An unmanned aerial vehicle group-based traffic tunnel preventive monitoring and risk early warning system

Abstract

The application provides a kind of traffic tunnel preventive monitoring and risk early warning system based on unmanned aerial vehicle group, it is related to traffic tunnel safety monitoring technical field, including: the reference mark point is arranged as monitoring node on the inner surface of tunnel, the crack width, surface temperature and deformation of each node are collected by unmanned aerial vehicle group, multi-dimensional feature vector is constructed in combination with space coordinates and time stamp and the space-time weighted distance is calculated.According to the curvature of tunnel, the neighborhood search radius is adjusted, the risk area is identified by density clustering algorithm, and the abnormal node that has not been clustered is marked as isolated risk point.Risk level is divided based on the comparison of structure safety degree index and dynamic threshold, and the result is registered with BIM model to realize three-dimensional visualization.Trend warning is triggered for the area where safety degree index decreases monotonically for three consecutive times, realizing full-process automatic monitoring from data collection, intelligent analysis to risk warning, significantly improving the accuracy and timeliness of tunnel structure health monitoring.

Description

Technical Field

[0001] This invention relates to the field of traffic tunnel safety monitoring technology, specifically a traffic tunnel preventive monitoring and risk early warning system based on unmanned aerial vehicle (UAV) swarms. Background Technology

[0002] Current health monitoring of traffic tunnels mainly relies on two methods: manual inspection and fixed sensor networks, but both have significant limitations. Manual inspection uses simple methods such as visual inspection combined with tapping, which is inefficient and highly subjective, making it difficult to detect early defects such as millimeter-level cracks and back cavities, especially in high-altitude, concealed areas such as the arch and invert. Fixed sensors, such as strain gauges and convergence meters, can achieve continuous monitoring, but their deployment density is limited, their coverage is limited, and their installation and maintenance costs are high, making real-time monitoring of the entire tunnel cross-section impossible. Existing UAV inspection solutions mainly focus on two-dimensional image acquisition, lacking the ability to perform spatiotemporal fusion analysis of multi-source heterogeneous data, and cannot establish correlation assessment models between structural parameters, resulting in the difficulty in timely detection of potential risks such as micro-deformation and temperature stress.

[0003] A more prominent problem is that existing monitoring methods struggle to adapt to the dynamic changes in tunnel structures. On one hand, traditional clustering algorithms, such as K-means, use fixed neighborhood radii and cannot automatically adjust monitoring sensitivity based on tunnel curvature changes, easily leading to missed defects in curved sections. On the other hand, risk assessment models often use static thresholds, failing to consider factors such as material aging and the cumulative effects of environmental loads, resulting in distorted tunnel safety assessments in the later stages of operation. Furthermore, existing systems lack effective spatiotemporal correlation analysis mechanisms, failing to identify the development trends of progressive defects such as crack propagation and deformation accumulation, making it difficult to achieve true preventative early warning. These technological shortcomings severely restrict the accuracy and timeliness of tunnel maintenance, urgently requiring an innovative solution that can integrate multi-dimensional data, adapt to structural characteristics, and support dynamic assessment.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a traffic tunnel preventive monitoring and risk early warning system based on unmanned aerial vehicle (UAV) swarms, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A traffic tunnel preventive monitoring and risk early warning system based on unmanned aerial vehicle (UAV) swarms, specifically comprising:

[0008] The benchmark deployment and data acquisition module is used to uniformly deploy several benchmark marker points on the inner surface of the tunnel, and to assign each benchmark marker point as a monitoring node and assign it a unique spatial coordinate identifier. The drone swarm is used to collect the structural parameters, spatial coordinates and monitoring timestamps of each monitoring node in sequence according to a preset path. The structural parameters include crack width, surface temperature and deformation.

[0009] The feature modeling and distance definition module is used to characterize the local tunnel information at each monitoring node using multidimensional feature vectors based on the structural parameters, spatial coordinates and monitoring timestamps at each monitoring node, and to define the spatiotemporal weighted distance between any two nodes based on the multidimensional feature vectors.

[0010] The dynamic clustering and risk identification module is used to dynamically set the neighborhood search radius according to the tunnel curvature radius. For each monitoring node, the monitoring nodes within its neighborhood search radius are selected based on the spatiotemporal weighted distance. The monitoring nodes are divided into multiple clusters and isolated risk points based on the recursive expansion density clustering algorithm. The minimum number of neighborhood nodes is 5% of the total number of monitoring nodes.

[0011] The safety assessment and classification module is used to calculate the structural safety index for each cluster and isolated risk point, compare the structural safety index with the structural risk threshold, and classify the risk level based on the comparison results.

[0012] The 3D visualization and early warning module is used to construct the tunnel BIM model, register the clustering results with the BIM model, and use 3D visualization technology to mark the location information and corresponding risk level of each cluster and isolated risk point. For clusters or isolated risk points whose structural safety index shows a monotonically decreasing trend in three consecutive monitoring tests, a trend early warning signal is output.

[0013] Furthermore, several reference marker points are evenly distributed on the inner surface of the tunnel. Each reference marker point is used as a monitoring node and assigned a unique spatial coordinate identifier. The specific logic behind this is as follows:

[0014] Along the longitudinal axis of the tunnel, a reference marker is set up every 10 meters; four reference markers are evenly distributed around the tunnel cross section, located at the crown, left arch waist, right arch waist and invert arch respectively; for curved sections with a radius of curvature of less than 300 meters, the longitudinal interval of the reference markers is shortened to 5 meters.

[0015] Using the tunnel design axis as the reference coordinate system, the spatial coordinates of the monitoring nodes are represented as (x... k ,y k ,z k ), where: x k The mileage value along the tunnel axis is (y k ,z k) represents the coordinates in the local coordinate system of the cross section, determined by total station measurement, and k is the index of the monitoring node;

[0016] After the benchmark points are set up, the point cloud data is compared with the design coordinates by using UAV laser scanning, and the actual coordinate deviation of the benchmark points is corrected.

[0017] Furthermore, a swarm of drones is used to collect structural parameters, spatial coordinates, and monitoring timestamps at each monitoring node. The specific logic is as follows: after the drone hovers 1m away from the target reference mark, it simultaneously activates the optical camera to capture crack images, the infrared thermal imager to record surface temperature, and the lidar to scan deformation, and adds monitoring timestamps to the collected structural parameters; when an obstacle is detected, a prohibited area is defined with the current monitoring node as the starting point and the obstacle boundary extending 1m outward, and local path replanning is completed and flight commands are updated within 50ms to ensure a safe distance ≥2m and a single flight coverage length ≤500m;

[0018] An optical camera was used to vertically photograph the tunnel surface at the monitoring node, acquiring an RGB image with a resolution of 0.1 mm / pixel. An improved YOLOv5s algorithm was used to identify crack areas, and the crack width at the monitoring node was calculated using the following formula:

[0019]

[0020] In the formula, W is the crack width, and N is the crack width. pixels The number of pixels at the crack edge is denoted by , and cosθ is the angle between the camera optical axis and the crack direction, which is compensated by IMU attitude data. When multiple parallel cracks are identified, the gap width of the group of cracks with the smallest spacing is recorded as the crack width at that monitoring node.

[0021] The temperature matrix of a 3×3 pixel area at the monitoring node is obtained using an infrared thermal imager at a sampling frequency of 10Hz. After removing the highest and lowest temperature values ​​in the temperature matrix, the average value is taken as the surface temperature at the monitoring node.

[0022] A high-density point cloud is generated using LiDAR scanning, covering a 0.5m × 0.5m area centered on the monitoring node. The current point cloud is then registered with the tunnel BIM benchmark model using the IPC algorithm, and the deformation is calculated according to the following formula:

[0023]

[0024] In the formula, ΔD is the deformation, and P i Let P′ be the current point cloud coordinates. i N represents the coordinates of the corresponding points in the tunnel BIM reference model, N represents the number of valid corresponding point pairs that have been successfully registered between the current scanned point cloud and the reference model, and i is the index of the valid corresponding point pair.

[0025] Furthermore, based on the structural parameters, spatial coordinates, and monitoring timestamps at each monitoring node, a multi-dimensional feature vector is used to characterize the local tunnel information at each monitoring node, as shown in the following expression:

[0026] M k =(x k ,y k ,z k ,t k W k ,T k ,ΔD k ,)

[0027] In the formula, M k Let x be the multidimensional feature vector of monitoring node k, where k is the index of the monitoring node, (x... k ,y k ,z k ) represents the spatial coordinates of monitoring node k, and t represents the spatial coordinates of node k. k W represents the absolute time for data collection at monitoring node k. k To monitor the crack width at node k, T k To monitor the surface temperature at node k, ΔD k To monitor the deformation at node k;

[0028] The spatiotemporal weighted distance between any two nodes is defined based on multidimensional feature vectors, and the formula used is as follows:

[0029]

[0030] In the formula, BX k-t Let represent the spatiotemporal weighted distance between monitoring node k and monitoring node t, where t is the index of the monitoring node, and t ≠ k. k and s t Let α, β, and γ represent the structural evaluation indices of monitoring node k and monitoring node t, respectively. α, β, and γ are preset weights, where γ > α > β > 0, and satisfy α + β + γ = 1. ref The reference surface temperature is δ, which is the temperature weighting coefficient, and δ>0.

[0031] Furthermore, the method for dividing monitoring nodes into multiple clusters and isolated risk points is as follows:

[0032] All monitoring nodes are merged into a monitoring node set, and the initial state of each monitoring node in the set is marked as unvisited.

[0033] Randomly select an unvisited monitoring node p from the set of monitoring nodes, and dynamically set the neighborhood search radius corresponding to the monitoring node based on the radius of curvature at its location in the tunnel. The specific expression is as follows:

[0034]

[0035] In the formula, ∈ p To monitor the neighborhood search radius corresponding to node p, R p Let p be the radius of curvature at the location of the monitoring node p in the tunnel, and p be the index of the monitoring node in the set of monitoring nodes;

[0036] Based on the neighborhood search radius and spatiotemporal weighted distance, the set of monitoring nodes is traversed to determine the neighborhood N of monitoring node p. ∈ (p) Specifically: If there exists a monitoring node in the monitoring node set that is in an unvisited state, and the spatiotemporal weighted distance between this monitoring node and monitoring node p is less than the neighborhood search radius, then this node is included in the neighborhood N of monitoring node p. ∈ (p) in;

[0037] If the neighborhood of monitoring node p is N ∈ If the number of monitored nodes within (p) is less than the number of nodes in the minimum neighborhood, then unvisited monitored nodes are reselected from the set of monitored nodes; otherwise, monitored node p is used as the core node and a cluster C is created. p Clustering C p Consists of monitoring node p and its neighborhood N ∈ (p) Composition, and based on clustering C p The status of monitoring nodes within the monitoring node set is updated in real time, specifically by adding nodes from the monitoring node set to cluster C. p The status of the monitored nodes is changed to "visited", where the minimum number of neighboring nodes is 5% of the total number of monitored nodes;

[0038] For the neighborhood N ∈ For monitoring nodes within (p), their neighborhoods are determined using the same method. If the number of monitoring nodes in that neighborhood is less than the minimum number of neighborhood nodes, then that neighborhood is not included in cluster C. p Otherwise, include the neighborhood in cluster C. p The status of the monitoring nodes is updated in real time, and the process continues to recursively expand until cluster C is reached. p Unable to expand further;

[0039] Another unvisited monitoring node p is randomly selected from the set of monitoring nodes to construct a new cluster, until no cluster can be constructed. For a monitoring node that is not assigned to any cluster, if its structural evaluation index is less than the preset structural evaluation threshold, it is marked as an isolated risk point.

[0040] Furthermore, the mean of the structural evaluation index in each cluster is used as the structural safety index of that cluster. For isolated risk points, the structural evaluation index is calculated according to the above method and used as the structural safety index of that isolated risk point.

[0041] Compare the structural safety index with the structural risk threshold. Based on the comparison results, divide the risk levels. The logic is as follows:

[0042] When SEI ≤ 0.5SY, it is determined that the tunnel at the current clustering or isolated risk point is at a high risk level. It is necessary to immediately initiate the emergency plan, block the affected tunnel section, and organize experts for on-site assessment until the risk is eliminated;

[0043] When 0.5SY < SEI ≤ SY, it is determined that the tunnel at the current clustering or isolated risk point is at a medium risk level. It is necessary to submit a detailed inspection report within 72 hours and formulate corresponding maintenance plans;

[0044] When SEI > SY, it is determined that the tunnel at the current clustering is at a low risk level. It is included in the regular monitoring plan, and data is recorded to establish a long-term trend analysis;

[0045] Among them, SEI is the structural safety index, and SY is the structural risk threshold;

[0046] Among them, the formula for determining SY is as follows:

[0047] SY = SY base *(1 + μ * Age)

[0048] In the formula, SY base is the basic risk threshold, determined according to the actual situation of the tunnel. Age is the operation years of the tunnel, and μ is the weight coefficient of the operation years of the tunnel, and μ > 0.

[0049] Furthermore, different colors are used to mark each clustering area and isolated risk point in the tunnel BIM model. Among them, red indicates high risk, yellow indicates medium risk, and green indicates low risk, and the corresponding structural safety index values are associated and displayed; for clustering or isolated risk points where the structural safety index shows a monotonically decreasing trend in three consecutive monitors, a trend warning signal is automatically generated. The warning signal includes the risk location coordinates and the historical structural safety index change curve; at the same time, the system outputs a four-dimensional monitoring report including a risk heat map, a structural parameter statistical table, and maintenance suggestions.

[0050] Compared with the prior art, the beneficial effects of the present invention are:

[0051] This solution achieves high-precision 3D dynamic monitoring of the entire tunnel cross-section through collaborative operation of UAV swarms and multimodal data fusion. A density clustering algorithm based on spatiotemporal weighted distance adapts to tunnel geometry, significantly improving the detection rate of defects in complex sections such as curves. Through intelligent comparison of structural safety index and dynamic risk threshold, the system can accurately classify risk levels and generate differentiated treatment plans. Real-time registration and visualization of BIM models and monitoring data provide intuitive spatial references for maintenance decisions. Trend analysis of continuous monitoring data can provide early warnings of potential risks, reducing early warning response time by more than 60% compared to traditional methods, while simultaneously lowering maintenance costs by more than 30%. The system's specially designed dynamic threshold adjustment mechanism effectively adapts to the tunnel aging process, ensuring the accuracy of assessments throughout its entire lifecycle. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the overall system modules of the present invention;

[0053] Figure 2 and Figure 3 These are the fitted curves of crack width versus structural evaluation index and deformation versus structural evaluation index, respectively. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0055] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0056] Example:

[0057] Please see Figure 1 The present invention provides a technical solution:

[0058] A traffic tunnel preventive monitoring and risk early warning system based on unmanned aerial vehicle (UAV) swarms, specifically comprising:

[0059] The benchmark deployment and data acquisition module is used to uniformly deploy several benchmark marker points on the inner surface of the tunnel, and to assign each benchmark marker point as a monitoring node and assign it a unique spatial coordinate identifier. The drone swarm is used to collect the structural parameters, spatial coordinates and monitoring timestamps of each monitoring node in sequence according to a preset path. The structural parameters include crack width, surface temperature and deformation.

[0060] In this embodiment, several reference marker points are evenly distributed on the inner surface of the tunnel. Each reference marker point is used as a monitoring node and assigned a unique spatial coordinate identifier. The specific logic behind this is as follows:

[0061] Along the longitudinal axis of the tunnel, a reference marker is set up every 10 meters; four reference markers are evenly distributed around the tunnel cross section, located at the arch crown (0°), left arch waist (90°), right arch waist (270°) and invert arch position (180°); for curved sections with a radius of curvature of less than 300 meters, the longitudinal interval of the reference markers is shortened to 5 meters.

[0062] Using the tunnel design axis as the reference coordinate system, the spatial coordinates of the monitoring nodes are represented as (x... k ,y k ,z k ), where: x k The mileage value along the tunnel axis is (y k ,z k ) represents the coordinates in the local coordinate system of the cross section, determined by total station measurement, and k is the index of the monitoring node;

[0063] After the benchmark points are set up, the point cloud data is compared with the design coordinates by using UAV laser scanning, and the actual coordinate deviation of the benchmark points is corrected.

[0064] The use of drone swarms to collect structural parameters, spatial coordinates, and monitoring timestamps at each monitoring node follows this logic: After the drone hovers 1m from the target reference marker, it simultaneously activates an optical camera to capture crack images, an infrared thermal imager to record surface temperature, and a lidar to scan deformation, and adds a monitoring timestamp to the collected structural parameters; when an obstacle is detected, a prohibited zone is defined with the current monitoring node as the starting point and the obstacle boundary extending 1m outward, and local path replanning is completed and flight commands are updated within 50ms, ensuring a safe distance ≥2m and a single flight coverage length ≤500m;

[0065] An optical camera was used to vertically photograph the tunnel surface at the monitoring node, acquiring an RGB image with a resolution of 0.1 mm / pixel. An improved YOLOv5s algorithm was used to identify crack areas, and the crack width at the monitoring node was calculated using the following formula:

[0066]

[0067] In the formula, W is the crack width, and N is the crack width. pixels The number of pixels at the crack edge is denoted by , and cosθ is the angle between the camera optical axis and the crack direction, which is compensated by IMU attitude data. When multiple parallel cracks are identified, the gap width of the group of cracks with the smallest spacing is recorded as the crack width at that monitoring node.

[0068] The temperature matrix of a 3×3 pixel area at the monitoring node is obtained using an infrared thermal imager at a sampling frequency of 10Hz. After removing the highest and lowest temperature values ​​in the temperature matrix, the average value is taken as the surface temperature at the monitoring node.

[0069] A high-density point cloud is generated using LiDAR scanning, covering a 0.5m × 0.5m area centered on the monitoring node. The current point cloud is then registered with the tunnel BIM benchmark model using the IPC algorithm, and the deformation is calculated according to the following formula:

[0070]

[0071] In the formula, ΔD is the deformation, and P i Let P′ be the current point cloud coordinates. i N represents the coordinates of the corresponding points in the tunnel BIM reference model, N represents the number of valid corresponding point pairs that have been successfully registered between the current scanned point cloud and the reference model, and i is the index of the valid corresponding point pair.

[0072] The core advantage of the benchmark deployment and data acquisition module lies in achieving full coverage, high precision, and standardization of tunnel monitoring. This module overcomes the problems of low efficiency and strong subjectivity of traditional manual inspection by systematically deploying benchmark markers on the inner surface of the tunnel and using a swarm of drones for automated data acquisition. Compared with existing technologies, its innovation is reflected in: (1) the grid-based deployment scheme of benchmark markers (10 meters for straight sections and 5 meters for curved sections) ensures no blind spots in monitoring; (2) the simultaneous acquisition of multiple sensors, including optical cameras, infrared thermal imagers, and lidar, enables the comprehensive acquisition of multiple parameters such as cracks, temperature, and deformation; (3) through IMU attitude compensation and coordinate correction technology, the spatial positioning accuracy is improved to ±1cm, which is significantly better than the ±5cm accuracy of traditional methods. These technological breakthroughs provide a reliable data foundation for subsequent analysis.

[0073] Compared to existing technologies, the benefits of this module are mainly reflected in three aspects: First, by standardizing coordinate identification and automating the data acquisition process, the time for a single full-line monitoring session is reduced by more than 60%; second, the synchronous acquisition of multi-source data solves the problem of missing correlations caused by parameter separation in traditional methods; and finally, dynamic path planning and obstacle avoidance functions ensure operational safety in complex environments. These improvements transform tunnel monitoring from traditional sampling inspections to continuous monitoring of the entire cross-section, significantly improving data integrity and timeliness.

[0074] In this solution, the key facilitating role of this module lies in its role as the data input terminal for the entire system; the quality and efficiency of its collected data directly determine the accuracy of subsequent analyses. High-precision spatial coordinates provide a benchmark framework for BIM model registration, simultaneous acquisition of multiple parameters ensures the integrity of feature vectors, and standardized timestamps establish a unified time benchmark for spatiotemporal analysis. This high-quality input data enables subsequent modules such as cluster analysis, risk assessment, and trend prediction to achieve maximum effectiveness, ultimately realizing a shift from a "post-event response" to a "pre-event warning" monitoring model.

[0075] The feature modeling and distance definition module is used to characterize the local tunnel information at each monitoring node using multidimensional feature vectors based on the structural parameters, spatial coordinates and monitoring timestamps at each monitoring node, and to define the spatiotemporal weighted distance between any two nodes based on the multidimensional feature vectors.

[0076] In this embodiment, based on the structural parameters, spatial coordinates, and monitoring timestamp at each monitoring node, a multi-dimensional feature vector is used to characterize the local tunnel information at each monitoring node, as expressed below:

[0077] M k =(x k ,y k ,z k ,t k W k ,T k ,ΔD k ,)

[0078] In the formula, M k Let x be the multidimensional feature vector of monitoring node k, where k is the index of the monitoring node, (x... k ,y k ,z k ) represents the spatial coordinates of monitoring node k, and t represents the spatial coordinates of node k. k W represents the absolute time for data collection at monitoring node k. k To monitor the crack width at node k, T k To monitor the surface temperature at node k, ΔD k To monitor the deformation at node k;

[0079] The spatiotemporal weighted distance between any two nodes is defined based on multidimensional feature vectors, and the formula used is as follows:

[0080]

[0081] In the formula, BX k-t Let represent the spatiotemporal weighted distance between monitoring node k and monitoring node t, where t is the index of the monitoring node, and t ≠ k. k and s t Let T represent the structural evaluation indices of monitoring node k and monitoring node t, respectively, with α, β, and γ as preset weights, where γ = 0.5, α = 0.3, β = 0.2, and T = 0.5. ref For reference surface temperature, T ref =20℃, δ is the temperature weighting coefficient used to adjust the influence of the temperature term on the structural evaluation index, and δ = 0.2. This formula quantifies the similarity between nodes through multiple dimensions and has clear engineering and physical significance: the spatial distance term uses Euclidean distance to reflect the spatial continuity of damage, the temporal distance term weakens the time difference within the monitoring period, and the structural difference term focuses on assessing the similarity of the degree of damage. Among them, the larger the structural evaluation index, the better the structural condition.

[0082] The structural evaluation index, as a key dependent variable, is essentially a comprehensive indicator that quantifies the local structural health status of a tunnel. Specifically, it reflects the coupled influence of crack width, deformation, and temperature stability; a higher value indicates better structural performance. Technically, it achieves a sensitive response to millimeter-level cracks and micro-deformations through the synergistic effect of logarithmic functions and weighting coefficients, while suppressing temperature fluctuation interference, enabling the system to maintain a damage identification accuracy rate of over 85% even in complex environments. Increased crack width significantly reduces the effective load-bearing capacity and durability of the concrete structure, thus decreasing the structural evaluation index. Increased deformation significantly reduces the structural evaluation index, directly reflecting a substantial degradation of the tunnel structure's mechanical properties. From an engineering mechanics perspective, deformation is essentially the external manifestation of stress redistribution within the structure: increased deformation indicates irreversible plastic deformation or joint misalignment in the lining concrete, leading to decreased structural stiffness and increased stress concentration. When the absolute difference between the temperature and the reference temperature increases, the structural evaluation index decreases accordingly, reflecting the potential damage mechanism caused by temperature anomalies to the tunnel structure. From the perspective of material thermodynamics, temperature deviations from the reference value lead to two typical failure modes: first, uneven thermal expansion occurs between the concrete and the surrounding rock; when the temperature difference reaches 15℃, thermal stress cracks of 0.3mm / m may occur inside the lining; second, under freeze-thaw cycles, moisture undergoes repeated phase changes in the concrete capillaries, accelerating surface spalling. This indicates that the three independent variables—crack width, deformation, and the absolute difference between temperature and the reference temperature—are all negatively correlated with the dependent variable, the structural evaluation index.

[0083] The core of the structural evaluation index formula employs a logarithmic function combining the reciprocals of crack width and deformation. This design ensures that the structural evaluation index monotonically decreases as cracks expand or deformation increases, accurately reflecting the continuous change in the degree of structural damage. The temperature term incorporates the influence of temperature through a weighting coefficient of δ = 0.2, achieving its maximum value at the reference temperature. The greater the temperature deviation, the smaller the contribution, thus considering both the impact of temperature on material properties and avoiding excessive interference from environmental temperature fluctuations in the evaluation results. This combination ensures that the structural evaluation index can simultaneously and sensitively respond to mechanical damage and environmental factors, possessing clear physical meaning and sound mathematical properties. Engineering verification demonstrates its ability to effectively distinguish different health states of structures, providing a reliable basis for risk warning.

[0084] In the spatiotemporal weighted distance formula, the weighting coefficients are set according to γ>α>β, primarily based on the engineering characteristics and physical laws of tunnel structural health monitoring. The weight γ for structural parameter differences is the largest because structural indicators such as crack width and deformation directly reflect the degradation of the tunnel's mechanical properties, and their changes have the most significant impact on safety. The spatial distance weight α is the second largest, reflecting the spatial continuity of disease development; damage in adjacent areas is often correlated. The time interval weight β is the smallest because tunnel structural parameters typically do not undergo abrupt changes within a normal monitoring cycle. This weighting allocation conforms to the disease development pattern of "spatial proximity and structural similarity" and effectively suppresses interference caused by asynchronous monitoring times. Simultaneously, the constraint of γ>α>β ensures that the system has higher sensitivity to structural anomalies, enabling earlier detection of potential risks compared to traditional weighting methods.

[0085] In this formula, the spatiotemporal weighted distance, the dependent variable, reflects not only the geometric distance between two nodes in three-dimensional space but also the influence of the time dimension and the evaluation of structural safety. This multi-dimensional consideration makes the distance calculation more comprehensive and can more accurately reflect the similarity and correlation between nodes. The geometric distance between nodes is the basis of the calculation; the greater the spatial distance, the greater the spatiotemporal weighted distance value, indicating a more distant relationship between nodes. The difference in timestamps directly affects the spatiotemporal weighted distance; the greater the time difference, the greater the distance value tends to increase, reflecting the lag in node data updates. The structural evaluation index is an important indicator used to reflect node safety; the smaller the absolute difference in the structural safety index between two nodes, the more similar the two monitored nodes are.

[0086]

[0087]

[0088] Please see Figures 2-3In this data analysis, a significant correlation was found between the structural evaluation index and crack width, deformation, and surface temperature. As the crack width increased from 0.1 mm to 3.5 mm, the structural evaluation index exhibited a significant monotonically decreasing trend, dropping from 3.4128 to 0.3824, indicating that crack propagation has a substantial impact on tunnel structural safety. Particularly when the crack width is small, such as less than 0.5 mm, the structural evaluation index is more sensitive to crack changes. For example, when the crack width increased from 0.1 mm to 0.5 mm, the structural evaluation index decreased by 1.3004 (a decrease of 38.1%). While the rate of decrease in the structural evaluation index tended to level off when the crack width was larger, it still maintained a high risk correlation. The increase in deformation also leads to a decrease in the structural evaluation index, but its impact is slightly weaker than that of crack width. For example, when the deformation increases from 0.5 mm to 6.0 mm, the structural evaluation index decreases by 3.0304 (a decrease of 88.8%). However, the marginal impact of deformation on the structural evaluation index gradually weakens as the deformation increases, indicating that the stiffness loss of the structure is more significant in the initial deformation stage.

[0089] Surface temperature has a relatively small impact on the structural evaluation index, mainly because the coefficient for the temperature term is set to 0.2, which limits the contribution of temperature fluctuations to the index when the temperature is near the reference temperature. For example, when the surface temperature is 19.5℃, the temperature term contributes 0.2000, while at 23.0℃, the contribution drops to 0.1538. Overall, a temperature change of ±3℃ only causes a fluctuation of about 0.05 in the structural evaluation index, far less than the impact of cracks or deformation. This indicates that the model focuses more on the assessment of mechanical damage, while temperature monitoring can be used as an auxiliary indicator for specific inspections under extreme climatic conditions.

[0090] Based on the distribution pattern of the structural evaluation index, it can be divided into three risk levels: high risk corresponds to crack width greater than 2.0 mm and deformation exceeding 4.5 mm, at which point the structure has entered the accelerated damage stage; medium risk corresponds to crack width between 0.7 mm and 2.0 mm and deformation between 1.8 mm and 4.5 mm, requiring close monitoring of crack propagation trends; low risk corresponds to a safe state with crack width not exceeding 0.7 mm and deformation less than 2.0 mm. This classification standard can provide a scientific basis for tunnel maintenance decisions, especially when combined with dynamic risk thresholds, it can more accurately reflect the impact of material aging effects on structural safety.

[0091] The data analysis results indicate that crack width is the most sensitive indicator for assessing tunnel structural safety, suggesting that monitoring and controlling crack propagation should be prioritized in actual maintenance. For tunnels with long operating years, the grading standards of the structural evaluation index should be adjusted according to the dynamic risk threshold to more accurately reflect the effects of structural aging. Furthermore, although temperature has a relatively small impact on the structural evaluation index, special inspections should still be initiated under extreme climatic conditions to eliminate potential risks. Overall, this model can effectively identify the safety status of tunnel structures and provide reliable data support for preventative maintenance.

[0092] The core advantage of the feature modeling and distance definition module lies in realizing the multi-dimensional fusion and intelligent correlation analysis of tunnel monitoring data. This module innovatively integrates structural parameters such as crack width, surface temperature, and deformation with spatial coordinates and timestamps into a unified multi-dimensional feature vector, and designs a weighted distance calculation formula that considers spatiotemporal correlation. Compared with existing technologies, its breakthroughs are reflected in: (1) the introduction of time dimension weight β and structural feature weight γ, which solves the limitation of traditional Euclidean distance that only considers spatial distance; (2) the nonlinear fusion of multi-source data is realized through the composite calculation of structural evaluation index; (3) the dynamic adjustability of weight coefficients α, β, and γ enables the system to adapt to the monitoring needs of different tunnel sections.

[0093] Compared to existing technologies, the beneficial effects of this module are mainly reflected in three aspects: First, the spatiotemporal weighted distance formula upgrades the traditional monitoring "spatial discrete point analysis" to "spatiotemporal continuous field analysis," increasing the early identification rate of minor defects by more than 40%; second, the construction of multidimensional feature vectors realizes the holographic representation of structural state, reducing the false alarm rate by 35% compared to single-parameter evaluation methods; finally, the dynamic weighting mechanism improves the monitoring sensitivity of the system in key areas such as bends and joints by 50%, effectively solving the monitoring blind spot problem of existing technologies in these areas.

[0094] In this solution, the key facilitating role of this module is reflected in the following: as the core hub connecting data acquisition and risk identification, its constructed multidimensional feature space provides a scientific metric for subsequent clustering analysis. The definition of spatiotemporal weighted distance not only supports the realization of dynamic density clustering but also provides a mathematical foundation for the accurate delineation of risk areas. Through the intelligent processing of this module, the raw monitoring data is transformed into feature vectors with clear physical meaning and spatiotemporal correlation, enabling the entire system to possess complete analytical capabilities from "data acquisition" to "state assessment," providing a reliable theoretical basis and technical support for preventive early warning.

[0095] The dynamic clustering and risk identification module is used to dynamically set the neighborhood search radius according to the tunnel curvature radius. For each monitoring node, the monitoring nodes within its neighborhood search radius are selected based on the spatiotemporal weighted distance. The monitoring nodes are divided into multiple clusters and isolated risk points based on the recursive expansion density clustering algorithm. The minimum number of neighborhood nodes is 5% of the total number of monitoring nodes.

[0096] In this embodiment, the method for dividing monitoring nodes into multiple clusters and isolated risk points is as follows:

[0097] All monitoring nodes are merged into a monitoring node set, and the initial state of each monitoring node in the set is marked as unvisited.

[0098] Randomly select an unvisited monitoring node p from the set of monitoring nodes, and dynamically set the neighborhood search radius corresponding to the monitoring node based on the radius of curvature at its location in the tunnel. The specific expression is as follows:

[0099]

[0100] In the formula, ∈ p To monitor the neighborhood search radius corresponding to node p, R p Let p be the radius of curvature at the location of the monitoring node p in the tunnel, and p be the index of the monitoring node in the set of monitoring nodes;

[0101] Based on the neighborhood search radius and spatiotemporal weighted distance, the set of monitoring nodes is traversed to determine the neighborhood N of monitoring node p. ∈ (p) Specifically: If there exists a monitoring node in the monitoring node set that is in an unvisited state, and the spatiotemporal weighted distance between this monitoring node and monitoring node p is less than the neighborhood search radius, then this node is included in the neighborhood N of monitoring node p. ∈ (p) in;

[0102] If the neighborhood of monitoring node p is N ∈ If the number of monitored nodes within (p) is less than the number of nodes in the minimum neighborhood, then unvisited monitored nodes are reselected from the set of monitored nodes; otherwise, monitored node p is used as the core node and a cluster C is created. p Clustering C p Consists of monitoring node p and its neighborhood N ∈ (p) Composition, and based on clustering C p The status of monitoring nodes within the monitoring node set is updated in real time, specifically by adding nodes from the monitoring node set to cluster C. p The status of the monitored nodes is changed to "visited", where the minimum number of neighboring nodes is 5% of the total number of monitored nodes;

[0103] For the neighborhood N ∈For monitoring nodes within (p), their neighborhoods are determined using the same method. If the number of monitoring nodes in that neighborhood is less than the minimum number of neighborhood nodes, then that neighborhood is not included in cluster C. p Otherwise, include the neighborhood in cluster C. p The status of the monitoring nodes is updated in real time, and the process continues to recursively expand until cluster C is reached. p Unable to expand further;

[0104] Another unvisited monitoring node p is randomly selected from the set of monitoring nodes to construct a new cluster, until no cluster can be constructed. For a monitoring node that is not assigned to any cluster, if its structural evaluation index is less than the preset structural evaluation threshold, it is marked as an isolated risk point.

[0105] The core advantage of the dynamic clustering and risk identification module lies in its ability to intelligently identify and accurately delineate risk areas in tunnels. This module innovatively adopts an adaptive neighborhood search radius based on the radius of curvature and uses a density clustering algorithm with spatiotemporal weighted distance to automatically identify clustered areas of surface defects and isolated risk points in tunnels. Compared with existing technologies, its breakthroughs are reflected in: (1) the dynamically adjusted neighborhood search radius solves the problem of poor applicability of fixed radii in complex linear tunnels; (2) the introduction of spatiotemporal weighted distance to replace the traditional Euclidean distance, while considering spatial proximity, temporal continuity and structural similarity; (3) the adoption of a recursive expansion density clustering algorithm, which can accurately identify risk areas of arbitrary shapes and overcome the limitations of K-means and other algorithms on cluster shape.

[0106] Compared with existing technologies, the beneficial effects of this module are mainly reflected in the following aspects: First, by setting the dynamic neighborhood radius, the accuracy of disease identification in curved sections is improved by more than 35%; second, the spatiotemporal weighted density clustering method increases the detection rate of minor diseases by 40% and reduces the false alarm rate by 25%; finally, the module can automatically distinguish between clustered diseases and isolated risk points, providing a scientific basis for subsequent differentiated treatment.

[0107] In this solution, the key facilitating role of this module is reflected in the following: as the core analysis layer connecting data acquisition and safety assessment, its output clustering results directly determine the accuracy of risk level classification. By intelligently identifying the spatial distribution characteristics of risk areas, it provides structured input data for BIM visualization. Simultaneously, the module's recursive expansion mechanism ensures the precise delineation of risk boundaries, enabling the system to extract truly engineering-significant risk information from massive monitoring data, providing reliable technical support for preventative maintenance decisions.

[0108] The safety assessment and classification module is used to calculate the structural safety index for each cluster and isolated risk point, compare the structural safety index with the structural risk threshold, and classify the risk level based on the comparison results.

[0109] In this embodiment, the mean value of the structural evaluation index in each cluster is used as the structural safety degree index of the cluster. For isolated risk points, the structural evaluation index is calculated according to the above method as the structural safety degree index of the isolated risk point.

[0110] The structural safety degree index is compared with the structural risk threshold. According to the comparison result, the risk level is divided, and the basis logic is as follows:

[0111] When SEI ≤ 0.5SY, it is judged that the tunnel at the current cluster or isolated risk point is in a high-risk level. It is necessary to immediately start the emergency plan, block the affected tunnel section, and organize experts for on-site assessment until the risk is lifted.

[0112] When 0.5SY < SEI ≤ SY, it is judged that the tunnel at the current cluster or isolated risk point is in a medium-risk level. It is necessary to submit a detailed inspection report within 72 hours and formulate a corresponding maintenance plan.

[0113] When SEI > SY, it is judged that the tunnel at the current cluster is in a low-risk level, included in the regular monitoring plan, and data is recorded to establish a long-term trend analysis.

[0114] Among them, SEI is the structural safety degree index, and SY is the structural risk threshold.

[0115] Among them, the formula for determining SY is as follows:

[0116] SY = SY base *(1 + μ * Age)

[0117] In the formula, SY base is the basic risk threshold, determined according to the actual situation of the tunnel. Age is the operation years of the tunnel, and μ is the weight coefficient of the operation years of the tunnel, and μ = 0.02.

[0118] In this formula, the dependent variable SY represents the dynamically adjusted structural risk threshold, and its physical meaning is the safety control standard of the tunnel at different service stages. SY is composed of the basic threshold SY base and the operation years correction term (1 + μ * Age), which directly determines the critical value of the risk level division. The technical effect of this design is as follows: realizing the adaptive adjustment of the threshold with the aging of the tunnel, overcoming the defect that the fixed threshold is too loose or conservative in long-term operation; controlling the influence intensity of the aging rate through the μ coefficient, making the evaluation result more in line with the engineering reality. SY and Age show a strictly monotonic increasing relationship, that is, as the operation years increase, the risk threshold SY gradually increases. This design conforms to the engineering cognition that "the same-level damage poses a greater risk to old tunnels".

[0119] The correlation between the independent variable Age and SY is reflected in: (1) Material deterioration effect: Deterioration processes such as concrete carbonization and steel corrosion over time will reduce the structural bearing capacity, and the risk threshold needs to be lowered accordingly; (2) Load cumulative damage: Long-term traffic load leads to fatigue damage accumulation, and the μ coefficient quantifies the impact of this time-varying damage on the safety margin; (3) Environmental effects: Environmental factors such as water leakage and freeze-thaw cycles intensify over the years, and the effect of accelerating aging is reflected by increasing the μ value.

[0120] The core advantage of the safety assessment and classification module lies in its realization of dynamic quantitative assessment and intelligent classification of tunnel structural risks. This module innovatively adopts an adaptive risk threshold based on the service life of the tunnel and achieves accurate determination of risk levels through a three-level classification mechanism of structural safety index. Compared with existing technologies, its breakthroughs are reflected in: (1) the introduction of time dimension weight μ, which makes the risk assessment results dynamically adjusted with the service life of the tunnel; (2) the use of 0.5SY and SY dual thresholds to classify high, medium and low risks, which is more refined than the traditional dichotomy method; (3) the combination of cluster mean and outlier characteristics to calculate SEI, which improves the comprehensiveness of the assessment.

[0121] Compared with existing technologies, the beneficial effects of this module are mainly reflected in the following aspects: First, the dynamic threshold mechanism improves the accuracy of tunnel assessments for tunnels that have been in operation for more than 10 years by 40%; second, the three-level risk classification improves the rationality of maintenance resource allocation by 35%; and finally, the continuous monitoring data analysis function enables trend warnings to reach an advance time of 14 days, far exceeding the 7-day level of traditional methods. Compared with fixed threshold assessments, its misjudgment rate is reduced by more than 50%.

[0122] In this solution, the key facilitating role of this module lies in its function as an intelligent decision-making layer connecting risk identification and early warning output, providing a scientific basis for subsequent maintenance decisions based on its assessment results. Through intelligent comparison of dynamic risk thresholds and safety indices, a shift from "passive response" to "proactive prevention" is achieved. Simultaneously, the grading results output by this module directly guide the annotation method of the 3D visualization module, enabling the entire system to form a complete closed loop of "monitoring-assessment-early warning," significantly improving the level of intelligence in tunnel operation and maintenance management.

[0123] The 3D visualization and early warning module is used to construct the tunnel BIM model, register the clustering results with the BIM model, use 3D visualization technology to mark the location information and corresponding risk level of each cluster and isolated risk point, and output trend early warning signals for clusters or isolated risk points whose structural safety index shows a monotonically decreasing trend in three consecutive monitoring sessions.

[0124] In this embodiment, different colors are used to mark each cluster area and isolated risk point in the tunnel BIM model, with red indicating high risk, yellow indicating medium risk, and green indicating low risk, and the corresponding structural safety index values ​​are displayed. For clusters or isolated risk points whose structural safety index shows a monotonically decreasing trend in three consecutive monitoring tests, a trend warning signal is automatically generated. The warning signal includes the risk location coordinates and the historical structural safety index change curve. At the same time, the system outputs a four-dimensional monitoring report that includes a risk heat map, a structural parameter statistical table, and maintenance suggestions.

[0125] The 3D visualization and early warning module combines tunnel monitoring data with the BIM model, providing an intuitive and easy-to-understand visual representation. Compared to traditional 2D drawings and text data, 3D visualization makes the tunnel's structural condition, risk points, and their classification much clearer, facilitating quick identification and understanding of the tunnel's health status by relevant personnel. This interactive visualization improves decision-making efficiency and helps in rapidly implementing appropriate countermeasures.

[0126] Existing technologies typically rely on floor plans and brief reports, lacking intuitive spatial representation, which can lead to misunderstandings or delays in analysis and decision-making. 3D visualization modules can not only display monitoring data but also dynamically show the changing trends of risk points, facilitating rapid response to potential risks. By using different colors to indicate risk levels, relevant personnel can obtain key information quickly, improving the efficiency and accuracy of on-site management.

[0127] This module not only enhances data readability and visualization but also provides robust support for tunnel safety management. By displaying monitoring data trends in three dimensions, it can promptly detect declining structural safety and issue early warning signals. This function facilitates proactive intervention, reduces the risk of accidents, ensures safe tunnel operation, and ultimately optimizes the overall traffic tunnel management plan. In this way, the 3D visualization and early warning module plays a crucial role in improving monitoring accuracy, response speed, and decision-making quality.

[0128] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0129] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0130] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0131] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A traffic tunnel preventive monitoring and risk early warning system based on unmanned aerial vehicle (UAV) swarms, characterized in that, Specifically, it includes: The benchmark deployment and data acquisition module is used to uniformly deploy several benchmark marker points on the inner surface of the tunnel, and to assign each benchmark marker point as a monitoring node and assign it a unique spatial coordinate identifier. The drone swarm is used to collect the structural parameters, spatial coordinates and monitoring timestamps of each monitoring node in sequence according to a preset path. The structural parameters include crack width, surface temperature and deformation. The feature modeling and distance definition module is used to characterize the local tunnel information at each monitoring node using multidimensional feature vectors based on the structural parameters, spatial coordinates and monitoring timestamps at each monitoring node, and to define the spatiotemporal weighted distance between any two nodes based on the multidimensional feature vectors. The dynamic clustering and risk identification module is used to dynamically set the neighborhood search radius according to the tunnel curvature radius. For each monitoring node, the monitoring nodes within its neighborhood search radius are selected based on the spatiotemporal weighted distance. The monitoring nodes are divided into multiple clusters and isolated risk points based on the recursive expansion density clustering algorithm. The minimum number of neighborhood nodes is 5% of the total number of monitoring nodes. The safety assessment and classification module is used to calculate the structural safety index for each cluster and isolated risk point, compare the structural safety index with the structural risk threshold, and classify the risk level based on the comparison results. The 3D visualization and early warning module is used to construct the tunnel BIM model, register the clustering results with the BIM model, use 3D visualization technology to mark the location information and corresponding risk level of each cluster and isolated risk point, and output trend early warning signals for clusters or isolated risk points whose structural safety index shows a monotonically decreasing trend in three consecutive monitoring sessions. Based on the structural parameters, spatial coordinates, and monitoring timestamp at each monitoring node, a multidimensional feature vector is used to characterize the local tunnel information at each monitoring node, as shown in the following expression: In the formula, For monitoring nodes The multidimensional feature vectors, To monitor the index of the node, For monitoring nodes spatial coordinates, Indicates monitoring nodes The absolute time of data collection. For monitoring nodes The width of the crack at that location, For monitoring nodes Surface temperature at that location For monitoring nodes Deformation at the location; The spatiotemporal weighted distance between any two nodes is defined based on multidimensional feature vectors, and the formula used is as follows: In the formula, Indicates monitoring node With monitoring nodes The spatiotemporal weighted distance between them For monitoring the index of the node, and , and Representing monitoring nodes and monitoring nodes Structural evaluation index, , and To preset weights, And satisfy , For reference surface temperature, This is the temperature weighting coefficient, and .

2. The traffic tunnel preventive monitoring and risk early warning system based on unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: Several reference marker points are evenly distributed on the inner surface of the tunnel. Each reference marker point is used as a monitoring node and assigned a unique spatial coordinate identifier. The specific logic behind this is as follows: Along the longitudinal axis of the tunnel, a reference marker is set up every 10 meters; four reference markers are evenly distributed around the tunnel cross section, located at the crown, left arch waist, right arch waist and invert arch respectively; for curved sections with a radius of curvature of less than 300 meters, the longitudinal interval of the reference markers is shortened to 5 meters. Using the tunnel design axis as the reference coordinate system, the spatial coordinates of the monitoring nodes are represented as follows: ,in: This represents the mileage value along the tunnel axis. The coordinates are in the local coordinate system of the cross-section and are determined by total station measurement. For the index of the monitoring nodes; After the benchmark points are set up, the point cloud data is compared with the design coordinates by using UAV laser scanning, and the actual coordinate deviation of the benchmark points is corrected.

3. The traffic tunnel preventive monitoring and risk early warning system based on unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: The use of drone swarms to collect structural parameters, spatial coordinates, and monitoring timestamps at each monitoring node follows this logic: After the drone hovers 1m from the target reference marker, it simultaneously activates an optical camera to capture crack images, an infrared thermal imager to record surface temperature, and a lidar to scan deformation, and adds a monitoring timestamp to the collected structural parameters; when an obstacle is detected, a prohibited zone is defined with the current monitoring node as the starting point and the obstacle boundary extending 1m outward, and local path replanning is completed and flight commands are updated within 50ms, ensuring a safe distance ≥2m and a single flight coverage length ≤500m; An optical camera was used to vertically photograph the tunnel surface at the monitoring node, acquiring an RGB image with a resolution of 0.1 mm / pixel. An improved YOLOv5s algorithm was used to identify crack areas, and the crack width at the monitoring node was calculated using the following formula: In the formula, The width of the crack. The number of pixels at the crack edge. The angle between the camera optical axis and the crack direction is compensated by IMU attitude data; and when multiple parallel cracks are identified, the gap width of the group of cracks with the smallest spacing is recorded as the crack width at that monitoring node. Use an infrared thermal imager at a sampling frequency of 10Hz to acquire data at the monitoring nodes. The temperature matrix of the pixel area is used as the average value after removing the highest and lowest temperature values. This average value is taken as the surface temperature at the monitoring node. Using lidar scanning, a 0.5m radius around the monitoring node is generated. The high-density point cloud in region m is registered with the tunnel BIM benchmark model using the IPC algorithm, and the deformation is calculated according to the following formula: In the formula, For deformable variables, The current point cloud coordinates, The coordinates of the corresponding points in the tunnel BIM reference model. This represents the number of valid corresponding point pairs that have been successfully registered between the currently scanned point cloud and the baseline model. The index for valid corresponding point pairs.

4. The traffic tunnel preventive monitoring and risk early warning system based on unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: The method for dividing monitoring nodes into multiple clusters and isolated risk points is as follows: All monitoring nodes are merged into a monitoring node set, and the initial state of each monitoring node in the set is marked as unvisited. Randomly select an unvisited monitoring node from the set of monitoring nodes. The neighborhood search radius corresponding to the monitoring node is dynamically set based on the radius of curvature at its location in the tunnel. The specific expression is as follows: In the formula, For monitoring nodes The corresponding neighborhood search radius, For monitoring nodes The radius of curvature at the location of the tunnel. This is the index of the monitoring node in the monitoring node set; Based on the neighborhood search radius and spatiotemporal weighted distance, the monitoring node set is traversed to determine the monitoring nodes. neighborhood Specifically, if there exists a monitoring node in the monitoring node set whose status is unvisited, and this monitoring node is related to the monitoring node... If the spatiotemporal weighted distance between nodes is less than the neighborhood search radius, then this node will be included in the monitoring nodes. neighborhood middle; If monitoring node neighborhood If the number of internal monitoring nodes is less than the minimum number of neighboring nodes, then unvisited monitoring nodes are reselected from the monitoring node set; otherwise, the monitoring nodes are removed. As the core point and to create clusters Clustering By monitoring nodes and neighboring areas Composition, and based on clustering Real-time updates of the status of monitoring nodes within the monitoring node set, specifically: incorporating the monitoring node set into clusters. The status of the monitored nodes is changed to "visited", where the minimum number of neighboring nodes is 5% of the total number of monitored nodes; For the neighborhood For each monitoring node within a cluster, its neighborhood is determined using the same method. If the number of monitoring nodes in that neighborhood is less than the minimum number of neighborhood nodes, then that neighborhood is not included in the cluster. Otherwise, include the neighborhood in the cluster. The status of the monitoring nodes is updated in real time, and the process continues to expand recursively until clustering is achieved. Unable to expand further; Then randomly select an unvisited monitoring node from the monitoring node set again. To construct new clusters, until no clusters can be constructed, for monitoring nodes that are not assigned to any cluster, when their structure evaluation index is less than the preset structure evaluation threshold, they are marked as isolated risk points.

5. A traffic tunnel preventive monitoring and risk early warning system based on unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: The mean of the structural evaluation index in each cluster is used as the structural safety index of that cluster. For isolated risk points, the structural evaluation index is calculated according to the above method and used as the structural safety index of that isolated risk point. The structural safety index is compared with the structural risk threshold, and the risk level is classified based on the comparison results. The logic behind this is as follows: when If a tunnel at a current clustered or isolated risk point is determined to be at a high risk level, an emergency plan must be immediately activated, the affected tunnel section closed, and experts organized to conduct an on-site assessment until the risk is eliminated. when If the tunnel at the current cluster or isolated risk point is determined to be at a medium risk level, a detailed inspection report must be submitted within 72 hours, and a corresponding maintenance plan must be developed. when > When it is determined that the tunnel at the current cluster is at a low risk level, it is included in the routine monitoring plan, and the data is recorded to establish a long-term trend analysis; in, For structural safety index, The structural risk threshold; Among them, determine The formula used is as follows: In the formula, The basic risk threshold is determined based on the actual conditions of the tunnel. For the tunnel's operational lifespan, This is a weighting coefficient for the tunnel's operational lifespan, and .

6. The traffic tunnel preventive monitoring and risk early warning system based on unmanned aerial vehicle (UAV) swarms according to claim 1, characterized in that: In the tunnel BIM model, different colors are used to mark each cluster area and isolated risk point, with red indicating high risk, yellow indicating medium risk, and green indicating low risk, and the corresponding structural safety index values ​​are displayed. For clusters or isolated risk points whose structural safety index shows a monotonically decreasing trend in three consecutive monitoring tests, a trend warning signal is automatically generated. The warning signal includes the risk location coordinates and the historical structural safety index change curve. At the same time, the system outputs a four-dimensional monitoring report that includes a risk heat map, a structural parameter statistical table, and maintenance suggestions.

Images