Vehicle cross-wind identification and early warning system for cable-stayed bridge

Abstract

This invention relates to the field of traffic communication technology, specifically to a vehicle crosswind identification and early warning system for cable-stayed bridges. The system obtains crosswind environment data based on wind and vibration data; analyzes the crosswind environment data to generate a traffic risk map for the cable-stayed bridge, including risk areas and risk levels; identifies vehicle information and crosswind data of sentry vehicles, and collects their travel trajectory data; quantifies the trajectory deviation points and degrees in the risk areas using the sentry vehicle's travel trajectory data; identifies the weight and outline of target vehicles to obtain vehicle similarity with the sentry vehicles; obtains crosswind difference based on real-time crosswind data and crosswind data during transit; and determines the lateral deviation risk index of the target vehicle within the risk area based on the risk level, vehicle similarity, and crosswind difference, combined with the sentry vehicle's travel speed, trajectory deviation points, and trajectory deviation degrees. This invention uses the lateral deviation risk index to provide early warning communication for target vehicles.

Description

Technical Field

[0001] This invention relates to the field of transportation communication technology, specifically to a vehicle crosswind recognition and early warning system for cable-stayed bridges. Background Technology

[0002] In certain traffic sections such as large bridges, strong lateral wind fields are easily generated due to their structural and geographical characteristics. These wind fields and the complex turbulent airflow formed by their interaction with road infrastructure, especially bridge towers and cable systems, can exert sudden and non-uniform aerodynamic interference on the transport units in transit, which can easily cause the transport units to deviate from their trajectory, become unstable, or even overturn, posing a serious threat to traffic flow safety.

[0003] In existing traffic control systems, early warning mechanisms for such environmental risks typically rely on roadside variable message signs to broadcast macroscopic environmental data collected by fixed sensors to traffic flow. This approach is an open-loop, non-personalized traffic information broadcasting method, which, while providing initial warnings, has significant technical limitations.

[0004] First, the control signals issued by the existing system lack spatiotemporal precision, failing to reflect the distributed characteristics of the wind field along the driving path and local high-risk areas, making it difficult to provide refined path risk indications. Second, the guidance information it issues lacks the ability to differentiate between target vehicle types; for vehicle units with different aerodynamic characteristics and anti-overturning capabilities, it sends homogenized warning instructions, failing to achieve differentiated traffic safety guidance and alerts.

[0005] Furthermore, the existing system is essentially a one-way environmental information dissemination facility, lacking the ability to collect and process dynamic response data generated by the on-the-go transportation units after being affected by the environment, such as real-time driving trajectory, lateral acceleration, yaw rate, etc., and lacking a dynamic traffic risk identification and control optimization mechanism based on real-time feedback.

[0006] To address this, a crosswind recognition and early warning system for vehicles on cable-stayed bridges is proposed. Summary of the Invention

[0007] The purpose of this invention is to provide a vehicle crosswind identification and early warning system for cable-stayed bridges, which provides early warning communication based on the lateral deviation risk index of the target vehicle.

[0008] To achieve the above objectives, the present invention provides a vehicle crosswind recognition and early warning system for cable-stayed bridges, comprising: The risk identification module obtains crosswind environmental data based on wind and vibration data of the cable-stayed bridge deck, cable array, and towers; it then analyzes the crosswind environmental data to generate a traffic risk map of the cable-stayed bridge, including risk areas and risk levels. The risk assessment module identifies the sentry vehicles that are currently passing through; acquires vehicle information and crosswind data of the sentry vehicles, and collects their travel trajectory data when crossing the cable-stayed bridge; and quantifies the trajectory deviation points and trajectory deviation degrees in the risk area using the travel trajectory data of the sentry vehicles. The traffic comparison module identifies target vehicles about to enter the cable-stayed bridge; based on the target vehicle's weight and outline, it identifies the vehicle similarity to the sentry vehicle; and based on real-time crosswind data and traffic crosswind data, it obtains the crosswind difference. The communication guidance module identifies the target vehicle's lateral deviation risk index within the risk area based on risk level, vehicle similarity, and crosswind difference, combined with the sentry vehicle's travel speed, trajectory deviation point, and trajectory deviation degree; and conducts early warning communication based on the lateral deviation risk index.

[0009] The process of identifying the traffic risk map of the cable-stayed bridge includes: Using the bridge deck of the cable-stayed bridge as a one-dimensional coordinate axis, wind force data and vibration data from the crosswind environment data are projected onto the bridge deck, and the identified wind force magnitude, wind turbulence degree, cable vibration frequency and cable vibration amplitude are used as standard environmental data. Rainfall is monitored by precipitation sensing units, and standard environmental data is used to identify inherent risk areas of the cable-stayed bridge. These inherent risks include wind and rain vibration and wake vibration. For non-inherent risk areas, clustering is performed based on wind force, wind turbulence, cable frequency, and cable amplitude. Based on the clustering results of inherent risk areas and non-inherent risk areas, risk areas are obtained; based on the standard environmental data and inherent risk types in the risk areas, the risk level is identified, resulting in a traffic risk map for the cable-stayed bridge.

[0010] The risk assessment module deploys an integrated vehicle perception unit on the cable-stayed bridge to acquire vehicle information of the sentry vehicles; the integrated vehicle perception unit includes: a visual recognition unit, a dynamic weighbridge unit, and a roadside communication unit. The visual recognition unit determines the three-dimensional geometric contour of the vehicle through images captured by a camera; the dynamic weighbridge unit measures the weight of the vehicle as it passes; and the roadside communication unit conducts vehicle-road cooperative communication with the sentry vehicle to obtain vehicle status data, including traffic speed.

[0011] The risk assessment module analyzes the travel trajectory data of the sentry vehicles using a trajectory stability quantification model. The trajectory stability quantification model acquires the vehicle's driving trajectory and heading angle data from the travel trajectory data and performs the following analysis: Position stability analysis: Calculate the standard deviation of the vehicle's lateral position relative to the lane centerline to quantify the drift amplitude within the risk area; Attitude stability analysis: Identify the rate of change of the vehicle's heading angle data to capture the rotational trend caused by crosswind impact; Handling oscillation analysis: Perform spectral analysis on the time series of the vehicle's lateral displacement, extract the spectral power density within a preset frequency range, and identify handling oscillation behavior caused by overcorrection; Using the sentry vehicle's trajectory before entering the cable-stayed bridge as a benchmark, weight multiple analysis indicators to obtain the trajectory deviation; Obtain time-series data of trajectory offset, and record the geographical locations where the time-series data reaches a maximum value as trajectory offset points.

[0012] The traffic comparison module uses a multi-dimensional similarity matching algorithm based on dynamic aerodynamic features to identify the vehicle similarity between the target vehicle and the sentry vehicle. The vehicle's three-dimensional geometric contour is obtained by a visual recognition unit, and combined with the real-time crosswind direction vector, the three-dimensional geometric contour is projected to determine the vehicle's effective projected side contour under the corresponding wind field. A comparative analysis is performed on the effective projected side profile to extract the effective side area associated with the crosswind force and the profile shape factor characterizing the aerodynamic properties. Vehicle similarity is obtained by normalizing and weighting the effective lateral area, contour shape factor, vehicle weight, and average vehicle speed of the target vehicle and the sentry vehicle.

[0013] The process by which the traffic comparison module identifies and obtains the crosswind difference includes: Crosswind data is abstracted into state vectors, the dimensions of which include wind force data and vibration data acquired from the bridge deck, cable array, and bridge tower of the cable-stayed bridge; state vectors are established for the crosswind data during the passage of the sentry vehicle and the current real-time crosswind data, respectively, to obtain the passage state vector and the real-time state vector; The crosswind difference is quantified based on the normalized distance between the passage state vector and the real-time state vector in the multidimensional space.

[0014] The communication guidance module determines the lateral deviation risk index of the target vehicle through a weighted empirical risk projection model; the identification logic of the weighted empirical risk projection model is as follows: Sentry vehicles that meet a preset threshold in terms of vehicle similarity are selected based on vehicle similarity. Based on the highest value of the trajectory deviation of the selected sentry vehicles in the risk area and the risk level of the corresponding risk area, the basic risk value is identified; the basic risk value is then weighted twice based on vehicle similarity and crosswind difference to calculate the expected risk projected onto the target vehicle, namely the lateral deviation risk index.

[0015] The warning communication of the communication guidance module is carried out by the roadside communication unit through the vehicle-road cooperative protocol, which sends the warning information to the on-board unit of the target vehicle. The warning information is classified according to the magnitude of the lateral deviation risk index, including low risk, medium risk and high risk.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention deploys multi-modal sensors for wind and vibration on the key structures of cable-stayed bridges and integrates and analyzes multi-dimensional data such as wind magnitude, turbulence level, and structural vibration to generate a high-resolution, dynamic risk field map; it accurately identifies local extreme danger areas caused by bridge tower wakes, cable interference, etc., eliminating the perception blind spots of traditional solutions; more importantly, it incorporates wind-induced structural vibration, a risk factor unique to cable-stayed bridges that directly affects tire grip, into the monitoring, achieving a fundamental upgrade in risk perception.

[0017] 2. This invention innovatively uses sentry vehicles traveling on bridges as dynamic, distributed risk probes; through visual tracking and other means, it quantifies the trajectory instability of these vehicles in real-world environments in real time, directly linking the abstract crosswind risk with the observable and quantifiable actual vehicle response; it achieves a fundamental transformation from passive environmental monitoring to active effect observation, providing risk assessment with real-world ground-based evidence.

[0018] 3. This invention can accurately identify the personalized risks of a target vehicle by constructing a similarity matching algorithm based on dynamic aerodynamic features; it can also predict the lateral deviation risk index that the target vehicle may have in the risk area it is about to pass through by using a weighted empirical risk projection model; and based on forward-looking quantitative indicators, it can accurately push graded traffic guidance containing specific operating instructions to the driver through vehicle-road cooperation. Attached Figure Description

[0019] Figure 1 This is a structural schematic diagram of a vehicle crosswind recognition and early warning system for cable-stayed bridges according to the present invention; Figure 2 This is a schematic diagram of the risk area identification process for cable-stayed bridges according to the present invention; Figure 3 This is a schematic diagram of the vehicle similarity comparison process of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1: This invention proposes a vehicle crosswind recognition and early warning system for cable-stayed bridges, the structure of which is as follows: Figure 1 As shown, it includes: a risk identification module, a risk calibration module, a traffic comparison module, and a communication guidance module.

[0022] The risk identification module obtains crosswind environmental data based on wind and vibration data of the cable-stayed bridge deck, cable array, and towers; it then analyzes the crosswind environmental data to generate a traffic risk map of the cable-stayed bridge, including risk areas and risk levels.

[0023] The process of identifying the traffic risk map of the cable-stayed bridge includes: Using the bridge deck of the cable-stayed bridge as a one-dimensional coordinate axis, wind force data and vibration data from the crosswind environment data are projected onto the bridge deck, and the identified wind force magnitude, wind turbulence degree, cable vibration frequency and cable vibration amplitude are used as standard environmental data. Rainfall is monitored by precipitation sensing units, and standard environmental data is used to identify inherent risk areas of the cable-stayed bridge. These inherent risks include wind and rain vibration and wake vibration. For non-inherent risk areas, clustering is performed based on wind force, wind turbulence, cable frequency, and cable amplitude. Risk regions are obtained based on the clustering results of inherent risk regions and non-inherent risk regions; the identification process of the risk regions is as follows: Figure 2 As shown, the risk level is identified based on standard environmental data and inherent risk types in the risk area, resulting in a traffic risk map for the cable-stayed bridge.

[0024] Considering the interaction between the bridge towers, cable arrays, and crosswinds, the risks at different locations on a cable-stayed bridge are not uniform or homogeneous. Therefore, spatial projection is used to transform discrete sensor data into environmental features continuously distributed along the longitudinal direction of the bridge, facilitating regional division. Secondly, two types of risk causes are distinguished: one is the inherent risk known from aerodynamic and structural dynamics theories and inevitably occurring under specific conditions (such as wind-rain coupling), such as wind-rain-induced vibration; the other is the non-inherent risk discovered from unknown real-time data patterns through data-driven clustering algorithms.

[0025] Specifically, anemometers, wind vanes, and vibration sensors are deployed along the bridge deck, towers, and cables of the cable-stayed bridge; all sensor data are mapped onto a one-dimensional coordinate system with the bridge deck centerline as the reference; for example, 0 meters at the bridgehead and 1200 meters at the bridge tail.

[0026] Meanwhile, precipitation sensors are installed on the bridge. When rainfall is detected, special attention is paid to areas near the bridge towers and other areas known to be prone to wind-rain vibration under wind-rain coupling, and these areas are directly marked as inherent risk areas. For other areas of the bridge, clustering algorithms such as K-Means are used to group segments with similar environmental data, such as wind speed, wind turbulence, and cable amplitude, into one category. For example, areas with high wind speed and high cable amplitude are clustered as high-risk, while areas with stable wind speed are classified as low-risk. Finally, combining the inherent risk areas and the clustering results, a risk map is generated, labeled with different risk levels and risk types.

[0027] Preferably, wind-induced vibration refers to a large-amplitude, low-frequency vibration generated in the cable-stayed bridge under the combined action of wind and rain. It is not caused by extreme winds, but usually occurs under conditions of moderate wind speed and moderate to heavy rain. Its core mechanism is as follows: Changing the aerodynamic shape: When there is no rain, the cross-section of the cable is close to a circle, and the airflow flows smoothly. When it rains, rainwater will form several water ridges on the top and bottom of the windward and leeward sides of the cable under the action of wind.

[0028] Formation of asymmetrical cross-section: The dynamically formed water ridges, especially the upper and lower ones, temporarily change the circular cross-section of the cable, making it an irregular, asymmetrical temporary aerodynamic shape.

[0029] Inducing aerodynamic instability: When wind flows through this temporary asymmetric circular cross section at a specific angle and speed, it will generate unstable lift and drag, forming aerodynamic negative damping effect; under these circumstances, the force of the air no longer hinders the vibration, but instead promotes the vibration, continuously injecting energy into the cable.

[0030] Resonance amplification: If the frequency of this aerodynamic change is close to or coincides with a certain natural vibration frequency (especially a low-order frequency) of the cable itself, resonance will occur, causing the vibration amplitude to increase rapidly in a short period of time. Sometimes the cable can be seen to swing violently like a string.

[0031] The identification process of wind and rain-induced vibration includes: Calculate the overthreshold excitation energy; establish a standard vibration energy spectrum baseline under no-rainfall conditions for the wind speed range; when rainfall and wind speed conditions meet the wind and rain excitation triggering conditions, collect the vibration signal of the cable in real time and calculate its real-time vibration energy spectrum; subtract the real-time energy spectrum from the standard vibration energy spectrum baseline of the corresponding wind speed range to obtain the net vibration energy spectrum; integrate the net vibration energy spectrum near the low-order natural frequency of the cable to obtain the quantized overthreshold excitation energy.

[0032] Vibration mode assessment: Based on the structural parameters of the cable, the visual saliency weights of the first three vibration modes are pre-defined; among them, the low-order modes (such as first-order symmetrical or antisymmetrical vibration, the form of large-amplitude swing of the entire cable) have the strongest visual impact and attention distraction effect on the driver due to their large amplitude and low frequency, so they are given the highest weight.

[0033] A comprehensive risk index is generated; the super-threshold excitation energy and the visual impact weight are weighted and fused, and the vibration duration is taken into account to generate a dimensionless wind and rain excitation comprehensive risk index.

[0034] This invention establishes a standard vibration energy spectrum baseline under rainless conditions and calculates the overthreshold excitation energy, which can accurately isolate the net increase in vibration caused by wind and rain coupling effect, effectively eliminating the interference of background noise such as traffic load and conventional wind vibration, and significantly improving the accuracy of identification. At the same time, it creatively introduces a visual impact assessment based on vibration mode, which incorporates the potential impact on driver psychology and attention distraction into risk consideration, so that the final comprehensive risk index not only reflects the physical state of the structure, but also directly relates to the core factors of traffic safety.

[0035] The process of obtaining the standard vibration energy spectrum baseline is as follows: collecting and learning the vibration acceleration data of the cable-stayed bridge under different wind speed levels and no rainfall conditions, and establishing a standard vibration energy spectrum baseline for each wind speed range through statistical averaging and filtering; the standard vibration energy spectrum baseline characterizes the background vibration caused by traffic load and conventional wind effects.

[0036] Furthermore, vibration modes are the inherent dynamic characteristics of elastic structures, defined sequentially from low to high natural frequencies as first-order, second-order, third-order, etc. The first-order mode (fundamental mode) is the most easily excited and fundamental vibrational form of the structure; its mode shape is a global vibration without an intermediate fixed point (node). The second-order mode has a slightly higher frequency, and its mode shape contains one node, dividing the structure into two segments that move in opposite directions. The third-order vibration mode has an even higher frequency, and its mode shape contains two nodes, dividing the structure into three segments that move in opposite directions.

[0037] For structures like bridge decks that have a central axis of symmetry, their first-order modes typically manifest in two basic forms: First-order symmetrical vibration: refers to the vibration deformation of a structure that is mirror-symmetrical about its axis of symmetry. For bridge decks, this usually manifests as a uniform, overall up-and-down bending vibration.

[0038] First-order antisymmetric vibration: refers to the vibration deformation of a structure that is point-symmetric about its axis of symmetry, meaning that the directions of motion on both sides of the axis of symmetry are exactly opposite. For a bridge deck, this manifests as torsional vibration, where one side moves upward while the other side moves downward.

[0039] Wake gallop is an aerodynamic disturbance phenomenon that occurs between parallel or nearly parallel structures. It occurs under strong wind conditions without rain and is highly sensitive to wind direction (usually requiring the wind direction to be basically perpendicular to the plane of the cable-stayed bridge array). It is a group phenomenon, meaning that the upstream cables are stable while the downstream cables vibrate violently. In cable-stayed bridges, it mainly manifests as the violent vibration of the downstream cables in the wake of the upstream cables in the cable-stayed bridge array.

[0040] Its core mechanism is as follows: Formation of a wake region: When the airflow passes over the front row of cable stays, a wake region with low speed, low pressure and full of vortices will be formed behind it. The airflow in the wake region is extremely unstable; the rear row of cable stays is located in the unstable wake region.

[0041] Aerodynamic negative damping effect: There is a sharp wind speed gradient in the wake region (i.e., the wind speed changes drastically in a very small space); when the rear cable begins to vibrate slightly, it will move back and forth between the high-speed and low-speed regions in the wake region, so that the aerodynamic force on the cable is in the same direction as its movement, thus forming a strong aerodynamic negative damping effect, continuously absorbing energy from the wind.

[0042] Self-excited vibration: When the energy absorbed from the wind exceeds the energy that the cable's own structural damping can dissipate, the vibration will self-excite and continuously increase, eventually forming a large-amplitude gallop.

[0043] The process for identifying wake gallop includes: The cable-stayed array is abstracted as a two-dimensional grid. The root mean square value of the vibration of each cable in the array is collected and used as the heat value of the grid node to generate a real-time cable array amplitude heat map. Based on the real-time wind direction, the upstream excitation cable and the downstream response cable in the wake region are identified; the wake interference gain factor is obtained by calculating the ratio of the amplitude between each pair of adjacent upstream and downstream cables; the average wake interference gain factor, which characterizes the wake interference intensity, is obtained by averaging the gain factors of all affected cable pairs. For high-amplitude clusters identified on the heat map, the vibration time-domain signals of all stay cables within the cluster are extracted; the consistency of vibration pace is quantified by calculating the cross-correlation coefficient between any two cable signals; and the cluster synchronization index is obtained by averaging the cross-correlation coefficients of all cable pairs within the cluster. By combining the average wake interference gain factor, the cluster synchronization index, and the number of affected cables, the wake gallop cluster risk index is calculated to determine the wake gallop phenomenon and its degree of danger.

[0044] This invention transforms abstract vibration data into intuitive spatial distribution images, enabling visualization of the range and intensity of high-risk vibration clusters. Furthermore, by using the wake interference gain factor cluster synchronization index, it identifies vibration amplitude, quantifies the aerodynamic excitation intensity of upstream cables on downstream cables, and assesses the coordinated consistency of cable group vibrations. It accurately captures the core hazardous characteristics of wake galloping phenomena occurring in clusters and exhibiting synchronous resonance, achieving precise diagnosis of this complex aerodynamic interference phenomenon and providing a reliable basis for developing lane-level, refined early warning strategies.

[0045] Among them, the real-time cable array amplitude heat map can visualize the spatial distribution, intensity and range of influence of vibration in the entire cable group; the closer the cluster synchronization index is to 1, the more synchronized the vibration of the cable group is, the more concentrated the energy is, and the stronger the potential for damage.

[0046] Furthermore, risk level identification is accomplished through a pre-trained decision tree classification model. This model uses wind force, wind turbulence, cable vibration frequency, and cable vibration amplitude as input features, and outputs one of three risk levels: 'high', 'medium', and 'low'. Additionally, the model's input features can also include a comprehensive wind and rain-induced vibration risk index and a wake-induced vibration cluster risk index.

[0047] This invention combines theoretical and data-driven approaches, enabling the generation of risk maps to have both a solid physical foundation and the ability to adapt to sudden and unknown situations; thus significantly improving the accuracy and scientific rigor of risk identification. Compared to treating the entire bridge as a single risk area, this invention achieves a refined depiction of risks, accurately identifying specific hazardous sections and their causes; making subsequent early warnings more targeted.

[0048] The risk assessment module identifies the sentry vehicles that are currently passing through; acquires vehicle information and crosswind data of the sentry vehicles, and collects their travel trajectory data when crossing the cable-stayed bridge; and quantifies the trajectory deviation points and trajectory deviation degrees in the risk area using the travel trajectory data of the sentry vehicles.

[0049] For example, an integrated device is installed on the gantry or roadside pole at the entrance of a cable-stayed bridge. When a truck acting as a sentry vehicle approaches, a top-mounted camera (visual recognition unit) captures multi-angle images of it, constructing its three-dimensional outline (length, width, height, etc.) using stereo vision algorithms or LiDAR. As the vehicle passes over an array of piezoelectric or quartz sensors (dynamic weighbridge unit) embedded beneath the road surface, its total weight and axle load can be measured in real time. Simultaneously, a roadside communication unit establishes communication with the truck's onboard unit via the C-V2X protocol to obtain precise real-time vehicle status data such as speed and acceleration; these data are then integrated to form a complete digital profile of the sentry vehicle.

[0050] This invention constructs an automated, high-precision vehicle feature acquisition system, which solves the problems of delay and inaccuracy caused by manual recording or reliance on vehicle self-reporting information, and realizes real-time and accurate capture of core parameters affecting crosswind stability. This standardized data acquisition entry is the cornerstone of the entire communication system, ensuring the quality and reliability of input data, thereby improving the accuracy of all subsequent analyses and warnings.

[0051] The risk assessment module analyzes the travel trajectory data of the sentry vehicles using a trajectory stability quantification model. The trajectory stability quantification model acquires the vehicle's driving trajectory and heading angle data from the travel trajectory data and performs the following analysis: Position stability analysis: Calculate the standard deviation of the vehicle's lateral position relative to the lane centerline to quantify the drift amplitude within the risk area; Attitude stability analysis: Identify the rate of change of the vehicle's heading angle data to capture the rotational trend caused by crosswind impact; Handling oscillation analysis: Perform spectral analysis on the time series of the vehicle's lateral displacement, extract the spectral power density within a preset frequency range, and identify handling oscillation behavior caused by overcorrection; Using the sentry vehicle's trajectory before entering the cable-stayed bridge as a benchmark, weight multiple analysis indicators to obtain the trajectory deviation; Obtain time-series data of trajectory offset, and record the geographical locations where the time-series data reaches a maximum value as trajectory offset points.

[0052] For example, continuous trajectory data (latitude and longitude sequence) and heading angle data of a sentry vehicle crossing a bridge were acquired using high-precision GPS or roadside sensing devices. Further, trajectory data of the vehicle within 1 kilometer of a straight road section before crossing the bridge were retrieved to quantify the standard deviation of its lateral position relative to the lane centerline, the rate of change of its heading angle data, and the rotational trend caused by crosswind impact, serving as a benchmark. For instance, its lateral position standard deviation was calculated to be 0.1 meters, quantifying the benchmark stability of normal driving.

[0053] When the vehicle enters a high-risk area on the bridge, the state changes are analyzed; positional stability is calculated, and the lateral standard deviation of its trajectory relative to the lane centerline increases to 0.4 meters; attitude stability is monitored, and its heading angle shows multiple rapid changes of 1-2 degrees; control oscillation analysis: a fast Fourier transform is performed on the vehicle's lateral displacement data, and significant energy is found in the 0.5-2Hz frequency band (typical human-vehicle closed-loop control oscillation frequency), indicating that the driver is overcorrecting the steering wheel.

[0054] Finally, the changes in these three indicators are weighted and summed relative to the benchmark to obtain a comprehensive trajectory deviation score, such as 0.75. The obtained trajectory deviation scores are arranged according to the one-dimensional coordinate axis of the cable-stayed bridge deck, and their maximum values ​​are identified as trajectory deviation points.

[0055] This invention achieves refined and multi-dimensional quantification of traffic risks, transforming a vague concept of danger into a calculable trajectory deviation composed of multiple physically meaningful indicators. It provides high-quality label data and empirical truth values ​​for subsequent risk prediction, enabling the system to not only know that there is risk, but also the degree and manifestation of the risk, greatly improving the scientificity and usability of risk labeling.

[0056] The traffic comparison module identifies target vehicles about to enter the cable-stayed bridge; based on the target vehicle's weight and outline, it identifies the vehicle similarity to the sentry vehicle; and based on real-time crosswind data and traffic crosswind data, it obtains the crosswind difference.

[0057] The traffic comparison module uses a multi-dimensional similarity matching algorithm based on dynamic aerodynamic features to identify the vehicle similarity between the target vehicle and the sentry vehicle. The vehicle's three-dimensional geometric contour is obtained by a visual recognition unit, and combined with the real-time crosswind direction vector, the three-dimensional geometric contour is projected to determine the vehicle's effective projected side contour under the corresponding wind field. A comparative analysis is performed on the effective projected side profile to extract the effective side area associated with the crosswind force and the profile shape factor characterizing the aerodynamic properties. Vehicle similarity is obtained by normalizing and weighting the effective lateral area, contour shape factor, vehicle weight, and average vehicle speed of the target vehicle and the sentry vehicle.

[0058] The process for comparing the vehicle similarity between the target vehicle and the sentry vehicle is as follows: Figure 3 As shown.

[0059] For example, when a van (target vehicle) is about to cross the bridge, its three-dimensional profile is captured by the visual recognition unit at the entrance; at this time, the crosswind direction on the bridge is 75 degrees from the bridge axis.

[0060] The 3D model of the truck is retrieved and projected onto a plane perpendicular to the 75-degree wind direction to obtain the effective projected side profile. The effective side area is calculated to be 25 square meters from the effective projected side profile, and its shape is analyzed to obtain a profile shape factor (e.g., 0.8, representing the quality of its aerodynamic shape). Then, a search is performed in the historical sentry vehicle database to find a sentry truck that passed under similar wind conditions, with an effective side area of ​​24.5 square meters, a shape factor of 0.82, and a similar weight. By comparing the four dimensions of effective side area, profile shape factor, vehicle weight, and average speed using normalized weighted averages, the similarity between the target vehicle and this sentry vehicle is calculated.

[0061] The process of obtaining the contour shape factor includes: The asymmetric moment index is determined by calculating the normalized longitudinal distance between the geometric centroid of the profile and the estimated pressure center point; the estimated pressure center point is obtained by integrating the area moment by applying a higher weight to the front region of the profile. This index aims to quantify the inherent tendency of the vehicle to generate yaw moment under crosswinds due to its asymmetric shape.

[0062] The profile turbulence index is determined by calculating a specific ratio of the profile perimeter to the area (such as the iso-perimeter ratio) and combining it with the analysis of the profile edge curvature. The profile turbulence index is designed to quantify the smoothness of the profile. A profile that is closer to a rectangle and has more sharp corners will produce stronger airflow separation and turbulence, thus obtaining a higher turbulence index.

[0063] The asymmetric moment index and the profile turbulence index are weighted and fused to form a profile shape factor that can comprehensively reflect the vehicle's attitude stability and force characteristics in crosswinds.

[0064] This invention directly correlates the asymmetric moment index with whether a vehicle is prone to nose-swinging or tail-wagging in crosswinds, and correlates the contour turbulence index with whether the lateral forces acting on the vehicle are stable. Furthermore, it obtains the contour shape factor by weighted fusion of the asymmetric moment index and the contour turbulence index, which can more accurately match the target vehicle with a sentry vehicle that is truly comparable, thereby significantly improving the prediction accuracy and reliability of the entire early warning system.

[0065] This invention introduces an effective projected side profile, which can dynamically calculate the vehicle's true windward surface based on real-time wind direction, making it more accurate than using a fixed side view area. At the same time, the profile shape factor further considers the streamlined nature of the vehicle's shape. By comprehensively comparing these physical quantities that directly affect the vehicle's crosswind stability, the physical meaning and accuracy of the similarity score are ensured, greatly improving the accuracy and physical relevance of vehicle matching.

[0066] The process by which the traffic comparison module identifies and obtains the crosswind difference includes: Crosswind data is abstracted into state vectors, the dimensions of which include wind force data and vibration data acquired from the bridge deck, cable array, and bridge tower of the cable-stayed bridge; state vectors are established for the crosswind data during the passage of the sentry vehicle and the current real-time crosswind data, respectively, to obtain the passage state vector and the real-time state vector; The crosswind difference is quantified based on the normalized distance between the passage state vector and the real-time state vector in the multidimensional space.

[0067] A sensor network covering the bridge deck, stay cables, and towers, comprising 50 measurement points, was deployed. Each point measures wind speed and amplitude, forming a 100-dimensional feature space. At a specific historical moment, when a sentry vehicle passes through, the system records the data at that time. For the process-time data generated by the sentry vehicle throughout its passage, the monitoring sequences of each sensor (such as anemometers and vibration meters) are deeply processed over the entire passage period. From this data, the mean representing the average state, the standard deviation representing environmental disturbance and volatility, and the maximum value representing the most extreme impact experienced are extracted to construct a passage state vector.

[0068] Accordingly, to ensure the fairness of the comparison, instead of using only the instantaneous data slices before the target vehicle goes onto the bridge, we collect near-real-time data of a stable period (e.g., 30 seconds) before it goes onto the bridge, and perform the same statistical operations as the sentry vehicle to obtain the mean, standard deviation and maximum value, and construct the real-time state vector.

[0069] To calculate the difference in crosswind conditions between two time points, the system first normalizes the two vectors and then calculates the Euclidean distance. This Euclidean distance, such as 0.15, is defined as the crosswind difference, which intuitively represents the overall degree of change in the current wind field environment compared to when the sentry vehicle passed.

[0070] Crosswind variability can capture the overall changes in wind field structure, not just changes in amplitude. For example, even if the average wind speed remains constant, changes in wind turbulence or distribution on the bridge surface will alter the state vector and increase the calculated variability.

[0071] This invention constructs a high-dimensional state vector, representing a snapshot of the wind environment of the entire bridge at a given moment as a point in a high-dimensional space. By comparing the environmental differences between two different moments, it intuitively transforms the problem into the distance between two points in the computational space. It comprehensively considers all relevant factors, such as wind speed, wind distribution, and the overall changes in bridge vibration response, making it more comprehensive and robust than comparing only a few isolated indicators. This allows risk prediction to take into account more dimensional environmental changes, thus making the early warning system more adaptable to complex weather conditions and more accurate in its judgments.

[0072] The communication guidance module identifies the target vehicle's lateral deviation risk index within the risk area based on risk level, vehicle similarity, and crosswind difference, combined with the sentry vehicle's travel speed, trajectory deviation point, and trajectory deviation degree; and conducts early warning communication based on the lateral deviation risk index.

[0073] The communication guidance module determines the lateral deviation risk index of the target vehicle through a weighted empirical risk projection model; the identification logic of the weighted empirical risk projection model is as follows: Sentry vehicles that meet a preset threshold in terms of vehicle similarity are selected based on vehicle similarity. Based on the highest value of the trajectory deviation of the selected sentry vehicles in the risk area and the risk level of the corresponding risk area, the basic risk value is identified; the basic risk value is then weighted twice based on vehicle similarity and crosswind difference to calculate the expected risk projected onto the target vehicle, namely the lateral deviation risk index.

[0074] For example, a target truck is about to cross a bridge; all sentry vehicles with a similarity of more than 90% to the target truck are selected from the database, totaling 10 vehicles; the historical records of these 10 vehicles when passing through the high-risk area ahead are reviewed, and it is found that their maximum trajectory deviation ranges from 0.3 to 0.8. The basic risk value is identified by combining the risk level of the risk area with the maximum value of 0.8; for example, a non-linear regional risk weight is set according to the risk level, and the maximum trajectory deviation is weighted according to the regional risk weight to obtain the basic risk value; for example, the basic risk value is obtained by multiplying the highest trajectory deviation value by the risk level weight.

[0075] At this point, the system calculates that the similarity between the target truck and the sentry vehicle B that generated the maximum value is 92%, while the crosswind difference between the current crosswind environment and that time is 0.2. Based on the vehicle similarity and crosswind difference, a dual weighting is initiated to obtain the lateral deviation risk index.

[0076] This invention achieves personalized and dynamic accurate prediction of target vehicle risks. Instead of simply applying historical data, it uses a scientifically weighted model to consider both individual vehicle differences and real-time environmental changes, resulting in a final risk index that closely reflects the actual situation the target vehicle will face. This precise risk assessment is key to effective and reliable early warning, significantly enhancing the intelligence and practical value of the entire communication system.

[0077] The warning communication of the communication guidance module is carried out by the roadside communication unit through the vehicle-road cooperative protocol, which sends the warning information to the on-board unit of the target vehicle. The warning information is classified according to the magnitude of the lateral deviation risk index, including low risk, medium risk and high risk.

[0078] When the system's internal warning logic determines that the lateral drift risk index falls into the high-risk range (e.g., >0.8), the roadside communication unit deployed at the bridgehead immediately sends a targeted, high-priority warning message to the onboard unit of the target vehicle via the C-V2X PC5 interface (or other vehicle-to-everything (V2X) protocol). For example: "Warning! Strong crosswind section ahead 200 meters. Your vehicle has an extremely high risk of lateral drift. Please immediately reduce speed to below 60 km / h, grip the steering wheel firmly, and stay in the center of the lane!" If the calculated risk index is 0.4, which falls under medium risk, the message might be: "Caution, moderate crosswind ahead on the bridge. Please drive with caution."

[0079] This solution ensures that all complex calculations and analyses are translated into effective interventions for the driver through a closed-loop system execution process. By using targeted, tiered, and low-latency communication, intangible risk indices are translated into specific operational suggestions that the driver can understand and execute. This not only enhances the driver's proactive safety but also allows the value of the entire system to be fully realized, truly achieving a complete safety assurance chain from environmental perception and risk analysis to communication guidance.

[0080] Example 2: This invention proposes a vehicle crosswind identification and early warning system for cable-stayed bridges, which is applied to the field of crosswind identification and early warning for cable-stayed bridges. Specifically, it includes a risk identification module, a risk calibration module, a traffic comparison module, and a communication guidance module.

[0081] The risk identification module obtains crosswind environmental data based on wind and vibration data of the cable-stayed bridge deck, cable array, and towers; it then analyzes the crosswind environmental data to generate a traffic risk map of the cable-stayed bridge, including risk areas and risk levels. The risk assessment module identifies the sentry vehicles that are currently passing through; acquires vehicle information and crosswind data of the sentry vehicles, and collects their travel trajectory data when crossing the cable-stayed bridge; and quantifies the trajectory deviation points and trajectory deviation degrees in the risk area using the travel trajectory data of the sentry vehicles. The traffic comparison module identifies target vehicles about to enter the cable-stayed bridge; based on the target vehicle's weight and outline, it identifies the vehicle similarity to the sentry vehicle; and based on real-time crosswind data and traffic crosswind data, it obtains the crosswind difference. The communication guidance module identifies the target vehicle's lateral deviation risk index within the risk area based on risk level, vehicle similarity, and crosswind difference, combined with the sentry vehicle's travel speed, trajectory deviation point, and trajectory deviation degree; and conducts early warning communication based on the lateral deviation risk index.

[0082] Preferably, a cross-sea cable-stayed bridge is used as the application scenario. The cross-sea cable-stayed bridge is 2,500 meters long, and its bridge towers, cable arrays and bridge deck are equipped with distributed fiber optic vibration sensors and ultrasonic anemometers.

[0083] When dividing non-inherent risk areas, the system specifically adopted the K-means clustering algorithm. Before executing the algorithm, the collected data in four dimensions—wind strength, wind turbulence, cable vibration frequency, and cable amplitude—were Z-score standardized to eliminate the influence of dimensions.

[0084] Risk Map Generation: The system monitored an average crosswind speed of 18 m / s on the bridge, indicating high wind turbulence near the bridge towers. Based on the aforementioned clustering and the identification of wake oscillations at the bridge towers, the system updated the traffic risk map for the cable-stayed bridge. Specifically, the road section downstream of bridge tower P2, located between 1500 and 1800 meters on the bridge, was identified as a Level 3 (high) risk area, with wake oscillations as the primary risk. The remaining road sections were classified as Level 1 or Level 2 risk areas.

[0085] A van is identified as a sentry vehicle. When it crosses the bridge, its information is obtained from the integrated vehicle sensing unit at the entrance as follows: Visual recognition unit: Determines its three-dimensional geometric outline as a regular cuboid with a length of 9.6 meters, a width of 2.5 meters, and a height of 4.0 meters; Dynamic truck scale unit: The total weight of the vehicle was measured to be 18.0 tons; Roadside communication unit: The speed of the road is 75 km / h.

[0086] For example, when a sentry vehicle passes through a high-risk area of ​​1500-1800 meters, the trajectory stability quantification model collects and analyzes its travel trajectory data, obtaining the following indicators: Position stability: The standard deviation of the lateral position of the trajectory relative to the lane centerline is 0.45 meters.

[0087] Attitude stability: The maximum rate of change of heading angle data is 1.8 degrees / second.

[0088] Manipulated oscillation analysis: Spectral analysis of the lateral displacement shows that the spectral power density integral in the frequency range of 0.5-2Hz is 0.12m² / Hz.

[0089] The trajectory deviation is calculated by comparing the above three indicators with the baseline values ​​before the vehicle went onto the bridge and normalizing them.

[0090] Calculations show that the trajectory deviation of Sentinel Vehicle A in the risk area is 0.78, and its trajectory deviation point is located 1650 meters from the bridge. The system will store the complete record containing vehicle information, crosswind data, trajectory deviation, and deviation point into the database.

[0091] A high-sided truck was detected about to enter the bridge and was identified as the target vehicle. The system obtained the target vehicle's weight as 16.5 tons and its outline as 12.0 meters long, 2.5 meters wide, and 4.2 meters high (including the sideboards).

[0092] Dynamic aerodynamic feature extraction: The current real-time crosswind direction is at an 80-degree angle to the bridge axis. The system performs projection calculations on the 3D contours of the target vehicle and sentry vehicles in the database to obtain the effective projected side contours. Based on this, the effective side area and contour shape factor are calculated.

[0093] Sentry vehicle: effective lateral area 39.1㎡, profile shape factor 0.85.

[0094] Target vehicle: Effective side area 50.9㎡, profile shape factor 0.75 (the aerodynamic shape is worse due to gaps in the side panels).

[0095] Weighted similarity calculation: First, the data for the four dimensions—effective lateral area, contour shape factor, vehicle weight, and average speed—are subjected to min-max normalization. Then, the similarity is calculated using a weighted formula. The calculated vehicle similarity between target vehicle B and sentry vehicle A is 0.92.

[0096] The system constructs multidimensional state vectors from crosswind environmental data at the time the sentry vehicle passes and at the current moment, including wind and vibration data acquired from the bridge deck, cable-stayed array, and bridge tower. By calculating the Euclidean distance between the two state vectors in the normalized multidimensional space, the crosswind difference is quantified to be 0.25.

[0097] The system selects van sentry vehicles as the best reference based on a preset threshold of vehicle similarity > 0.9. It calculates the basic risk value based on the sentry vehicle's highest trajectory deviation (0.78) and risk area level (Level 3-High). The basic risk value is then weighted twice based on vehicle similarity and crosswind difference to calculate the final lateral deviation risk index.

[0098] The system's internal risk levels are divided into: low risk (<1.5), medium risk (1.5-2.5), and high risk (>2.5).

[0099] When the calculated lateral drift risk index indicates a high risk, the communication guidance module immediately sends the following warning message to the onboard unit of target vehicle B via the vehicle-to-infrastructure (V2I) protocol: "[High Risk Warning] You are entering the bridge tower wake zone 1.5 kilometers ahead. Your vehicle has an extremely high risk of lateral drift. Please immediately reduce your speed to 60 km / h, grip the steering wheel firmly, and stay in the center of your lane!"

[0100] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A vehicle crosswind recognition and early warning system for cable-stayed bridges, characterized in that, include: The risk identification module obtains crosswind environmental data based on wind and vibration data of the cable-stayed bridge deck, cable array, and towers; it then analyzes the crosswind environmental data to generate a traffic risk map of the cable-stayed bridge, including risk areas and risk levels. The risk assessment module identifies the sentry vehicles that are currently passing through; acquires vehicle information and crosswind data of the sentry vehicles, and collects their travel trajectory data when crossing the cable-stayed bridge; and quantifies the trajectory deviation points and trajectory deviation degrees in the risk area using the travel trajectory data of the sentry vehicles. The traffic comparison module identifies target vehicles about to enter the cable-stayed bridge; based on the target vehicle's weight and outline, it identifies the vehicle similarity to the sentry vehicle; and based on real-time crosswind data and traffic crosswind data, it obtains the crosswind difference. The communication guidance module identifies the target vehicle's lateral displacement risk index within the risk area based on risk level, vehicle similarity, and crosswind difference, combined with the sentry vehicle's travel speed, trajectory deviation point, and trajectory deviation. Early warning communication is conducted based on the aforementioned lateral offset risk index.

2. The vehicle crosswind recognition and early warning system for cable-stayed bridges according to claim 1, characterized in that: The process of identifying the traffic risk map of the cable-stayed bridge includes: Using the bridge deck of the cable-stayed bridge as a one-dimensional coordinate axis, wind force data and vibration data from the crosswind environment data are projected onto the bridge deck, and the identified wind force magnitude, wind turbulence degree, cable vibration frequency and cable vibration amplitude are used as standard environmental data. Rainfall is monitored by precipitation sensing units, and standard environmental data is used to identify inherent risk areas of the cable-stayed bridge. These inherent risks include wind and rain vibration and wake vibration. For non-inherent risk areas, clustering is performed based on wind force, wind turbulence, cable frequency, and cable amplitude. Based on the clustering results of inherent risk areas and non-inherent risk areas, risk areas are obtained; based on the standard environmental data and inherent risk types in the risk areas, the risk level is identified, resulting in a traffic risk map for the cable-stayed bridge.

3. A vehicle crosswind recognition and early warning system for cable-stayed bridges according to claim 1, characterized in that: The risk assessment module is equipped with an integrated vehicle sensing unit on the cable-stayed bridge to obtain vehicle information of the sentry vehicles. The integrated vehicle perception unit includes: a visual recognition unit, a dynamic weighbridge unit, and a roadside communication unit; The visual recognition unit determines the three-dimensional geometric contour of the vehicle through images captured by a camera; the dynamic weighbridge unit measures the weight of the vehicle as it passes; and the roadside communication unit conducts vehicle-road cooperative communication with the sentry vehicle to obtain vehicle status data, including traffic speed.

4. A vehicle crosswind recognition and early warning system for cable-stayed bridges according to claim 1, characterized in that: The risk assessment module analyzes the travel trajectory data of the sentry vehicles using a trajectory stability quantification model. The trajectory stability quantification model acquires the vehicle's driving trajectory and heading angle data from the travel trajectory data and performs the following analysis: Position stability analysis: Calculate the standard deviation of the vehicle's trajectory relative to the lane centerline to quantify the drift amplitude within the risk area; Attitude stability analysis: Identify the rate of change of vehicle heading angle data and capture the rotational trend of the vehicle caused by crosswind impact; Control oscillation analysis: Perform spectral analysis on the time series of vehicle lateral displacement, extract the spectral power density within a preset frequency range, and identify control oscillation behavior caused by overcorrection; Using the travel trajectory of the sentry vehicle before entering the cable-stayed bridge as a benchmark, multiple analytical indicators were weighted to obtain the trajectory deviation. Obtain time-series data of trajectory offset, and record the geographical locations where the time-series data reaches a maximum value as trajectory offset points.

5. A vehicle crosswind recognition and early warning system for cable-stayed bridges according to claim 3, characterized in that: The traffic comparison module uses a multi-dimensional similarity matching algorithm based on dynamic aerodynamic features to identify the vehicle similarity between the target vehicle and the sentry vehicle. The vehicle's three-dimensional geometric contour is obtained by a visual recognition unit, and combined with the real-time crosswind direction vector, the three-dimensional geometric contour is projected to determine the vehicle's effective projected side contour under the corresponding wind field. A comparative analysis is performed on the effective projected side profile to extract the effective side area associated with the crosswind force and the profile shape factor characterizing the aerodynamic properties. Vehicle similarity is obtained by normalizing and weighting the effective lateral area, contour shape factor, vehicle weight, and average vehicle speed of the target vehicle and the sentry vehicle.

6. A vehicle crosswind recognition and early warning system for cable-stayed bridges according to claim 1, characterized in that: The process by which the traffic comparison module identifies and obtains the crosswind difference includes: Crosswind data is abstracted into state vectors, the dimensions of which include wind force data and vibration data acquired from the bridge deck, cable array, and bridge tower of the cable-stayed bridge; state vectors are established for the crosswind data during the passage of the sentry vehicle and the current real-time crosswind data, respectively, to obtain the passage state vector and the real-time state vector; The crosswind difference is quantified based on the normalized distance between the passage state vector and the real-time state vector in the multidimensional space.

7. A vehicle crosswind recognition and early warning system for cable-stayed bridges according to claim 1, characterized in that: The communication guidance module determines the lateral deviation risk index of the target vehicle through a weighted empirical risk projection model; the identification logic of the weighted empirical risk projection model is as follows: Sentry vehicles that meet a preset threshold in terms of vehicle similarity are selected based on vehicle similarity. Based on the highest value of the trajectory deviation of the selected sentry vehicles in the risk area and the risk level of the corresponding risk area, the basic risk value is identified and obtained. The base risk value is weighted by both vehicle similarity and crosswind difference to calculate the expected risk projected onto the target vehicle, namely the lateral offset risk index.

8. A vehicle crosswind recognition and early warning system for cable-stayed bridges according to claim 1, characterized in that: The warning communication of the communication guidance module is carried out by the roadside communication unit through the vehicle-road cooperative protocol, which sends the warning information to the on-board unit of the target vehicle. The warning information is classified according to the magnitude of the lateral deviation risk index, including low risk, medium risk and high risk.

Images