[Objective] As a key foundation for human survival, production, life and development, the urban transportation system suffers from spatiotemporal imbalance between supply and demand, leading to severe problems, e.g., congestion, accident, emission and inefficiency. Research on urban road traffic congestion tracing is vital for achieving the dynamic balance between supply and demand on road networks. This study systematically reviews its research status and development trends. [Method] First, the study was based on the distribution characteristics of travel demand, the evolution of traffic bottlenecks, as well as the deep-rooted causes of traffic congestion. It constructed a three-layer methodological framework for urban road traffic congestion tracing analysis. The first layer was Origin-Destination (OD) tracing analysis, which traced the origins and destinations of traffic flows causing congestion. The second layer was traffic bottleneck tracing analysis, which characterized the spatiotemporal evolution patterns of traffic states at bottlenecks. The third layer was congestion causality tracing analysis, which investigated the fundamental reasons behind the supply-demand mismatch. Building upon this framework, a literature search was conducted utilizing CNKI database (from 2003 to 2025) and Web of Science core collection (from 1994 to 2026) as data sources. The search used combined keywords, e.g., congestion tracing, OD estimation, bottleneck identification, and causality analysis. A total of 183 papers were included. Finally, existing research was systematically summarized and reviewed through VOSviewer for bibliometric analysis. [Result] The current tracing analysis has initially achieved a transition from single-section traffic volume statistics to network-level spatiotemporal trajectory reconstruction. OD tracing can effectively capture the network distribution of traffic demand leading to congestion, achieving vehicle source localization. The traffic bottleneck tracing enables spatiotemporal analysis on congestion bottleneck formation, revealing the evolution patterns of traffic states. The congestion causality tracing attempts to investigate the deep-rooted origins of congestion generation from the perspectives of plan, design, and management. The accurate and reliable multi-source heterogeneous data continue to accumulate with the development of traffic information technology. It promotes the innovation of urban road traffic congestion tracing analysis and related applications, improving the accuracy and real-time capabilities of results. [Conclusion] The findings offer guidance for urban road traffic congestion tracing analysis, holding significance for the sustainable and innovative development of urban transportation systems in the future. [Prospect] Regarding future research trends, the urban road traffic congestion tracing analysis should integrate multi-source data and modeling approaches, with a specific focus on analyzing critical links within the urban road network. Future research directions are summarized into three main aspects, i.e., the tracing analysis methods combining data-driven approaches and traffic flow models, the traffic congestion tracing and urban transportation system reconstruction, and the translational application from tracing diagnosis to targeted governance.
[Objective] To effectively alleviate congestion propagation in accident scenarios, the studies on identification methods for critical paths in congested areas were conducted for coordinated control. [Method] This study updated the definition of maximum flow by improving the parameter defects in existing relational degree models; as well as constructed a relational degree model for urban road network accidents. Considering the influence of right-turning and U-turning vehicles, a state disturbance relational degree was proposed. It was weighted and fused with the improved traffic demand relational degree to obtain a comprehensive relational degree. The critical path was determined based on the improved algorithm for breadth recognition. Simulation tests were conducted using VISSIM to examine the effectiveness of the improved model in practical applications. [Result] After interference on critical paths, the time for the average speed at road section to fully recover is 447 s, which is reduced by 108 s compared with that at self-organized state of traffic flow; and only increases by 25 s compared with that at the regional full path interference state. In addition, the dissipation time at congested section caused by accident is reduced by 164.9 s compared with that at self-organized state of traffic flow; and only increases by 2.3 s compared with that at regional full path interference state. [Conclusion] The improved comprehensive relational degree model can accurately identify critical paths within a unit of time, achieve precise interference, and improve the efficiency of road network control. The findings provide theoretical support and case reference for the study on social vehicle traffic organization, congestion control and induction in case of sudden traffic accidents.
[Objective] To meet the real-time governance requirements of urban traffic congestion, this study focuses on critical links that significantly influence traffic operational efficiency and congestion propagation within local networks. The spatio-temporal propagation characteristics of traffic congestion is analyzed, aiming to establish an efficient method for identification of critical congested links, thereby providing robust decision-making support for dynamic traffic control. [Method] A dynamic spatio-temporal graph attention network model was proposed. It innovatively constructed a dynamic graph representation learning framework, integrating the topological structure of road network with spatio-temporal evolution characteristics of traffic flow. The model quantified the time-varying traffic correlation among road network nodes. On this basis, the mathematical representation of local congestion clusters was defined by combining graph theory. The congestion propagation influence index was designed to quantify the criticality of links in the local road network, in order to accurately identify critical congested links. The study conducted empirical analysis based on real taxi GPS trajectory data from Chengdu, and carried out case verification through traffic simulation software SUMO. [Result] The proposed model can effectively capture the spatial proximity of road network nodes and the spatio-temporal dynamics of traffic flow propagation. The critical links identification algorithm achieves sub-second processing latency, maintaining the real-time performance even at scales of 100 000 trajectory points. The traffic control measures implemented in case studies significantly improve the network efficiency, reducing the congestion duration by more than 50%. [Conclusion] The formation of local congestion clusters is typically dominated by a small number of critical links. Targeted interventions on such critical nodes can effectively curb the further spread of congestion. The proposed algorithm for identification of critical congested links is both real-time and effective, providing reliable decision-making support for urban traffic governance and control.
[Objective] The existing lane change data extraction methods struggle to adapt to diverse lane change trajectories and have low extraction accuracy. An adaptive lane change data extraction method based on trajectory pattern classification is proposed. [Method] First, the dual dynamic time warping method was adopted, which integrated the similarities of trajectory morphology and motion states, thereby enhancing the quantifiable representation of trajectory morphological similarity. Second, DDTW-NN classification method was proposed by incorporating DDTW into Nearest Neighbours classification to perform fine-grained identification and classification of lane change trajectories on the initial lane and the target lane respectively. Furthermore, a pattern-adaptive feature selection mechanism and a conditional strategy for identifying start and end points of lane change were proposed for different categories of trajectory data. Based on the currently determined trajectory data category, the proposed method could dynamically select the most discriminative feature combinations and set the corresponding determination thresholds for the start and end points. It improved the accuracy of extracting lane change data on both the initial and target lanes, avoiding the issue where a single fixed threshold failed to adapt to different lane change scenarios. Finally, experiments and tests were conducted on NGSIM dataset. [Result] DDTW-NN achieved the classification accuracy of 85.45% for lane change trajectories, which was improved by 8.52% compared with DTW-NN method. The adaptive lane change data extraction method reduced the probability of determined start and end points deviating from lane centreline to 12%, compared with 28%-44% by using existing methods. Simultaneously, it reduced the probability of vehicles remaining in an unstable driving state, i.e., 2%, before start point and after end point, compared with 18%-26% by using existing methods. [Conclusion] The proposed adaptive lane change data extraction method can accurately identify the start and end points of lane changes and improve extraction accuracy, providing a reliable data foundation for the understanding of lane change behaviours and high-precision modelling in various scenarios.
[Objective] This study deeply investigates the issue of large-capacity traffic flow on future expressways, especially in ultra-high-speed driving conditions. How to accurately assess and calculate the traffic capacity of expressway? The key considerations include the changes in driver response time under different following behaviors, as well as the influence of penetration rate of connected and automated vehicles (CAV) on expressway traffic capacity. [Method] A reaction time-speed gradient prediction model was constructed by collecting driver reaction time data within the speed range of 120 km/h. The response time parameters in ultra-high-speed driving conditions of 150 km/h and 180 km/h were extrapolated and obtained. The kernel density estimation method was used to establish a dynamic speed distribution model for preceding vehicles; and the speed correlation mechanism in different following scenarios was analyzed. Three types of following models for mixed traffic flow of human-driven vehicles and CAV were innovatively established, i.e., human-driven vehicle following preceding vehicle, CAV following human-driven vehicle, and CAV following CAV. The super expressway scenario was constructed through SUMO simulation platform. The theoretically calculated and simulated traffic capacity values were compared; and the deviation rate between two methods was controlled within 5%. The consistency of results with two methods verified the accuracy of theoretical model.The study further quantified the improvement effect of penetration rate of CAV on traffic capacity. [Result] When the penetration rate reaches 80%, the correction coefficient of traffic capacity can reach 2.08, and the increase in penetration rate has significant influences on traffic capacity. [Conclusion] The goodness of fit (i.e., R2) of reaction time prediction model reached 0.99 and 0.97 respectively. MSE, RMSE, MAE of the theoretical model and simulation results had errors within a reasonable range, verifying the scientificity and applicability of traffic capacity calculation method in ultra-high-speed environments.
[Objective] The traditional expressway hub signs generally suffer from fixed information and limited capacity, e.g., information overload, insufficient guidance accuracy, and difficulties in driver recognition and decision-making in complex scenarios with multiple intertwined paths. A dynamic selection method for intelligent connected sign information at hubs based on deep reinforcement learning was proposed to break through the bottleneck of traditional hub sign information display. It aims to achieve precise matching of hub sign information with real-time traffic conditions and driver's need. [Method] First, an evaluation system for hub sign information was established using analytic hierarchy process. The comprehensive weight affecting selection result of sign information was determined by combining the standard deviation method, effectively avoiding the bias of single subjective weighting. Second, the decision matrix was established by normalizing various indicators of sign information. The sign information selection problem was described as Markov decision process. A deep reinforcement learning model was established, defining the state space, action space, and reward function for sign selection. Finally, a double deep Q-network algorithm was proposed to solve this model, and the optimal sign information selection strategy was obtained after offline training. [Result] Compared with similar algorithms, the proposed algorithm achieves 34.6% improvement in score deviation rate and 45% reduction in single decision calculation time with the same indicator data. [Conclusion] The proposed method can dynamically adjust the display content of signs according to the real-time traffic conditions of expressway hubs, providing technical support for the intelligent connected upgrade of expressway hub signs. It achieves the multi-path and efficient guidance for expressway hub sign information, effectively improving the service efficiency and level of expressway networks.
[Objective] This study investigates continuous real-time monitoring technology based on SmartRock sensing for rutting deformation in asphalt pavement. [Method] First, the evolution characteristics of asphalt mixture skeleton were investigated by integrating SmartRock sensing with discrete element numerical simulation technology. Subsequently, the real-time monitoring data of SmartRock during evolution process of rutting deformation was analyzed. The mechanical and motion response rule of SmartRock during monitoring process were revealed. Furthermore, a feature importance analysis method based on average fusion was proposed to extract rutting monitoring indicator characteristics. The relative cumulative rotation angles of SmartRock around three axes were selected as monitoring indicators for rutting deformation. Finally, a regression model of rutting deformation depth at varying distances from the wheel path was established. A monitoring method for rutting deformation based on SmartRock was proposed, which could be used to identify and assess characteristics in different rutting evolution stages. [Result] SmartRock sensor embedded with the medium skeleton had the most obvious mechanical and motion response among the coarse, medium, and fine skeletons of AC-13 asphalt mixture designed based on graded skeleton theory. The peak vertical contact force on medium skeleton reached 4.2 kN, while SmartRock sensor primarily interacted with aggregates of 2.36 mm. The dominant motion modes of SmartRock sensor embedded in skeleton were primarily characterized by translational motion in vertical axis and rotation around vertical axis. SmartRock were primarily subjected to vertical forces during the rutting deformation process, and their motion was dominated by translational movement in the vertical direction and rotational movement around the vertical axis. The vertical peak displacement was 0.85 mm, and the maximum relative cumulative rotation angle around the vertical axis was 1.3°. [Conclusion] The importance weight of SmartRock's rotational responses was nearly 10 times higher than that of other indicators. The average goodness of rutting deformation depth regression model established through rotation angles of SmartRock was 0.965. The model demonstrated strong predictive capability for rutting deformation depth.
[Objective] The asphalt mixture performance reinforced with plant fibers at high and low temperature can be enhanced by modifying the surface of plant fibers. This study aims to clarify the influence of modification methods on microscopic interface and macroscopic properties of materials. [Method] The high-temperature stability and low-temperature breaking resistance of asphalt mixture reinforced with melamine-formaldehyde modified bamboo fibers were measured through rutting tests and bending tests. The molecular structure and dynamic behavior of interfaces, between asphalt and bamboo fibers before and after modification, were investigated through molecular dynamics simulations. The quantitative analysis on binding energy, diffusion coefficient, and relative concentration of asphaltene molecules was employed to study the adhesion to bamboo fibers and the distribution pattern at different temperatures. [Result] The asphalt mixture reinforced with modified bamboo fibers exhibited 10.4% increase in high-temperature dynamic stability and 16.6% increase in low-temperature bending strength. The interfacial adhesion between bamboo fibers and asphalt improved significantly as grafting density increased from 0 to 1.11×10-7 mol/m2. The interfacial binding energy of modified system increased by 266.2% at 338 K, and the diffusion coefficient decreased by 49.3%. The binding energy increased by 145.8% at 258 K, and the diffusion coefficient decreased by 38.3%. Electrostatic forces dominated the interfacial binding energy. The relative concentration profiles revealed that further increasing grafting density from 1.11×10-7 mol/m2 to 4.43×10-7 mol/m2 inhibited direct contact between asphalt and bamboo fibers due to steric hindrance from hydroxyl groups in the grafted molecular chains, leading to degraded interfacial performance. Consequently, the relative concentration of asphalt at bamboo fiber interface decreased by 25.0% at 258 K, and by 34.4% at 338 K. [Conclusion] This study provides insights for optimizing high and low temperature performances of asphalt mixture reinforced with modified plant fibers.
[Objective] This study investigated the deformation response mechanism of fiber-solidified lightweight soil due to combined effect of intermittent loading and freezing-thawing cycle. The mechanics under dynamic load was investigated. [Method] First, the dynamic triaxial test, control variable method and orthogonal test were conducted to simulate road conditions in regions with large temperature differences between winter and summer. Second, the cumulative strain of soil was studied due to three main factors, i.e., temperature, freezing-thawing cycles, and dynamic stress amplitude. The soil deformation due to single-factor and double-factors was analyzed respectively. Finally, a prediction model for cumulative deformation was established, considering the influences of freezing-thawing cycle and fiber reinforcement coefficient on fiber-solidified lightweight soil. The validity of model was verified as well. [Result] The fiber reinforcement effect can effectively inhibit soil deformation. The cumulative strain curve of fiber-solidified lightweight soil exhibits the types of stable, critical, and destructive in different freeze-thaw conditions. An increase in freezing-thawing cycles, a decrease in freezing temperature, or an increase in dynamic stress amplitude will all exacerbate the soil deformation. The combined effect of freezing-thawing cycle and dynamic stress amplitude has significant influence on the dynamic characteristics of soil. When the dynamic stress amplitude is less than 150 kPa, the increase in temperature has a limited effect on the axial strain. The temperature rise leads to a significant increase in axial strain, and the failure rate of soil is accelerated when the dynamic stress amplitude is 250 kPa and the freezing-thawing cycle exceeds three times. The dynamic deformation of fiber-solidified lightweight soil basically reaches a stable state when the freezing-thawing cycle exceeds 9 times. The cumulative axial strain of soil, in the condition of 9 freezing-thawing cycles and 400 kPa dynamic stress amplitude, is two to three times higher than that in the condition of 1 freezing-thawing cycle and 200 kPa dynamic stress amplitude; the difference is especially obvious at ―15 ℃. [Conclusion] The findings can provide a scientific basis for the design and construction of fiber-solidified lightweight soil in subgrade filling engineering in cold regions.
[Objective] This study investigates the swelling deformation and settlement of swelling soil subgrade. It took the swelling soil from Zaoyang-Qianjiang expressway as the study object, as well as taking plant ash and fly ash as modifying materials. [Method] This study carried out experiments on free swelling ratio, unloaded swelling ratio and direct shear strength. It analyzed the variation rules of swelling ratio and shear strength of modified swelling soil with different proportions of modifying materials. Meanwhile, the micro-modification mechanism of swelling soil was investigated from a fine-grained point of view through X-ray diffraction tests and scanning electron microscopy observations. [Result] The incorporation of plant ash and fly ash changed the particle composition and structure within soil, which effectively suppressed the swelling of soil and significantly increased the shear strength. The free swelling ratio decreased from 48% to 34% when modified with 10% plant ash and 14% fly ash; the shear strength increased by 46.70% and 42.71% respectively under 100 kPa and 300 kPa of vertical pressure; the cohesion of modified soil increased from 14.20 kPa to 22.57 kPa; and the angle of internal friction increased from 23.61° to 30.20°. The modified swelling soil exhibited new hydration products, e.g., C—S—H and C—A—H. The internal porosity of modified soil was significantly reduced due to the filling of modifying materials and hydrates. The large pore size almost disappeared. The contact among soil particles was more compact. It improved the strength of swelling soil and reduced the swelling ratio. [Conclusion] The findings will provide a scientific basis for solving the deformation of swelling soil subgrade; as well as provide a new idea for realizing the application of solid waste resources in subgrade improvement, which has good engineering application prospects.
[Objective] This study proposed a control method for subgrade settlement based on suspended cement mixing piles to address the uneven settlement in expressway widening projects on weak subgrade. [Method] First, the field tests of suspended cement mixing piles were conducted. A monitoring network, including surface displacement and horizontal deep displacement, was established to obtain the settlement and lateral displacement data of subgrade during both construction and operation. Subsequently, the settlement deformation of widened subgrade was precisely numerical simulated by using the field monitoring data and a soft soil creep model. The model parameters were validated through back analysis. Finally, three-point modified hyperbolic method was employed to predict the long-term development of subgrade settlement by using the observed settlement data. [Result] The strength of cement mixing piles generally ranges from 0.1 MPa to 0.6 MPa. The pile formation effect is good, effectively improving the bearing capacity of composite foundation. The construction of suspended cement mixing piles has a significant influence on the settlement of adjacent aged subgrade. The settlement exhibits an approximately linear growth trend during the initial construction, i.e., the first three days. The settlement rate gradually decreases during the intermediate stage, i.e., from the 3rd to the 7th day. The settlement process tends to stabilize in 7 days. The significant differential settlement is observed during the widening construction on the left lane. The settlement of left shoulder ranges from 50.83 mm to 56.61 mm on sections with general fill, i.e., a monthly settlement rate of 3.99-6.78 mm. The right-side settlement ranges from 6.43 mm to 9.50 mm. The differential settlement reaches 44.40-47.11 mm. The left-side settlement ranges from 18.48 mm to 24.05 mm on sections with lightweight soil, i.e., a monthly settlement rate of 1.44-3.06 mm. The differential settlement ranges from 12.95 mm to 16.69 mm. The total settlement is approximately one-third of that on sections with general fill. The shoulder of adjacent aged subgrade is predicted to undergo an additional settlement of 4.5 cm without widening load over the next 20 years. The cumulative settlement over 20 years is predicted to be 11.4-19.9 cm when the widening load is considered. The settlement on lightweight soil sections is approximately one-third lower than that on sections with general fill. [Conclusion] This study systematically investigates the deformation characteristics of composite foundations reinforced with suspended cement mixing piles in weak subgrade. The superiority of lightweight soil is demonstrated in controlling settlement and cross slope variation of weak subgrade. The findings provide a quantitative basis for settlement control and material selection in expressway widening projects over deep soft soil areas. Future study should focus on the long-term performance evolution of suspended cement mixing piles.
[Objective] Road defect detection is a critical component of intelligent highway maintenance, as its performance directly determines the quality and efficiency of road upkeep. A road defect detection algorithm based on dynamic convolution and cross-attention was proposed to address the collaborative optimization challenges of dynamic receptive field adaptability, feature disentanglement efficiency, and inter-layer feature consistency in existing methods. [Method] First, a collaborative mechanism, combining deformable convolution's spatial adaptive offset with dynamic receptive field's scale adaptive adjustment, was employed to improve the multi-scale feature extraction. Second, the spatial-channel cross-attention was adopted to reduce the computational complexity while improving the target extraction in complex backgrounds. Simultaneously, the dynamic cross-layer feature fusion was implemented to strengthen semantic correlation among channels. Finally, a down-sampling module ADown was designed to optimize the network topology, further reducing the model complexity. [Result] The proposed algorithm achieved 90.7% mAP, improving 2.6% compared with baseline models. The number of parameters decreased by 9.5%. Ablation tests indicated optimal collaborative optimization across all modules. [Conclusion] The proposed algorithm improves the detection accuracy while effectively reducing the model complexity, offering the efficient and lightweight technical solutions for intelligent highway maintenance.
[Objective] To improve the detection capability of road defects in logistics parks, this paper introduces LFGFBlock module into Mamba model and combines the improved Mamba model with YOLOv8 head, thereby constructing a novel intelligent road defects detection model for logistics parks, i.e., Road Defects Detect-MambaYOLOv8 (RDD-MY). [Method] To effectively capture the cross-scale and long-range dependent features of road damage images in logistics parks, the LFGFBlock module was employed to improve the interaction between local and global information, thereby improving the detection accuracy of model. The synergistic effect of two models combined the efficiency of CNNs in local feature extraction with the global modeling capability of Transformers. A dataset comprising 6 500 typical road damage images from logistics parks was constructed to provide data support for model training. The effectiveness of incorporating LFGFBlock module into Mamba model and the overall performance advantages of RDD-MY model was validated through ablation test and comparative test. [Result] The proposed RDD-MY model improves precision, recall, PmA, 50, and PmA, 95 by 16.9%, 13.3%, 16.8%, 17.2% respectively compared with Mamba model. The proposed model demonstrates significant advantages across all evaluation metrics, compared with YOLOv5n, YOLOv6n, YOLOv7-tiny, and YOLOv8n models. [Conclusion] RDD-MY model exhibits stronger detection robustness and generalization capability compared with the aforementioned models. The findings can provide effective technical support for intelligent detection and routine maintenance of road defects in logistics parks, demonstrating promising engineering application potential and promotional values.
[Objective] To improve the accuracy and generalization of prediction models for concrete shrinkage and creep, a novel model was proposed, utilizing the sparrow search algorithm (SSA) to optimize the random forest (RF) algorithm for concrete shrinkage and creep prediction. [Method] American Northwestern University's database of shrinkage and creep was employed, undergoing rigorous screening, completion, and cleaning. The selected data were then split into training sets and testing sets, which were randomly divided in a 9:1 ratio. Multiple prediction models, i.e., SSA-RF, RF, LSTM, and BP, were used to predict shrinkage and creep. The evaluations were conducted using three metrics, i.e., coefficient of determination, mean absolute error, and root mean square error, to assess the reliability and accuracy of prediction. [Result] SSA-RF model exhibited the best performance in predicting concrete shrinkage and creep, followed by the RF model, then the BP model, with the LSTM model performing the worst. Notably, when compared with traditional RF model, the SSA-RF model, optimized by SSA, showed an improvement of 4.8% in the coefficient of determination for creep prediction and 2.6% for shrinkage prediction. Additionally, SSA-RF model achieved a reduction of 25.5% in mean absolute error for creep prediction and 40.8% for shrinkage prediction. Similarly, the RMSE decreased by 23.2% for creep prediction and 40.8% for shrinkage prediction. Concrete creep tests further indicated that SSA-RF model's predictions were closer to the actual values, with an average relative error of 6.7%, indicating its superior predictive performance. [Conclusion] SSA-RF concrete shrinkage and creep prediction model proposed in this study significantly improves prediction accuracy, offering valuable insights for practical engineering applications.
[Objective] Existing inclination-based bridge flexural deformation algorithms do not fully utilize the boundary conditions at intermediate supports. This paper proposes the algorithm with quartic Hermite interpolation, combining inclination and boundary displacement. [Method] The proposed algorithm aims to achieve zero deflection at intermediate supports, which determines the optimal shape control parameters in fourth-order Hermite interpolation, thereby improving calculation accuracy. First, the proposed algorithm was compared with the sub-span least squares method and the piecewise cubic spline interpolation method by using inclination data obtained from numerical analysis; in terms of global and local accuracy for calculating flexural deformation under concentrated load, temperature gradient load, and self-weight load. Subsequently, the proposed algorithm allowed flexural deformation to approach the known deflection at the intermediate support through the interval iteration of shape control parameters, resulting in high accuracy in calculating flexural deformation under all three types of loads. Finally, the calculated deflection time-history at each mid-span was compared with that measured with linear variable differential transformer through field test. [Result] The result indicates strong consistency between the calculated deflections and measured deflections. [Conclusion] The proposed algorithm demonstrates high accuracy under multiple types of load and good stability with field-measured inclinations, rendering it applicable to bridge flexural deformation monitoring.
[Objective] Automatic identification of modal parameters is necessary to realize real-time online monitoring of bridge health monitoring systems without human intervention. An automatic identification method for modal parameters, using Gaussian mixture model clustering based on variational Bayes, was proposed to automatically analyze stabilization diagrams and identify modal parameters. [Method] First, the noise reduction method for bridge vibration signals based on adaptive variational modal decomposition was used to denoise the measured signals. Subsequently, the stabilization diagrams were formed by using the covariance-based stochastic subspace method. The pseudo-modalities were eliminated using two widely validated and used modal validation criteria. Second, the improved DBSCAN algorithm was targeted to automatically determine the optimal number of clusters. Gaussian mixture models based on variational Bayes was used for clustering to realize the automatic identification of stabilization diagrams. Finally, the proposed method was applied to a large suspension bridge to verify its effectiveness. [Result] Seven modal frequencies within the first 1 Hz were successfully identified through the analysis on the measured data from the large suspension bridge. The relative errors between identification results and literature values were all less than 3%, with the minimum error being only 0.01%. [Conclusion] The proposed method for automatic identification of bridge modal parameters does not require any manually adjusted parameters or thresholds. It can effectively identify dense modes and be applied to real-time monitoring systems for bridge health monitoring.
[Objective] This study investigated the influence of reamed plate bearing on pile shaft resistance. It proposed the calculation method for vertical bearing capacity of squeezed branch pile, based on the mechanical parameters of soil around pile. [Method] First, the pile shaft above and below a bearing plate was divided into four segments due to lateral friction resistance, i.e., normal segment, reduced segment, nonfunctional segment and enhanced segment. Because the settlement of loaded bearing plates could respectively lead to unloading of soil on pile side above the plate and squeezing of soil around pile under the plate; as well as lead to the variation of pile-soil relative settlement to adjacent segmental piles and soil around piles. The characteristic differences of lateral friction resistance of each pile segment were investigated. Second, the bearing failure mode of squeezed branch pile with multi-plates was constructed, combining with the characteristics of common shear failure of soil between small-spacing bearing plates. Third, the soil arching theory and spherical cavity expansion theory were introduced respectively according to the stress and deformation characteristics of each pile segment. The calculation methods were established for the end bearing capacity of bearing plate and pile tip, as well as the lateral friction resistance of each segment. The calculation formula for ultimate bearing capacity of piles was obtained by superposing the resistance of each unit segment. Finally, Kondner hyperbolic constitutive model was introduced to construct the formula for bearing capacity of piles according to the pile settlement. The rationality of the proposed method was verified through existing test data. [Result] The segmental stress characteristics of pile were determined by analyzing the stress and deformation effects of soil around pile induced by the settlement of bearing plates. It leads to a more comprehensive investigation on load-bearing mechanism of squeezed branch pile.The calculation method for bearing capacity of squeezed branch pile can be established based on the mechanical parameters of surrounding soil by introducing applicable theories for different segments to perform quantitative calculations of pile-soil interaction. [Conclusion] The proposed method has a scientific theoretical basis, compared with existing studies that still rely on tests and experience to determine the end resistance and lateral resistance of piles. The findings could provide theoretical guidance for the bearing capacity calculation of squeezed branch piles and the collation of data for test piles.
[Objective] Addressing the vulnerability of RC high-rise piers to damage under strong earthquakes, this study aims to establish a rapid and accurate prediction model for the seismic limit state and to thoroughly reveal the influence of structural design parameters. [Method] Various machine learning algorithms, including decision tree, artificial neural network, random forest, and support vector machine, were employed to construct nonlinear prediction models for the seismic limit state of RC high-rise piers. The models utilized pier height, cross-sectional area, concrete material properties, and reinforcement material properties as input indicators. On this basis, Shapley Additive Explanations (SHAP) and Partial Dependence Plots (PDP) analysis were introduced to evaluate the feature importance of the optimal model, exploring the contribution and influence trends of each design parameter on the prediction result. [Result] Among various machine learning models, the artificial neural network model showed the highest accuracy in predicting the ultimate drift ratio of RC high-rise piers. The coefficient of determination reached 1.0. The root mean square error was only 0.012. The model accurately captured and reflected the seismic capacity and behavior of piers under earthquakes. The pier height and cross-sectional area significantly influenced the prediction result. The importance of design parameters was proportional to their influence on the predicted value range of ultimate drift ratio. A significant positive correlation existed between pier height and ultimate drift ratio. [Conclusion] The findings provide a significant reference for the rapid seismic design and safety assessment of RC high-pier bridges. Furthermore, they offer a theoretical and practical basis for the prediction on the limit state of bridge structures.
[Objective] This study investigates to improve punching shear resistance of wet joints on fabricated ribbed-girder bridges.It focuses on the mortise-tenon joint structure. [Method] First, static loading tests were performed on four joint types, i.e., mortise-and-tenon, inclined, V-shaped, and conventional straight joints. The study examined the influence of joint types on punching shear resistance of fabricated ribbed-girder bridges.Subsequently, reinforcement layouts and geometric details of all joint types were designed.To replicate field construction conditions, the specimens were cast in two stages, and joint interfaces were roughened. Finally, a conservative analytical expression for punching shear capacity was developed. [Result] All joint specimens showed the tendency for joint concrete to spall vertically as a frustum-shaped block.The ultimate failure mode was punching shear. It manifested as a conical breakout localized at the joint.Following steel yielding, the arch-truss mechanism in inclined, V-shaped, and straight joints became ineffective; in contrast, the mortise-tenon joint exhibited markedly higher punching shear resistance.Cracking loads were nearly identical across all specimens.Yield loads were also comparable among all specimens.The ultimate punching shear capacity of mortise-tenon joint exceeded that of the conventional straight joint by 10.5%. [Conclusion] The mortise-tenon wet joint significantly improves punching shear resistance of fabricated ribbed-girder bridge.The findings provide valuable guidance for improving the structural performance of wet joints in such systems.
[Objective] This study investigates the temperature field evolution rules in cold-region tunnels and embedded utility galleries during operation; as well as clarifying the influence of ventilation conditions on temperature distribution. [Method] Taking Dalu No.2 Tunnel on Benxi-Ji'an Expressway as the engineering background, the on-site monitoring of temperature field in surrounding rock was carried out. A numerical model of temperature field in cold-region tunnels was established. The longitudinal and radial distribution characteristics of temperature field in embedded utility gallery were systematically investigated due to the combined effect of different inlet wind speeds and air temperatures. [Result] The temperature field in surrounding rock at sidewalls and invert generally exhibits a sinusoidal variation pattern. With tunnel depth increasing, the annual temperature amplitude decreases, while the annual mean temperature increases. The temperature at different monitoring points in surrounding rock shows V-shaped trend in spring and winter, i.e., decreasing first and then increasing. Moreover, with the increase in radial distance, the temperature at all measuring points shows an increasing trend. The temperature exhibits an opposite trend in summer and autumn, i.e., increasing first and then decreasing. As the radial distance increases, the measured temperature shows a downward trend. The greater the radial distance, the more gradual the temperature decreases. As the inlet wind speed increases, the temperature in utility gallery gradually decreases. The closer to the center of gallery, the faster the temperature changes; the closer to the bottom, the slower the temperature changes. The greater the inlet air velocity and temperature, the more significant the convective heat transfer between gallery and tunnel air. It makes the temperature field at gallery top to be significantly affected by the air velocity and temperature. [Conclusion] The findings provide a theoretical basis for the thermal insulation design and operational safety assessment on cold-region tunnels and their embedded utility galleries. Meanwhile, the proposed temperature field model and analytical approach offer guidance for temperature control and ventilation optimization on similar cold-region tunnels with embedded utility galleries.
[Objective] This study investigates the influence of nano-CSH gel on performance of shotcrete, thereby solving the excessive rebound rate of shotcrete for tunnel engineering. [Method] This study conducted laboratory tests to investigate the influence of different nano-CSH gel contents (0%-4%) on setting time, compressive strength and rebound rate of cement paste, based on Tongqing No.2 tunnel project. The mechanism of shotcrete performance improvement by using nano-CSH gel materials was analyzed through scanning electron microscopy. The engineering field test was carried out for verification. [Result] nano-CSH gel can significantly improve the procoagulant effect, mechanical properties and rebound rate of shotcrete. With the increase of nano-CSH gel content, the initial setting time and final setting time of cement paste show a decreasing trend. The 1-day and 28-day compressive strength of concrete shows an increasing trend, while the rebound rate shows the trend of first decreasing and then increasing. When nano-CSH gel content is 2%, the initial setting time and final setting time are shortened by 31.8% and 39.4% respectively compared with the control group; 1-day and 28-day compressive strength is increased by 19.7% and 4.4% respectively; and the rebound rate is reduced by 35.6%. The engineering field test result indicates that when nano-CSH gel content is 2%, the 1-day and 28-day compressive strength is increased by 16.4% and 5.6% respectively compared with the control group; and the rebound rates at arch top and arch waist are reduced by 37.0% and 29.6% respectively. [Conclusion] Due to the nucleation effect and filling effect of nano-CSH gel, it can accelerate the hydration speed of shotcrete and increase the amount of hydration products, as well as densify the internal structure of concrete, thereby improving the mechanical properties of shotcrete and reducing the rebound rate of shotcrete.
[Objective] Water injection technology is extensively applied in oilfield areas across the Loess Plateau. The leakage from water injection wells poses a significant threat to the safety of adjacent tunnel engineering. This study investigates the moisture movement patterns in water-rich loess strata following water injection well leakage, as well as the corresponding adverse influence on tunnel stability. [Method] The systematic analysis and simulation on water injection well leakage were conducted based on practical engineering cases, combining field engineering investigations with numerical simulation via COMSOL Multiphysics. [Result] The saturated scope of surrounding rock at spandrel on tunnel's far-well side accounts for merely 0.43 times that of the undisturbed stratum, whereas the saturated rock range at spandrel on the near-well side is 12.5 times greater than that on the far-well side. As water injection and leakage time elapses, the tunnel surrounding rock achieves saturation in the sequential order of arch springing, arch invert, arch waist, near-well side spandrel, far-well side spandrel and vault. The moisture content of surrounding rock on near-well side increases at a markedly faster rate, and the upward vertical movement of groundwater within surrounding rock is more pronounced than horizontal movement. In comparison with the original undisturbed stratum, the pore water pressure of surrounding rock on tunnel's near-well side increases by a factor of 1.26-1.31, while that on the far-well side decreases by 0.58-0.91. Additionally, a rapid growth phase of pore water pressure is observed in the arch surrounding rock during the continuous leakage process. As the duration of leakage increases, a triangular-shaped zone of stress reduction appears on the lining support. The support stress at near-well side spandrel, vault and arch invert rises by a magnitude ranging from 200.63 kPa to 294.31 kPa. [Conclusion] Water injection leakage accelerates the deterioration rate of surrounding rock on tunnel's near-well side, and intensifies the upward movement tendency of groundwater. With the extension of leakage time, stress redistribution occurs in the strata surrounding tunnel, accompanied by the stratum of a stress-weakened zone in the lining and a substantial increase in support stress on near-well side. The findings can provide valuable theoretical and engineering references for tunnel construction and long-term operation in loess-covered oilfield areas.
