanism of passenger transfer mode choice from two analytical layers, i.e., independent analysis on transfer mode and aggregated analysis on similar mode. Logit model was established for subway, bus, taxi, ride-hailing, and private car at the layer of independent analysis. Logit model was established for scheduled-route modes and on-demand modes at the layer of aggregated analysis. [Result] The model results indicate significant influencing factors and heterogeneity in transfer mode choice. It identifies four travel scenarios that significantly affect transfer mode choice at the layer of independent analysis, i.e., whether to have an emergency, whether to carry bulky luggage, whether to travel with the elderly and children, and whether to have physical discomfort. It identifies two additional common influencing factors (private car ownership and residence place) at the layer of aggregated analysis, serving as a supplement to the layer of independent analysis. It is also found that the aggregated similar mode alters the randomness of parameters, thereby eliminating the heterogeneity effects on mode choice. The aggregated analysis on similar mode leads to the loss of certain factors; specifically, whether to carry bulky luggage is a significant variable in the independent analysis, but becomes insignificant in the aggregated analysis. [Conclusion] The findings contribute to the precise scheduling of capacity at comprehensive passenger transport hubs, enabling rapid evacuation of passenger flows.
[Objective] This study proposed a data-driven method for the layout of driver stations, addressing four difficulties, i.e., parking, dining, toileting, and resting commonly faced by ride-hailing drivers. It aimed to improve drivers' occupational well-being and operational safety, as well as to provide scientific basis for planning urban ride-hailing supporting facilities. [Method] This study took ride-hailing order endpoints as proxy features for drivers' potential rest demands; and based on approximately 3.5 million order records from a provincial capital city in July 2024. First, a demand identification model with multi-factor weighting was constructed to quantify the demand intensity at each point, integrating trip duration, distance, end time period, and day type. Second, the spatiotemporal clustering method was applied to identify the demand hotspot areas. Four indicators from complex network analysis were introduced, i.e., centrality of degree, centrality of closeness, centrality of betweenness, and intensity of node demand. A comprehensive evaluation model was established based on the coefficient of variation method and TOPSIS, identifying key demand nodes in the network. Finally, a hierarchical layout scheme for driver stations was proposed based on ranking of demand centrality and principle of maximum coverage. The layout effectiveness was verified through service coverage rates and comparison with alternative schemes. [Result] A total of 42 demand hotspot nodes was identified. The key nodes with the highest demand centrality were identified through comprehensive evaluation. Based on this, 8 driver stations were deployed. They achieved an overall coverage rate of 81.5% for order endpoints, and 76.2% for order start points, representing improvements of 6.2% and 6.4% respectively compared with the demand-intensity-only ranking scheme, and 9.4% and 4.7% compared with the centrality-only ranking scheme. All driver stations maintained good service accessibility across different time periods. [Conclusion] The proposed driver station layout method can accurately identify demand hotspots. It covers major rest demands with fewer stations, demonstrating strong practicality and operability; as well as providing data support and decision-making reference for urban traffic management departments and ride-hailing platform enterprises in planning service facilities.
[Objective] Rutting is a major distress in asphalt pavement, which affects road safety and service quality. Current prediction methods have poor generalization ability, low prediction accuracy, and need too much parameters. A deep learning model, integrating CNN, BiLSTM and Attention mechanism, was proposed to predict the development trend of rut depth, based on measured data from a full-scale test track, i.e., RIOHTrack. [Method] The model used seven key factors as inputs, i.e., cumulative equivalent standard axle load repetitions, falling weight load, deflection basin area, texture depth, central deflection, surface roughness, and average temperature. Support vector regression, random forest, CNN and BiLSTM were selected as benchmark models for comparison. In addition, the ablation test was designed to analyze the influence of different feature combinations on prediction accuracy. [Result] CNN-BiLSTM-Attention model outperformed the benchmark models across all evaluation metrics, demonstrating higher prediction accuracy and better generalization capability. The ablation test result indicates that the highest prediction accuracy was achieved when all input features were retained. Removing texture depth resulted in the most significant decrease in prediction accuracy, followed by the removal of deflection basin area and surface roughness. In contrast, removing falling weight load, central deflection, and average temperature led to relatively small reductions in prediction accuracy. In addition, the prediction accuracy remained higher than that of feature combinations lacking key features, when only cumulative equivalent standard axle load repetitions and average temperature were retained as input features. [Conclusion] The proposed model extracted spatial features through CNN, captured temporal dependencies through BiLSTM, and optimized feature weights through Attention mechanism, thereby achieving accurate rutting prediction in multi-factor coupling conditions. The ablation test further identified the key input features.
[Objective] Loess is main material for subgrade filling in loess areas. Post-construction uneven settlement and collapsible deformation induced by its unique engineering properties have long been critical problems threatening the safety and stability of roads in such areas. It is urgent to systematically summarize the research findings concerning mechanical properties of loess and subgrade deformation control, supporting the design and construction practices of subgrade engineering in loess regions. [Method] This study presents a comprehensive state of the art review of relevant domestic and international studies from four perspectives of loess subgrade, i.e., compression deformation characteristics, creep properties and constitutive modelling, settlement mechanisms, and settlement calculation and prediction method ologies. The applicability and technical attributes of different research approaches are critically compared and evaluated, including laboratory model tests, in situ field tests, and numerical simulations. In addition, current remedial measures for subgrade distress and their documented field performance are systematically summarised. [Result] The deformation of loess is intimately governed by its moisture content, dry density, and stress path. Existing settlement computation methods and predictive models both have their own specific applicable conditions and limitations. [Conclusion] The ever increasing complexity of engineering demands continues to pose new practical challenges, although considerable efforts have been devoted to investigating the fundamental properties and constitutive relationships of loess in the context of infrastructure development. Future study should place particular emphasis on moisture fluctuations, cyclic loading, long term creep behaviour, and multi field coupling effects. The findings provide theoretical basis and technical reference for subgrade construction and safe operation in loess areas.
[Objective] This study investigates the dynamic response evolution of lateritic clay subgrade due to coupling effect of wetting and dynamic loading. It analyzes the performance deterioration of lateritic clay subgrade on expressways in humid and rainy areas. [Method] The test section of lateritic clay subgrade was on Hengyang-Yongzhou expressway in Hunan Province. The field driving tests were conducted with different axle loads, vehicle speeds, and moisture contents. The distribution and attenuation rules in different wetting and dynamic loading conditions were investigated on dynamic stress amplitude, dynamic acceleration amplitude, and dynamic displacement amplitude along depth and lateral direction of lateritic clay subgrade. [Result] The amplitudes of dynamic stress, dynamic acceleration, and dynamic displacement increase with the increase of axle load and vehicle speed. An increase in moisture content will lead to an increase from 40% to 70% of dynamic response in the same dynamic loading conditions. The dynamic stress amplitude, dynamic acceleration amplitude, and dynamic displacement amplitude are varying with the increase of depth and lateral distance of subgrade; it exhibits a nonlinear decay that decreases rapidly and then stabilizes. The increases in axle load and vehicle speed lead to an increase in the ratio of dynamic stress to static stress on the subgrade. When the vehicle speed is 60 km/h, the influence depth of vehicle under empty load is 1.6 m, and that of vehicle under full load is higher than 2.3 m. Based on the field test result, a prediction model of dynamic stress amplitude, dynamic acceleration amplitude, and dynamic displacement amplitude is established considering moisture content, axle load and vehicle speed. [Conclusion] The coupling effect of wetting and dynamic loading aggravates the dynamic response of subgrade, resulting in the increased depth of subgrade workspace and greater deformation of subgrade. The findings provide reference for the performance evaluation of durable lateritic clay subgrade in humid and rainy areas.
[Objective] The treatment effect evaluation of existing foundation is a key technical challenge for new and aged subgrade coordination in expressway expansion. This study quantitatively evaluates the long-term reinforcement effects with three treatment schemes, i.e., dynamic compaction, compacted gravel piles, and dry jet mixing piles. [Method] Seven typical cross-sections were selected due to the widely distributed liquefiable silty soil and soft soil along the expansion project of Xinyi-Yangzhou expressway. The insitu tests, combined with cone penetration test and high-frequency long-span cross-hole CT, were carried out. Comparative tests were conducted at seven typical sections to systematically evaluate the performance of different treatments. [Result] The cone tip resistance of shallow silt (less than 10 m) increased by up to 350% using dynamic compaction, markedly densifying the surface layer. Compacted gravel piles showed the best overall reinforcement with the cone tip resistance rising by over 400%, especially with 1.2 m pile spacing and under surcharge preloading. Dry jet mixing piles produced minor and sporadic increments in shallow silt, while indicating limited improvement in the underlying soft soil. [Conclusion] Reinforcement differences stem from mechanistic compatibility between methods and foundation conditions. Compacted gravel piles yield balanced improvement in liquefiable silt and soft clay by delivering drainage and densification. Dynamic compaction is effective in silt, but its energy attenuates in deep clay. Dry jet mixing has the minimal effect on inter-pile soil. This findings offer a reference for treatment selection for similar layered foundations.
[Objective] This study investigated the influence of aging on low-temperature cracking resistance of asphalt, as well as established the evaluation indexes for low-temperature rheological properties of asphalt. It provided a theoretical basis for the prevention and control of low-temperature cracking risks in asphalt pavement. [Method] First, RTFOT and pressure aging vessel test were conducted to simulate asphalt aging. Second, the low-temperature rheological properties of aged asphalt were measured by using the bending beam rheometer test. The creep curve was obtained by fitting the data to Burgers viscoelastic model. The aging sensitivity analysis was performed on the rheological parameters, e.g., relaxation time, low-temperature comprehensive compliance parameter, and low-temperature ratio of creep rate to bending creep stiffness. Finally, a low-temperature rheological evolution model based on comprehensive compliance parameter was developed. FTIR was employed to investigate the correlation between functional group variation and low-temperature rheological parameters. The reliability of the proposed model was further verified by using this technique. [Result] The low-temperature cracking resistance of asphalt is affected by both aging and ambient temperature. A certain ambient temperature threshold exists at which the influence of temperature on low-temperature performance surpasses that of aging. Below this threshold, temperature will replace aging as the main factor reducing low-temperature performance of asphalt. The compliance parameter exhibits higher sensitivity to aging than relaxation time and low-temperature ratio of creep rate to bending creep stiffness. This parameter can serve as an effective index for evaluating low-temperature cracking resistance of asphalt. The low-temperature rheological evolution model of asphalt based on compliance parameter exhibited correlation coefficients all above 0.85, with significance levels of all variables below 0.05. FTIR analysis indicates that the compliance parameter effectively reflects the variations in chemical composition of asphalt during aging. This parameter exhibits high correlations with the sulfoxide index, carbonyl index, and aliphatic functional group index. The compliance parameter of aged asphalt demonstrates a multivariate linear relation varying with the content of chemical functional groups. [Conclusion] The proposed model can be used to predict low-temperature crack resistance of asphalt during aging at various temperatures, providing technical support for the design and maintenance of asphalt pavements.
[Objective] This study aims to provide one-stop digital and intelligent design and analysis services for bridge structures on BIM platform. A technical workflow is proposed for concrete segmental beam bridges. The workflow combines BIM parametric modelling with the output of intelligent bridge calculation results. [Method] First, a programme for the parametric input and extraction of data for segmental concrete beam bridge was developed by using the general-purpose segmental beam bridge generic family and Revit APl, enabling the rapid digital updating and automatic identification of bridge information.An engineering parameter user interface was developed for BIM platform based on MVVM architecture within WPF framework. The interface allowed users to input the data required for bridge analysis without migrating the model. It also converted and exported standard data files automatically. A theory-driven intelligent calculation method was then proposed for beam bridges. A one-dimensional temporal convolutional network was used as the core deep learning model, which was jointly trained with a pre-trained boundary neural network to improve calculation efficiency. It output the internal force distribution of bridge sections, nodal reaction forces, and deformation results. The calculation model could be connected with BIM model by quickly parsing the data exported from the engineering parameter interface. Structural mechanics equations were also introduced as prior knowledge. This would help ensure the correctness of calculation results. [Result] Numerical tests showed good agreement with MIDAS Civil results. The maximum bending moment error was less than 3%. The maximum mid-span deflection error was less than 6%. The proposed method can meet the accuracy requirements of engineering applications. [Conclusion] The proposed BIM parametric modelling and intelligent calculation workflow integrates model updating, analysis parameter organization, and calculation result output. It provides technical support for the end-to-end digital design and analysis of segmental beam bridges.
[Objective] Bolts are critical connection components in steel bridges. Their defects, e.g., loosening, loss and corrosion, pose serious threats to structural safety. Traditional manual inspection is inefficient, while existing deep learning methods often suffer from insufficient transferability and limited detection performance in real bridge environments due to data scarcity, small and dense targets, and class imbalance. This study proposes a bridge bolt defect detection method integrating transfer learning with object detection. [Method] First, a weighted cosine similarity was employed to select source domain samples with features close to the target domain for pre-training, thereby alleviating data scarcity and reducing the risk of negative transfer. Subsequently, a lightweight IRMA module was further embedded into the backbone of YOLOv8s to enhance feature extraction for dense small targets. In addition, Varifocal Loss was introduced to balance positive and negative samples through a dynamic weighting mechanism, reducing the missed detection rate of defect categories. Finally, comparative experiments and ablation studies were conducted on the self-constructed bridge bolt defect dataset. The proposed method was evaluated against mainstream approaches, including Faster R-CNN, SSD, and YOLO variants. [Result] The proposed method achieved an accuracy of 89.9% and an mAP@0.5 of 91.1% on the dataset. [Conclusion] It effectively mitigates the issues of limited samples and class imbalance in real bridge scenarios, enhances the detection capability for dense small targets, and improves cross-domain generalization, providing a reliable technical solution for the visual inspection of bridge bolts.
[Objective] This study proposes optimized configurations and quantitative design parameters for mild steel dampers to determine reasonable longitudinal and transverse seismic constraint systems for asymmetric single-pylon cable-stayed bridges in high-intensity seismic zones. It address the deviation between mass center and stiffness center, as well as the concentration of seismic responses caused by structural asymmetry. [Method] An asymmetric single-pylon cable-stayed bridge in the high-intensity seismic zone was selected as study object. The nonlinear time-history analysis was adopted.The seismic performance of four systems was compared, including floating system, restraint system, viscous damping system, and mild steel damping system. Parameter sensitivity analysis and optimization design of dampers were then performed. The optimal specifications and arrangement of longitudinal and transverse mild steel dampers were determined. The seismic response reduction was verified. [Result] The mild steel damping system achieved the optimal seismic response reduction among longitudinal seismic systems. Its structural bending moment was reduced by approximately 70% compared with other systems; the displacement was reduced by 58.3% compared with floating system, and reduced by 28.3% compared with viscous damping system. The mild steel dampers, installed at both pylon-girder and pier-girder connections, yielded the best control effect among transverse seismic systems. The bending moment and displacement at key sections were reduced by up to 21.5% and 29.2% respectively. The optimal configurations were determined through parameter optimization. Four longitudinal mild steel dampers with capacity of 1 500 kN and two transverse mild steel dampers with capacity of 1 000 kN were installed between pylon and girder; two transverse mild steel dampers with capacity of 500 kN were installed between pier and girder. The maximum seismic response reduction rate reached 45.99% in the longitudinal direction and 40.44% in the transverse direction.The concentration of seismic responses in asymmetric structure was significantly alleviated. [Conclusion] The following constraint system is recommended for such bridges in high-intensity seismic zones. Mild steel dampers should be installed longitudinally between pylon and girder. The pier-to-girder connection should remain longitudinally movable. Mild steel dampers should be installed transversely at both pylon-girder and pier-girder connections. The damper specifications and quantities should be determined according to the optimized parameters. Quantitative references are provided for the seismic design of similar cable-stayed bridges.
[Objective] This study proposes a method based on influence matrix sensitivity analysis to control the complete cable replacement of a single-cable-plane CFST truss cable-stayed bridge with pylon-pier-girder consolidated system.Its main girder alignment, pylon-top deviation, and stay-cable forces are strongly coupled.The measured alignment is also affected by temperature. [Method] First, a coupling influence matrix for stay-cable force, main girder alignment, and pylon-top deviation was established. Key cable positions and sensitive construction stages were then identified. Subsequently, a control strategy was developed, including symmetrical zoning, batch replacement, graded loading, and synchronous tensioning. A matrix of time-temperature-main girder alignment was also established to identify and correct the temperature effect in the measured alignment. Finally, the proposed method was verified by using full-process finite element simulation and field monitoring data. [Result] The proposed method reduces the peak vertical deflection of main girder by about 8% compared with the conventional sequential replacement method. It also reduces the 95th-percentile deflection in the high-deflection stage by nearly 20%. The field test results indicate that the elevation deviation of main girder is controlled within ±10 mm after cable replacement. The cable force deviation of 59 stay cables is controlled within the optimal control range of ±5%, accounting for 77.6% of 76 stay cables of the entire bridge.The remaining cables all meet the engineering control limit of ±8%. The pylon-top deviation changes smoothly during replacement. No continuous accumulation is observed. [Conclusion] The proposed method enables coordinated control of main girder alignment, pylon-top deviation, and stay-cable force under uninterrupted traffic and strong temperature disturbance. It can provide technical support for replacement design, construction monitoring, and process acceptance of similar bridges.
[Objective] The mechanical properties of steel-reinforced ultra-high performance concrete (UHPC) structures at early curing age depend on the bond property of UHPC with steel reinforcement. It is crucial to understand the evolution rule of bond property of UHPC with ribbed steel reinforcement at early curing age for achieving good bond property. It is as well as the key to ensure effective force transfer and structural safety of reinforced UHPC components during the early stages of construction. [Method] It examined the bond strength, bond-slip behavior, and failure mode of ribbed steel reinforcement through pull-out test at six curing ages, i.e., 1-day, 1.5-day, 2-day, 3-day, 7-day, and 28-day. The influences of curing age on compressive strength, bond strength, and anchorage critical length were analyzed. [Result] The bond strength develops rapidly at early curing age, then slows down with increasing age. Bond strengths at 1-day, 2-day, 3-day reach 40%, 60%, 70% respectively of that at 28-day. The bond strength shows a linear relatedness with compressive strength at the same curing age. UHPC is prone to splitting cracks at early curing age once the steel reinforcement pulls out. The failure mode of UHPC with curing age exceeds 3 days is steel reinforcement slip failure. The steel reinforcement yielding occurs in 28 days. The anchorage critical length should take the influence of early curing age into account; it is recommended to be four times the steel reinforcement diameter when curing age of UHPC exceeds 3 days, while it is recommended to be five times the steel reinforcement diameter when curing age less than 3 days. [Conclusion] The bond property of UHPC develops quickly at early curing age; it is highly age dependent. The existing bond strength formulas have poor applicability to the early age, and the age-based modification is required.
[Objective] This study aims to establish a field-test-based time-varying mode of effective prestress for PC small box girders. It investigates the longitudinal distribution pattern of prestress loss along tendons, providing a basis for the prestress design and construction control of similar bridges. [Method] Three spans (20 m, 25 m, and 30 m) of PC small box girders were instrumented for long-term in-situ monitoring of effective prestress distribution. The duct friction factor and deviation coefficient were back-calculated via multivariate linear regression. The early-age elastic modulus of concrete was inversely determined from the measured elastic shortening loss in 7 days. Subsequently back-calculated the early-age elastic modulus of concrete, and corrected the calculation of elastic compression loss accordingly. A time-varying mode of effective prestress was then constructed and validated through long-term monitoring data. [Result] The measured duct friction factor is 0.264, which is in good agreement with the current specified value of 0.250. The deviation coefficient is 0.003 3, which is about 2.2 times the specified value of 0.001 5. It indicates that the duct positioning deviations in small-section box girders substantially amplify the friction loss. The 7-day concrete elastic modulus reaches only 72.2% of the standard value, resulting in elastic shortening losses exceeding specified predictions. The proposed time-varying mode achieves an average prediction error of 2.25%, with long-term (50 years) effective prestress assessment accuracy exceeding 90%. [Conclusion] The strict control of tendon duct positioning accuracy is essential during construction of small box girders. The measured elastic modulus should replace the specified standard value for early-age tensioning when computing elastic shortening loss. The proposed time-varying mode provides a reliable reference for long-term performance evaluation of similar PC small box girders.
[Objective] The location-routing problem (LRP) in rural e-commerce logistics networks is a critical link of rural e-commerce development. The effective location-routing problem research can significantly alleviate the issues of high costs and low efficiency in rural e-commerce logistics distribution networks. [Method] The distribution costs (including sites and routes), penalty costs of village-level service points, and carbon emission costs of multiple types of vehicles (fuel vehicles and electric vehicles) were taken into account. With the goal of maximizing profit of logistics enterprises with government subsidies, the pickup demand at village-level service points was treated as a fuzzy variable to construct a two-echelon LRP with multi-vehicle (2E-LRP-MV) model. The triangular fuzzy numbers were introduced to characterize demand; and the expected value planning method was applied to clarify the fuzzy demand. A two-echelon hybrid heuristic algorithm was designed to address the complexity of solving LRP. In the first echelon, the genetic algorithm was used to determine the location of township distribution centers; while in the second stage, the ant colony optimization algorithm was employed to optimize the two-echelon routing. Finally, an empirical analysis was conducted in Baishui County, Weinan City, Shaanxi Province. [Result] The hybrid heuristic algorithm can effectively solve the 2E-LRP-MV model. When government's subsidies for two-tier networks are set at 0.4 CNY per item and 0.6 CNY per item respectively, the profitability of logistics enterprises is ensured while promoting the development of rural logistics. Additionally, using electric vehicles as delivery tools significantly reduces the carbon emission costs of logistics network. [Conclusion] It provides a theoretical foundation for optimizing the county-township-village e-commerce logistics network. The well-designed government subsidies and the adoption of electric vehicles can jointly improve both economic and environmental performance of logistics system.
[Objective] This study investigates the socio-economic and environmental benefits of transforming provincial rural E-commerce logistics from traditional independent distribution to joint distribution. It pinpoints the key indicators influencing distribution costs. [Method] Commencing from factors, e.g., rural economy, rural demand levels, logistics enterprise supply capacities, energy, and environment, this study analyzed the variations in vehicle trip volume, distribution costs, energy consumption, carbon dioxide emissions, and distribution frequency following the implementation of joint distribution. The joint distribution model for rural E-commerce logistics was constructed using system dynamics by analyzing system interconnections across four dimensions, i.e., rural economy, demand levels, logistics enterprise supply, and energy environment. The simulation analysis was performed by using Vensim PLE, based on the data of Liaoning Province from 2016 to 2021. [Result] Implementing a joint distribution model for the continuous growth of supply capabilities in rural express and logistics enterprises is expected to reduce the total vehicle trip by 73.75%-79.0%. The express delivery cost is expected to decline from 1.31 CNY per item to a range of 0.64-0.91 CNY per item. The energy consumption and carbon dioxide emissions are expected to decrease by 30%-51%. The primary indicators affecting joint distribution costs are the full load rate, average distance, and distribution frequency. The optimal decision-making range for vehicle full load rate lies between 80% and 90%. The distribution cost increases by 0.10-0.12 CNY per item within this range for every 5 km increase in average transport distance. The distribution frequency can be reduced without compromising rural residents' satisfaction when distribution volumes are low. [Conclusion] Implementing joint distribution in rural areas can effectively reduce distribution costs, decrease vehicle trip volume, and lower energy consumption and pollutant emissions. The findings can serve as a reference for investigating new models for provincial rural logistics development.
[Objective] A route optimization model for multi-depot and heterogeneous vehicles with time window is constructed, due to the practical challenges of high safety risks, cost-sensitive operation and strict low-carbon constraints in the road distribution of refined oil. The model takes into account the premise of safety, economy, and environmental friendliness. An efficient solution method is proposed to provide theoretical support and technical approaches for safe and green transportation as well as refined scheduling decision-making of refined oil. [Method] A risk assessment model, integrating real-time vehicle load and population density, was developed; as well as a comprehensive modal emission model, incorporating driving speed, travel time, vehicle hardware parameters and real-time load to make the assessment of risk and carbon emissions more consistent with actual transportation scenarios. A multi-objective optimization algorithm named NSGA Ⅱ-Tent-VNS was developed based on NSGA Ⅱ framework to solve the model efficiently. This algorithm adopted the improved Tent chaotic mapping for population initialization to improve population diversity. Subsequently, individual evolution was realized through an adaptive crossover and mutation mechanism. The routes after crossover and mutation were optimized combining with problem-specific variable neighborhood search, so as to further improve the quality of solutions and search efficiency. Finally, simulation tests were carried out based on real data of the refined oil transportation network in Xi'an City and Solomon benchmark instances of different scales. [Result] NSGA Ⅱ-Tent-VNS has significant advantages in convergence stability and solution distribution compared with the classical NSGA Ⅱ algorithm. The total risk, total cost and carbon emissions of its optimal solution are reduced by 7.5%, 6.5% and 16.3% respectively. Meanwhile, heterogeneous vehicles configuration can simultaneously reduce the overall system risk, transportation costs and carbon emission levels. [Conclusion] The proposed model and improved algorithm have favorable applicability and effectiveness with the complex constraints of multi-depot collaboration and heterogeneous fleet mixing. They can provide a scientific decision-making basis for refined oil transportation enterprises to formulate transportation route schemes, optimize vehicle configuration and improve the operation efficiency of transportation systems.
