Method for determining detection distance of bridge inspection unmanned aerial vehicle based on wind field pressure distribution and application

Abstract

The application discloses a bridge inspection unmanned aerial vehicle detection distance determination method and application based on wind field pressure distribution, relates to the technical field of bridge detection by means of unmanned aerial vehicles, and obtains the three-dimensional wind field pressure distribution around the bridge through numerical simulation, establishes a space risk grade division model, and combines the wind resistance performance parameters of the unmanned aerial vehicle to determine the safe detection distance of the unmanned aerial vehicle around the bridge, wherein the height in the safe detection distance only relates to the upper side of the bridge. The application obtains the most unfavorable wind field pressure by coupling flow field analysis, uses the most unfavorable wind field pressure as a flight decision constraint condition, and for the first time, on the basis of quantitative wind field calculation, the most unfavorable wind field pressure of the oncoming heavy truck is taken into the core basis of the bridge inspection unmanned aerial vehicle, so that the blank of the existing research is filled, and the flight safety is improved.

Description

Technical Field

[0001] This invention relates to the technical field of bridge inspection using unmanned aerial vehicles (UAVs). More specifically, this invention relates to a method for determining the detection distance of bridge inspection UAVs based on wind pressure distribution and its application. Background Technology

[0002] With the continuous expansion of bridge transportation, bridge structures bear the burden of frequent heavy vehicle traffic, making their structural and operational safety increasingly important. To improve the efficiency and safety of bridge inspection, drones are gradually being applied to tasks such as bridge visual inspection, structural inspection, and defect identification. Compared to traditional manual inspection methods, drone inspection offers advantages such as high mobility, high efficiency, and low safety risks.

[0003] However, in actual bridge inspections, heavy-duty trucks traveling on the bridge deck significantly disturb the surrounding airflow, especially when traveling in opposite directions. Vehicle-induced wind fields create complex pressure fluctuations and turbulent structures above, below, and to the sides of the bridge. These unstable aerodynamic disturbances can easily affect the flight attitude and control stability of drones, potentially leading to flight deviations or even loss of control, thus increasing inspection risks.

[0004] Existing UAV bridge inspection methods are mostly based on empirical flight altitude or simple safety distance settings, lacking systematic analysis of the impact of wind fields on heavy-duty vehicles. They fail to combine vehicle structural dimensions with the aerodynamic distribution characteristics of the bridge space for risk zoning, and also fail to implement graded matching control based on the UAV's own wind resistance capabilities, making it difficult to scientifically and quantitatively determine safe flight distances. In complex wind field environments, if the UAV detection distance is set inappropriately, problems such as unstable flight attitude, blurred images, reduced detection accuracy, and even flight instability can easily occur, posing significant safety risks.

[0005] Therefore, it is necessary to propose a method for determining the detection distance of UAVs based on coupled flow field analysis of heavy-duty truck bridges, so as to determine the safe flight altitude range and improve the safety and reliability of UAV bridge inspection operations. Summary of the Invention

[0006] One object of the present invention is to solve at least the above-mentioned problems and / or defects, and to provide at least the advantages described below.

[0007] To achieve these objectives and other advantages of the present invention, a method for determining the detection distance of a bridge inspection UAV based on wind field pressure distribution is provided, comprising: S1. Establish a coupled flow field model of heavy-duty trucks and bridges, and the coupled flow field model includes a three-dimensional computational domain of the bridge structure and the heavy-duty trucks traveling in opposite directions. S2. Perform numerical simulation calculations on the coupled flow field model to obtain three-dimensional wind pressure distribution data around the bridge, whereby the bridge perimeter includes the vertical and lateral spatial regions of the bridge. S3. Based on three-dimensional wind field pressure distribution data, and taking the height H of the heavy-duty truck body as the benchmark, the wind field risk level around the bridge is classified to obtain the corresponding safe flight altitude range. S4. Calculate the pressure gradient distribution data of the spatial region corresponding to the safe flight altitude range, and select low-risk flight areas from them. S5. In a low-risk flight area, based on the performance parameters of the UAV, determine the safe detection distance of the UAV around the bridge, wherein the height in the safe detection distance only involves the area above the bridge.

[0008] Preferably, in S1, the coupled flow field model is based on the six-axle semi-trailer tractor traveling in opposite directions, established using ANSYS software. The coupled flow field model adopts the Realizable-K-∑ model, and the coupled flow field is calculated using a three-dimensional turbulence numerical simulation method. The length of the bridge in the longitudinal direction in the three-dimensional computational domain is not less than three times the bridge span, and the three-dimensional computational domain is discretized using structured or semi-structured meshes, with local mesh refinement applied to the vehicle surface and the area near the bridge deck.

[0009] Preferably, in S3, the wind field risk level classification method for the area above the bridge is as follows: Areas with an altitude range of 0 to 1.8H are designated as extremely high-risk zones; Areas with an altitude range of 1.8H to 2.5H are designated as high-risk areas; Regions with a height greater than 2.5H are designated as stable regions; For the area under the bridge, the area with a height range of 0 to H from the two sides of the bridge towards the middle is designated as a high-risk area, and the remaining area is a low-risk area. For the lateral space area of ​​the bridge, the area from 0 to 1.2H in the vertical direction is designated as a high-risk area, and the other areas are designated as low-risk areas; In the horizontal direction, the area 0 to 1.2H from the side edge of the bridge is designated as a high-risk area, and other areas are designated as low-risk areas.

[0010] Preferably, in S4, the method for selecting low-risk flight areas is as follows: S40. Calculate the spatial gradient of the wind and pressure fields in the space region corresponding to the safe flight altitude range to obtain the distribution data of the pressure gradient vector field. S41. Calculate the pressure gradient amplitude of each spatial unit based on the pressure gradient vector field; S42. Compare the pressure gradient amplitude with a preset gradient threshold, and designate the space unit with a pressure gradient amplitude less than the preset gradient threshold as a low-risk flight area.

[0011] Preferably, in S5, the performance parameters include: volume parameters, mass parameters, and wind resistance performance parameters; The safe detection distance is determined by comparing the wind resistance capability of the drone with a preset benchmark threshold. If the wind resistance capability is less than the preset benchmark threshold, the drone's flight altitude is limited to the stable zone above the bridge; otherwise, the drone's flight altitude is limited to the medium-risk zone and the stable zone above the bridge. When the drone flies under the bridge, the vertical distance between the drone and the bottom of the bridge shall not be less than H; The wind resistance capability is determined based on the maximum wind speed parameter or the maximum wind pressure parameter that the UAV can withstand, and the preset benchmark threshold is determined based on the wind field pressure parameter of the corresponding risk zone in the bridge space.

[0012] A path planning method for a bridge inspection drone includes: Step 1: Simulate the wind pressure distribution characteristics under heavy-duty vehicle driving conditions using a coupled flow field model; Step 2: The discrete wind pressure data output by the coupled flow field model is converted into a continuous cost field that can be directly queried in path planning using spatial interpolation techniques. Step 3, design an improved version of A * -RRT * For bridge detection, a fusion algorithm using a 3D grid A is employed. * The algorithm generates a global guiding path from the continuous cost field as the initial tree node of the RRT* algorithm, and obtains the three-dimensional trunk path I through iterative optimization; For under-bridge detection, a deterministic traversal method is used to generate a two-dimensional grid point sequence covering the entire under-bridge area from a continuous cost field, and three-dimensional attributes are assigned through height optimization to obtain the three-dimensional main path II; Among them, the safe detection distance of the UAV around the bridge obtained from point S5 in the three-dimensional backbone path I and three-dimensional backbone path II; Step 4: Sample the main path at equal intervals to generate a path that continuously covers the entire width of the bridge deck; In step four, the main path is divided into several equally spaced sub-segments based on adjacent points, and RRT is run within a 2.0 m cube local area around each sub-segment. * The algorithm finds a line segment with a lower wind pressure integral than the corresponding sub-segment path, and replaces the corresponding sub-segment to complete the local optimization of the segmentation. After splicing all the optimized segments, perform cubic B-spline smoothing to obtain a continuous and smooth path graph.

[0013] Preferably, in step three, the global guiding path is obtained by sequentially adding the point sequence from the initial path obtained by the A* algorithm into the tree and forming it through accumulated cost; The RRT * The algorithm iterates around the global guiding path, continuously optimizing local connections by re-completing lines to obtain the three-dimensional main path I; In a continuous cost field, the cost C(p1,p2) of the line segment A formed by points p1 and p2 is defined by the following formula: In the above formula, m is the midpoint of line segment A, and c wind (∙) represents the normalized wind pressure cost function. α As weight, and In the above formula, clip (∙) represents the clamping function. The pressure value at the query point. P min , P max These represent the minimum and maximum global pressure values, respectively. NaN This indicates an invalid value.

[0014] The present invention has at least the following beneficial effects: Firstly, this invention obtains the most unfavorable wind field pressure through coupled flow field analysis, which serves as a constraint for flight decision-making. Furthermore, based on quantitative wind field calculations, it is the first to incorporate the most unfavorable wind field pressure of a heavy-duty truck traveling in the opposite direction into the core basis of bridge inspection UAVs, filling a gap in existing research and improving flight safety.

[0015] Secondly, this invention directly links risk zoning with vehicle structural dimensions, enabling the bridge spatial risk level to be dynamically determined according to the range of vehicle disturbance, thus realizing the dynamic correlation between risk level and vehicle wind field, and enhancing the physical basis and engineering rationality of risk classification.

[0016] Third, this invention establishes a flight altitude selection mechanism that matches the wind resistance performance of UAVs with the risk area. By comparing and filtering the wind resistance threshold of the UAV with the wind field data of the bridge space, the flight altitude is matched with the performance parameters of the UAV, ensuring both safety and detection efficiency.

[0017] Fourth, regarding the flight path on the bridge, the drone only needs to perform a full-coverage scan to avoid the risk area according to its own performance to complete the crack detection on the bridge. This invention simulates the wind pressure field generated by the intersection of two trucks and divides the wind pressure gradient based on the wind pressure value. Users can flexibly select a suitable wind pressure gradient for flight according to the wind resistance performance and weight of the drone to avoid being drawn into the traffic flow due to the negative pressure generated by the vehicles. Regarding flight paths under bridges, the most significant impact comes from the negative pressure wind field generated on the sides of the bridge by vehicles passing over it. Therefore, this invention, after dividing the gradient, plans a path that avoids this high-risk area. By avoiding this area, the wind pressure generated by vehicles has a smaller impact and almost no effect on the drone's flight status. Thus, planning a flight path close to the underside of the bridge results in more accurate images. In other words, the path planned by this invention allows the drone to avoid high-risk areas, preventing crashes, loss of control, and large-angle rolls and swaying. Although the path length may increase, it avoids the risk of higher energy consumption.

[0018] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the system flow for the bridge inspection drone detection distance determination method of the present invention; Figure 2 This is an ANSYS model diagram of the bridge inspection UAV detection distance determination method of the present invention; Figure 3 This invention presents a three-dimensional wind field transverse pressure distribution diagram around a bridge under the condition of heavy-duty trucks traveling in opposite directions. Figure 4 This is a three-dimensional wind field pressure distribution along the bridge around the bridge under the condition of heavy-duty trucks traveling in opposite directions. Figure 5 This invention presents a three-dimensional wind field transverse pressure curve of the bridge centerline under the condition of heavy-duty trucks traveling in opposite directions. Figure 6 In Example 2, a continuous footprint full-coverage round-trip path is obtained by generating a two-dimensional grid point sequence using a deterministic method; Figure 7 The three-dimensional trunk path obtained by applying Gaussian filtering in the traversal order in Example 2; Figure 8 In Example 2, a continuous smooth path diagram is obtained by splicing together all the optimized sub-segments. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0021] A method for determining the detection distance of a bridge inspection drone based on wind field pressure distribution is proposed. This method obtains the three-dimensional wind field pressure distribution around the bridge through numerical simulation, establishes a spatial risk level classification model, and combines this with the drone's wind resistance performance parameters, such as... Figure 1 As shown, it includes the following steps: Step 1: Establish a coupled flow field model of heavy-duty trucks and bridges, and construct a three-dimensional computational domain that includes the bridge structure and the heavy-duty trucks traveling in opposite directions. In step one, a coupled flow field model of the heavy-duty truck and the bridge is established using ANSYS software. The most unfavorable working condition is selected as the opposite direction of a six-axle semi-trailer tractor. The parameters of the coupled flow field model are set as follows: 1. Bridge type: A typical highway simply supported beam bridge is selected, which conforms to the standard dimensions for highway transportation.

[0022] 2. Heavy-duty trucks: The vehicle height H is the maximum structural outline height of a heavy-duty truck. H serves as the benchmark parameter for spatial stratification and wind field risk assessment in this invention. It should be noted that, based on continuous statistical analysis of traffic flow on typical highway simply supported beam bridge sections, heavy-duty trucks account for a relatively high proportion of the overall traffic flow, with six-axle semi-trailer tractors accounting for approximately 8% to 10% of the total traffic flow. Although this proportion is insufficient to constitute the main vehicle type, it has a stable frequency of occurrence in traffic flow and is prone to generating high-intensity wake disturbances under two-way oncoming traffic conditions. Due to the large frontal area and obvious vehicle boundary layer separation characteristics of six-axle semi-trailer tractors, strong turbulent wake regions are formed in the rear and side areas during their operation. These wake regions interact with the bridge structure within the limited width of the bridge, thereby affecting the pressure distribution in the space above the bridge deck, the lateral space of the bridge, and the space under the bridge. Therefore, it is fully reasonable to model and analyze six-axle semi-trailer tractors as typical disturbance sources in bridge-vehicle coupled flow field analysis.

[0023] 3. Computational Domain: The length of the computational domain in the longitudinal direction of the bridge is not less than 3 times the bridge span. The computational domain is discretized using structured or semi-structured meshes, and local mesh refinement is performed on the vehicle surface and the area near the bridge deck.

[0024] 4. Turbulence Model: The Realizable-K-∑ model is selected, and a turbulence model suitable for complex separated flows and near-wall flows is used for solution.

[0025] Step 2: Perform numerical simulation calculations on the coupled flow field model to obtain three-dimensional wind pressure distribution data around the bridge; Step 3: Using the height H of the heavy-duty truck as a benchmark, classify the wind field risk levels of the space above the bridge, below the bridge, and the space on the side of the bridge. Based on the wind pressure distribution results, using the height H of a heavy-duty truck as the spatial stratification benchmark, a proportional risk level classification is performed for the spaces above, below, and lateral to the bridge. This risk level classification includes: (1) Area above the bridge Areas with an altitude range of 0 to 1.8H are designated as extremely high-risk areas; Areas with a B-degree range of 1.8H to 2.5H are designated as high-risk areas; Regions with a height C greater than 2.5H are designated as stable regions; (2) Area under the bridge The area between 0 and H in height from the edges of both sides of the bridge toward the center is designated as a high-risk area, and the remaining area is a low-risk area.

[0026] (3) Lateral spatial area of ​​the bridge In the vertical direction, the area from 0 to 1.2H is designated as a high-risk area, and the other areas are designated as low-risk areas; In the horizontal direction, the area 0 to 1.2H from the side edge of the bridge is designated as a high-risk area, and other areas are designated as low-risk areas.

[0027] Step 4: Within the safe flight range, perform spatial gradient calculation on the wind and pressure fields of the corresponding space regions to obtain pressure gradient vector field distribution data. Calculate the pressure gradient amplitude of each space unit based on the pressure gradient vector field, compare the pressure gradient amplitude with a preset gradient threshold, and select space unit regions with pressure gradient amplitudes less than the preset gradient threshold as low-risk flight areas.

[0028] Step 5: Based on the performance parameters of the drone, determine the safe detection distance of the drone. The altitude only involves the area above the bridge, and the detection distance involves three directions: the area above the bridge, the area under the bridge, and the side area. The performance parameters refer to the UAV's volume parameters, mass parameters, and wind resistance performance parameters. The safe detection distance is determined by dividing the UAV into different wind resistance levels and matching the corresponding safe flight altitude range based on the risk level classification of the wind field in the bridge space. The wind resistance is determined based on the UAV's maximum wind speed parameter or maximum tolerable wind pressure parameter, and a preset threshold is determined based on the wind field pressure parameter of the corresponding risk zone in the bridge space. When the wind resistance of the drone is lower than a preset first threshold, its flight altitude is limited to the stable area above the bridge, which is a region with a height greater than 2.5H. B. When the wind resistance capability of the drone is not lower than the first threshold, its flight altitude is limited to the medium-risk area and the stable area above the bridge, wherein the medium-risk area is the region with an altitude of 1.8H to 2.5H; When C flies under the bridge, the vertical distance between the drone and the bottom of the bridge shall not be less than H.

[0029] Example 1: 1. Taking the oncoming traffic of heavy-duty trucks as the most unfavorable working condition, establish a system as follows: Figure 2 The coupled flow field model of the heavy-duty truck and the bridge shown includes the following sub-steps in its implementation: A selects a medium-span highway bridge as the research object. The bridge is a simply supported beam structure, with a total length of 40m, a deck width of 9.5m, a deck thickness of 1.5m, a main beam height of 2.5m, and a clearance height of 7m. The bridge adopts a two-way single-lane layout. B simplifies the geometric model of the six-axle semi-trailer tractor, setting its parameters as follows: vehicle width 2.5m, vehicle length 7m, main body height 2.5m, and tire height from the bridge surface 0.5m. Therefore, the height of the vehicle's top from the bridge surface is 3.0m. C. Construct a three-dimensional coupled flow field model of the bridge and heavy-duty trucks. In the model layout, two six-axle semi-trailer tractors are positioned at the center line of the two-way single lane of the bridge to simulate the oncoming traffic meeting situation.

[0030] Furthermore, to unify the spatial stratification standard, the vehicle body height of 2.5m is defined as the reference height H for subsequent calculations of bridge spatial stratification and risk classification; when it comes to the relative spatial height of the bridge deck, the height of the vehicle top from the bridge deck of 3.0m is used as the reference value for the actual disturbance impact height.

[0031] Based on the aforementioned bridge structural parameters and vehicle model parameters, a three-dimensional coupled flow field model of the bridge and heavy-duty trucks was established. The model incorporates the bridge structure, bridge deck, and two opposing six-axle semi-trailer tractors into the computational domain. The two vehicles are positioned at the centerline of a single lane in both directions to simulate oncoming traffic. Since the total width of the bridge deck is 9.5m, the lateral distance between the two vehicles is small during oncoming traffic, leading to compression and superposition of the vehicle's lateral wake regions, creating a relatively unfavorable disturbance condition and thus obtaining more representative wind field distribution data. The computational domain extends five times the bridge length both forward and backward along the traffic direction, with a lateral width set to three times the bridge deck width and a vertical height set to 6H above and below the bridge deck to ensure sufficient flow field development and reduce boundary influences. The computational domain is discretized using structured or semi-structured meshes, with local mesh refinement applied to the vehicle surfaces and areas near the bridge deck.

[0032] 2. Perform numerical simulation calculations on the coupled flow field model to obtain, for example... Figure 3 The diagram shows the three-dimensional wind field transverse pressure distribution around the bridge. Based on this diagram, the wind field risk level above the bridge is classified as follows: For the space above the bridge deck, with the bridge deck as the zero point height, it is divided into three levels: 0–1.8H (0–4.5m), 1.8H–2.5H (4.5–6.25m), and above 2.5H (above 6.25m). The 0–1.8H range corresponds to the area directly affected by disturbances above the vehicles, the 1.8H–2.5H range is the wake diffusion and attenuation zone, and the area above 2.5H is the relatively stable airflow field zone.

[0033] 3. Perform numerical simulation calculations on the coupled flow field model to obtain, for example... Figure 4 The diagram shows the 3D wind field pressure distribution along the bridge, and based on this diagram, the wind field risk levels for the lateral and sub-bridge spaces are classified as follows: For the lateral space of the bridge, based on the outer edge of the bridge side guardrail, it is divided into two sections along the lateral extension direction: 0 to 1.2H and above 1.2H. Since the bridge deck is relatively narrow, the lateral space disturbance is more concentrated when vehicles are passing each other. Therefore, the 0 to 1.2H section is usually a high-risk disturbance area.

[0034] The space under the bridge is divided into sections from 0 to H and below H, based on the lowest point of the bridge structure. The space under the bridge is significantly affected by the bridge structure, but when vehicles travel at high speeds, the airflow forms a backflow zone at the bridge edge, which still creates some pressure fluctuations within the 0 to H range under the bridge.

[0035] By adopting the above-mentioned hierarchical approach, a correspondence between "vehicle scale - spatial scale - risk level" is established, giving the risk classification a clear physical meaning rather than arbitrarily setting the partition height, thereby enhancing the rationality and interpretability of the technical solution.

[0036] 4. Perform numerical simulation calculations on the coupled flow field model to obtain, for example... Figure 5 The three-dimensional wind field transverse pressure curve around the bridge under opposing traffic conditions is shown in the diagram. This curve is used to screen low-risk flight zones. The specific implementation method is as follows: After completing the model construction, a three-dimensional turbulence numerical simulation method was used to calculate the coupled flow field, with the incoming wind velocity set to 0 m / s and the vehicle speed set to 60 km / h. Three-dimensional wind field pressure distribution data around the bridge were obtained through numerical calculation, and pressure and pressure gradient data were extracted for the space above the bridge deck, the space laterally to the bridge, and the space under the bridge. The calculation results show that under the condition of oncoming traffic, the pressure gradient in the range of 0–4.5 m above the bridge deck is approximately 7.5 Pa / m, and there is a significant wake superposition phenomenon in the 0–4.5 m region on the side of the bridge. The wind field gradually stabilizes when the height exceeds 6.25 m.

[0037] Furthermore, after obtaining the spatial wind field distribution data, the safe flight altitude range is determined by combining the performance parameters of the UAV. In this embodiment, a small rotary-wing UAV for bridge inspection is selected, with a mass of approximately 246 g. Its maximum permissible vertical wind speed is 5–8 m / s, and the maximum permissible vertical pressure gradient for safe flight is approximately 5–7 Pa / m. Comparison and analysis of the UAV's wind resistance threshold with the bridge's spatial wind field data reveals that the pressure gradient in the 0–1.8H region above the bridge deck is approximately 7.5 Pa / m, unsuitable for flight; the pressure gradient in the 1.8H–2.5H region is approximately 4.3 Pa / m, within a tolerable range; and the pressure gradient above 2.5H is approximately 1.2 Pa / m, indicating a more stable wind field. Considering both the bridge inspection operation distance requirements and flight safety requirements, the safe flight altitude range for the UAV is determined to be above 6.25 m above the bridge deck.

[0038] Furthermore, pressure gradient distribution data for the corresponding spatial region was extracted and analyzed. The results showed that beyond 5m from the bridge edge, the pressure gradient was below 1.2 Pa / m, classifying it as a low-risk flight zone, while strong wake disturbances still existed within 0–5m of the bridge edge. Therefore, the area beyond 5m from the bridge edge was selected as the preferred deployment area for UAV inspection routes. In actual flight verification, the amplitude of UAV attitude angle fluctuations was significantly reduced within this area, and image acquisition stability was improved.

[0039] Example 1 determines the optimal detection path for a UAV within a specific flight altitude and safe distance range, enabling the UAV to stably acquire high-precision target information, achieve accurate identification and classification of target features, effectively reduce operational risks, improve detection accuracy and efficiency, and provide a reliable basis for operational decisions. It should be noted that the optimal detection path here refers to avoiding areas with excessive wind pressure, preventing the UAV from going out of control or crashing, ensuring that the UAV always maintains a suitable detection distance and shooting angle with the bridge surface, and avoiding areas with strong wind disturbances, so that the images are clearer and the measurements are more accurate, thereby supporting the accurate identification and classification of defects, minimizing the total flight mileage, reducing invalid climbs and turnarounds, and saving power and time.

[0040] A path planning method for bridge inspection UAVs based on CFD wind pressure field modeling and an improved A*-RRT*+LRM algorithm is proposed. The method employs the Realizable-K-∑ turbulence model in ANSYS Fluent to perform high-precision numerical simulation of the flow field around the bridge, constructing a continuous wind pressure cost map. Secondly, considering the different environmental constraints above and below the bridge, a hierarchical planning architecture is designed, using an A*-RRT* hybrid algorithm. A* is used to quickly generate a global guiding path, followed by RRT* refinement optimization. Shape constraint discrete sampling and Gaussian smoothing are combined to achieve structural fit. Finally, a local refinement module is introduced to perform segmented RRT* optimization of the scanning path. Multiple ablation and comparative experiments under wind pressure fields show that the proposed method achieves optimal or near-optimal levels in six dimensions: path length, wind pressure integral, average wind pressure, maximum wind pressure, curvature, and planning time. This provides an effective solution for autonomous path planning of bridge inspection UAVs in complex wind pressure environments, effectively reducing operational risks and improving detection accuracy and efficiency.

[0041] Specifically, the path planning method for the bridge inspection UAV includes: Step 1: Simulate the wind pressure distribution characteristics under heavy-duty vehicle driving conditions using a coupled flow field model; In step one, a coupled flow field model of the heavy-duty truck and the bridge is established using ANSYS Fluent numerical analysis software to simulate the wind pressure distribution characteristics under the driving conditions of heavy-duty vehicles, and to obtain the wind pressure distribution cloud map and the pressure change law of key sections. In the model layout, two six-axle semi-trailer tractors are located at the center line of the two-way single lane of the bridge to simulate the oncoming traffic conditions.

[0042] A 3D geometric model of the bridge was created in the Workbench platform, and a sufficiently large external fluid domain was constructed. The fluid domain was discretized using an unstructured mesh. In the Fluent solver, a pressure-based solver was selected, and double precision was enabled. The Realizable-K-∑ turbulence model was used, which has better predictive capabilities for rotating flows and separated flows with strong adverse pressure gradients, and is suitable for complex flows such as those around bridges.

[0043] After convergence, the spatial coordinates (X, Y, Z) and pressure values ​​of all grid nodes are saved as an Excel file using the data export function. To ensure the accuracy of subsequent interpolation, a suitable interpolation method needs to be selected in subsequent modules based on the data distribution. The wind pressure cost map constructed by this module provides an environmental awareness basis for subsequent path planning. This modeling method can accurately reflect the wind pressure differences on the bridge surface, providing reliable environmental data support for the effectiveness of subsequent path planning.

[0044] Step 2: The discrete wind pressure data output by the coupled flow field model is converted into a continuous cost field that can be directly queried in the path planning by using spatial interpolation technology. In step two, the discrete wind pressure data output from the ANSYS Fluent simulation is converted into a continuous cost field that can be directly queried in path planning using the environmental cost map builder module. The environmental cost map builder module first reads the data using pandas and extracts the coordinate matrix points (N x 3) and the pressure orientation (N). To select the optimal interpolation strategy, this module extracts the unique values ​​for each dimension of the coordinates. If the total number of data points equals the product of the number of unique values ​​in each dimension, it is determined to be a structured mesh; otherwise, it is an unstructured mesh.

[0045] In practice, tensor product linear interpolation is used for data with a regular grid distribution. Let the grid node coordinates be as follows: , , Pressure value p ijk Corresponding node (x) i ,y j ,z k The stress estimate for any query point (x, y, z) As shown in the following formula: in, , , All are one-dimensional linear interpolation basis functions, and We obtain it from the following formula: And for , The same function is then used to obtain the result. SciPy employs piecewise linear-multilinear interpolation over the computational domain, performing linear interpolation in the order of x, y, z directions to ensure both computational efficiency and the continuity of the interpolation function.

[0046] For unstructured meshes, linear interpolation based on Delaunay triangulation is used. The data point set is partitioned into a three-dimensional Delaunay tetrahedron, dividing the space into a series of non-overlapping tetrahedra. Using four data points as vertices of a tetrahedron, for any query point x = (x, y, z), its corresponding tetrahedron is located. The pressure value of that point can then be obtained by linearly combining the pressure values ​​of the four vertices of that tetrahedron using the centroid coordinates.

[0047] Step 3: The purpose of generating the initial path is to provide high-quality initial guidance for subsequent trunk path optimization. Since the detection task covers both the area above and below the bridge and the environmental constraints of the two are significantly different, different initial path generation strategies need to be designed for the area above and below the bridge.

[0048] Design Improvement A * -RRT * Fusion algorithm (it should be noted that A) * The algorithm refers to the Wind Adaptive Lifetime Programming (WA-LPA) proposed by Lian et al. * RRT * The algorithm refers to an asymptotically optimal RRT (Randomized Fast Exploratory Tree) that improves upon the RRT. * The algorithms are all existing technologies, so this step will not describe their specific operation process in detail, but only integrate the two algorithms, that is, through A. * The algorithm provides a coarse path that takes into account both distance and wind pressure. However, this path is limited by grid discretization and may not fully utilize low-wind-pressure areas in the continuous space, and the path smoothness is insufficient. Therefore, the optimization of the main path on the bridge adopts an asymptotically optimal fast exploratory random tree (RRT) approach. * The algorithm performs global optimization in a continuous space (minimizing wind pressure integral cost while maintaining a short flight distance). For bridge detection, a three-dimensional grid A is used. * The algorithm generates a global guiding path from a continuous cost field as the initial tree node for the RRT* algorithm (i.e., incorporating wind pressure cost into the standard RRT* framework to significantly accelerate the convergence process, ensuring that the optimized path maintains the overall direction and can be locally adjusted to avoid high wind pressure areas), and obtains the three-dimensional trunk path I through iterative optimization. Specifically, it uses a three-dimensional grid A. * The algorithm discretizes the space, uniformly dividing the 3D environment into grids with a resolution of r = 2.0 m. Choosing the grid center as the node coordinate reduces the positional error caused by discretization while ensuring that the path point always remains within the flyable space. Movement between grids allows for 26 neighborhoods, including 6 principal axis directions (forward, backward, left, right, up, down) and 12 face diagonal directions (upper left, right front, etc.). In 3D space, the UAV can fly in any direction, allowing all possible movements and avoiding unnecessary steering constraints introduced by discretization. To ensure that the initial path prioritizes paths with lower wind pressure for the same distance and shorter paths for the same wind pressure, the single-step cost C(p1,p2) from grid node p1 to grid node p2 is defined as:

[0049] In the above formula, Let m be the Euclidean distance from grid node p1 to grid node p2, representing flight energy consumption. m = (p1 + p2) / 2 is the midpoint of the line segment formed by grid nodes p1 and p2 (by sampling at the midpoint, the wind pressure cost of the entire edge can be approximated as equal to the wind pressure cost at the midpoint), representing the average wind pressure impact of the entire edge. c wind (∙) is the normalized wind pressure cost function provided by the environmental cost map construction module, and , P min , P max These represent the minimum and maximum global pressure values, respectively. clip (∙) is a clamping function, and the 0 and 1 in the parentheses refer to the upper and lower limits of the value, used to restrict the result to between 0 and 1, ensuring the validity of the cost. If the calculated value is between 0 and 1, the value is retained; if it exceeds the upper limit, the upper limit value is returned; if it is below the lower limit, the lower limit value is returned. This performs three-dimensional... Delaunay Tetrahedral partitioning divides space into a series of non-overlapping tetrahedra. Using four data points as vertices of a tetrahedron, for any query point x = (x, y, z), the tetrahedron containing that query point is located. The pressure value at the query point; NaN NaN is an abbreviation for Not a Number. NaN represents an invalid value. In formulas, if... If the result is a valid value, normalization is performed; otherwise, a constant of 0.1 is assigned. α This is the wind pressure weighting coefficient (or simply weight), used to achieve a reasonable balance between distance and wind pressure in the path. By sampling at the midpoint, it can be approximated that the wind pressure cost of the entire edge is equal to the wind pressure cost at the midpoint. Single-step cost. It can be used to ensure that the A* algorithm can find the optimal path under the grid graph and cost function, and also to ensure search efficiency and avoid blind diffusion. RRT* is guided by the A* path, and the initial nodes of the tree are already distributed near a low-cost path from the starting point to the goal, which greatly improves the convergence speed.

[0050] For under-bridge detection, the UAV must perform precise detection along the width of the bridge deck. Therefore, a deterministic traversal method is used to generate a sequence of two-dimensional grid points covering the entire under-bridge area from a continuous cost field, and three-dimensional attributes are assigned through height optimization. This ensures that each (x,y) position is visited. Separating planar coverage and height optimization makes each sub-problem easier to optimize, resulting in the three-dimensional main path II. It should be noted that the height optimization refers to: Let the sampling step size I along the length direction (x-axis) of the bridge be d. x The sampling step size II along the bridge width direction (y-axis) is d. yY-direction coverage area The number of sampling points is specifically represented by the following formula: For each sampling point x in the X direction i Take the Y sequence in order, and generate the sequence (x) i ,y i The height is determined by unidirectionally increasing in the X direction and cyclically moving back and forth in the Y direction. The output two-dimensional coordinates, combined with the conditional constraints and wind pressure optimization, determine the height. The Z value with the minimum wind pressure and shape penalty is selected, and then Gaussian smoothing is applied to obtain the final height.

[0051] As can be seen from the above, this step will involve A. * The algorithm adds the initial path's point sequence to the tree sequentially and accumulates the cost to form an initial feasible path. Subsequent RRT... * The algorithm iterates around this path, continuously optimizing local connectivity by re-completing lines. The advantage of this deeply integrated hybrid planning strategy lies in A * The algorithm is RRT * The algorithm provides a clear search direction, avoiding blind sampling from zero, and significantly improving the convergence speed. RRT * The algorithm compensates for A * Despite the accuracy loss caused by discretization, the algorithm can achieve fine local adjustments in continuous space. Furthermore, this step has already taken wind pressure factors into account, and the output supports the subsequent path refinement module. For environments with complex constraints, intelligent search and deterministic traversal are used to ensure both path quality and computational efficiency.

[0052] It should be noted that this step includes two aspects of path planning: firstly, the flight path planning of the UAV above the bridge surface (i.e., Figures 6-7 The two aspects are: firstly, optimizing the path on the bridge (wind pressure optimization); and secondly, planning the flight path for UAV detection under the bridge (i.e., Figures 6-7 The optimization path under the bridge (wind pressure optimization) is as follows. Specifically, for the flight path planning of UAVs above the bridge, according to the pressure cloud map calculated by numerical simulation, the UAV needs to maintain a certain safe distance in altitude from vehicles passing by on the bridge to avoid being interfered with or even crashing under the action of strong wind pressure. Therefore, the flight path planning of UAVs above the bridge itself includes the element of maintaining a safe distance. Regarding the flight path of drones under bridges, the traditional flight path is to fly along a plane in a "bow" shape (e.g., Figures 6-7The image shows the original path of the UAV under the bridge. However, according to the pressure cloud map calculated by numerical simulation, the wind pressure is higher at the two sides of the bridge edge and lower in the middle. Therefore, this step plans the flight path of the UAV under the bridge as a "W" shaped flight path. The UAV flies along this flight path, which keeps it away from the high wind pressure area, ensuring its flight stability and guaranteeing the detection accuracy. This path planning also reflects the element of safe distance.

[0053] Therefore, regardless of whether the inspection is conducted on or under the bridge, the final inspection path = flight altitude + flight route. The flight altitude (i.e., the UAV detection distance for bridge inspection) is determined using a pressure cloud map calculated through numerical simulation, and the flight route is based on the aforementioned improved A... * -RRT * Obtained by fusion algorithm or deterministic traversal method.

[0054] Step 4: Sample the main path at equal intervals to generate a path that continuously covers the entire width of the bridge deck; Step four mainly involves transforming the optimized 3D backbone path into a scanning trajectory capable of fully covering the bridge surface. This is achieved by dividing the backbone path into several equidistant segments based on adjacent points, and then running RRT within a 2.0 m cube local area around each segment. * The algorithm finds a line segment with a lower wind pressure integral than the corresponding sub-segment path, and replaces the corresponding sub-segment to complete the local optimization of the segmentation; after splicing all the optimized sub-segments, it performs cubic B-spline smoothing to obtain a continuous and smooth path map.

[0055] Specifically, to transform the path into a continuous curve model, the cumulative path length for each point needs to be obtained by sequentially accumulating the Euclidean distance between adjacent points along the path, starting from the starting point. The total path length is the cumulative length value of the last point. Equidistant sampling is performed along the main path with a fixed step size. The number of sampling points is determined by dividing the length of the main path by the sampling step size. The sequence of sampling points determines the distribution density of the scan line along the bridge's length. A reciprocating alternating strategy is used for the scanning direction: the direction variable is initialized to positive, and for the m-th sampling point... Points are generated sequentially, and if the direction is reversed, they are generated in reverse order. This mode ensures continuous connection between adjacent scan lines, avoids large-scale jumps by the UAV, and achieves a complete scan of the entire bridge surface. However, the scan path usually consists of a large number of straight line segments, with approximately 90° bends at the corners. This path, as a flight trajectory, causes frequent acceleration and deceleration of the UAV, affecting the stability of the detection. Therefore, cubic B-spline interpolation is used to smooth the scan path, and node vector homogenization is defined to ensure that the curve maintains its state near the control points while having good second-order continuity. Uniform resampling within the node intervals can eliminate the bends and inflections of the path and make the path curvature smoother.

[0056] Based on the path planning results for UAV bridge inspection, the path length L and wind pressure integral are used. I wind Average wind pressure and maximum instantaneous wind pressure c max Conduct quantitative evaluation.

[0057] The total path length L refers to the sum of the Euclidean distances between all adjacent grid nodes on the path, reflecting the total distance and time of the UAV flight mission. The lower the value, the higher the flight efficiency and the less energy consumed, as shown in the following:

[0058] In the above formula, N This represents the number of raster nodes on a path when calculating the total length L of that path. p i+1 Indicates the first i +1 grid node position, p i Indicates the position of the i-th grid node, | |pi +1- pi || represents the Euclidean distance between two grid nodes, and the total distance (i.e., the total path length) corresponding to the sum of the Euclidean distances. L This reflects the time and energy consumption of the flight mission; Wind pressure integral I wind This refers to the line integral of the normalized wind pressure cost path, which is the weighted accumulation of wind pressure over the path length. It reflects the total wind pressure exposure experienced by the UAV throughout the entire flight. The lower the value, the lower the overall flight risk, and the more successfully the path avoids high wind pressure areas. Specifically, it is expressed as follows:

[0059] in, Let c be the midpoint of the i-th line segment. wind (∙) represents the normalized wind pressure cost function.

[0060] Average wind pressure This refers to the ratio of wind pressure integral to path length, i.e., the average wind pressure per unit length. It reflects the overall risk density of the path. The lower the value, the more the path tends to be in a low-wind-pressure area, thus eliminating the influence of path length on risk assessment. Specifically, it is expressed as follows: Maximum instantaneous wind pressure c max This refers to the maximum wind pressure cost at the midpoint of all line segments along the path, reflecting the extreme wind pressure risks that the drone may encounter. The lower the value, the more the path avoids local high wind pressure areas, serving as a safety indicator. Specifically, it is expressed as follows: Example 2: 1. Environmental and data preparation was carried out using Example 1; 2. Path planning In wind pressure field modeling, the first step is to read the file exported by Fluent and parse the spatial coordinates and pressure values. After reading the data, the network structure needs to be determined, specifically as follows: Read the unique values ​​of each dimension X, Y, and Z. If the total number of points is equal to the product of the number of unique values ​​in the three dimensions, it is judged as a structured mesh, and a tensor product linear interpolator is constructed using SciPy. Otherwise, it is judged as an unstructured mesh, and linear interpolation is performed using Delaunay triangulation.

[0061] After constructing the wind pressure field interpolator, the global maximum and minimum pressure values ​​are statistically obtained from CFD data, and normalization is performed on any query point. During the actual interpolation process, for query points that may be located outside the data convex hull or beyond the grid boundary, a small cost constant is assigned to these points to ensure the continuity of the path search. Through normalization and anomaly handling, the physical simulation data is transformed into a unified cost map, providing an environmental awareness foundation for subsequent path planning.

[0062] The initial path is generated using a 3D raster A. * Algorithm Example 2 discretizes the space into a grid with a resolution of 2.0, the number of grids being calculated based on the bridge's extent. Each grid acts as a node with 26 neighbor connections. A priority queue is used to search from the starting grid to the ending grid, resulting in a coarse path, which will serve as the initial tree node for subsequent RRT*.

[0063] With A * The algorithm's path is used as the initial tree, and RRT is run. * The algorithm iterates 2000 times with a step size of 2.0m, a search radius of 5.0m, and a target bias probability of 0.1. The height is allowed to fluctuate within ±1.0m of the base height, but must not exceed the global safe height. Conditional constraints are added under the bridge. In each iteration, random sampling is performed to find the nearest node. New nodes are generated by expanding towards the sampling point. Nearby nodes are searched within the radius, and the parent node that minimizes the cumulative cost is selected. If passing through a new node reduces the cost of neighboring nodes, its parent node is updated. Finally, the optimized path is obtained by backtracking from the target node.

[0064] like Figure 6 As shown, a continuous footprint full-coverage round-trip path is obtained by generating a two-dimensional grid point sequence using a deterministic method. Gaussian filtering is applied in traversal order to eliminate abrupt changes between adjacent points, ultimately resulting in a smooth path. Figure 7 The three-dimensional trunk path is shown.

[0065] Furthermore, equidistant sampling is performed along the optimized main path, generating a path that continuously covers the entire bridge deck width. The path is then segmented into local optimizations, dividing it into several short segments based on adjacent points. For each segment, a local optimization technique (RRT) is run within a 2.0 m cube-shaped local area. * The algorithm replaces a segment if the found path has a lower wind pressure integral than the original straight segment; otherwise, it retains the original straight segment. All optimized segments are then concatenated to form the final path. Finally, the path is smoothed using cubic B-spline smoothing to increase the number of points and ensure curvature continuity, facilitating UAV flight. The result is shown below. Figure 8 The diagram shows a continuous smooth path.

[0066] The above solution is merely an illustration of a preferred example and is not limited thereto. When implementing this invention, appropriate substitutions and / or modifications can be made according to the user's needs.

[0067] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. It can be applied to various fields suitable for the present invention. Other modifications can be readily made by those skilled in the art. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and examples shown and described herein.

Claims

1. A method for determining the detection distance of a bridge inspection UAV based on wind field pressure distribution, characterized in that, include: S1. Establish a coupled flow field model of heavy-duty trucks and bridges, and the coupled flow field model includes a three-dimensional computational domain of the bridge structure and the heavy-duty trucks traveling in opposite directions. S2. Perform numerical simulation calculations on the coupled flow field model to obtain three-dimensional wind pressure distribution data around the bridge, whereby the bridge perimeter includes the vertical and lateral spatial regions of the bridge. S3. Based on three-dimensional wind field pressure distribution data, and taking the height H of the heavy-duty truck body as the benchmark, the wind field risk level around the bridge is classified to obtain the corresponding safe flight altitude range. S4. Calculate the pressure gradient distribution data of the spatial region corresponding to the safe flight altitude range, and select low-risk flight areas from them. S5. In a low-risk flight area, based on the performance parameters of the UAV, determine the safe detection distance of the UAV around the bridge, wherein the height in the safe detection distance only involves the area above the bridge.

2. The method for determining the detection distance of a bridge inspection UAV based on wind field pressure distribution as described in claim 1, characterized in that, In S1, the coupled flow field model is based on the six-axle semi-trailer tractor traveling in opposite directions, established using ANSYS software. The coupled flow field model adopts the Realizable-K-∑ model, and the coupled flow field is calculated using the three-dimensional turbulence numerical simulation method. The length of the bridge in the longitudinal direction in the three-dimensional computational domain is not less than three times the bridge span, and the three-dimensional computational domain is discretized using structured or semi-structured meshes, with local mesh refinement applied to the vehicle surface and the area near the bridge deck.

3. The method for determining the detection distance of a bridge inspection UAV based on wind field pressure distribution as described in claim 1, characterized in that, In S3, the wind field risk level classification method for the area above the bridge is as follows: Areas with an altitude range of 0 to 1.8H are designated as extremely high-risk zones; Areas with an altitude range of 1.8H to 2.5H are designated as high-risk areas; Regions with a height greater than 2.5H are designated as stable regions; For the area under the bridge, the area with a height range of 0 to H from the two sides of the bridge towards the middle is designated as a high-risk area, and the remaining area is a low-risk area. For the lateral space area of ​​the bridge, the area from 0 to 1.2H in the vertical direction is designated as a high-risk area, and the other areas are designated as low-risk areas; In the horizontal direction, the area 0 to 1.2H from the side edge of the bridge is designated as a high-risk area, and other areas are designated as low-risk areas.

4. The method for determining the detection distance of a bridge inspection UAV based on wind field pressure distribution as described in claim 1, characterized in that, In S4, the method for selecting low-risk flight areas is as follows: S40. Calculate the spatial gradient of the wind and pressure fields in the space region corresponding to the safe flight altitude range to obtain the distribution data of the pressure gradient vector field. S41. Calculate the pressure gradient amplitude of each spatial unit based on the pressure gradient vector field; S42. Compare the pressure gradient amplitude with a preset gradient threshold, and designate the space unit with a pressure gradient amplitude less than the preset gradient threshold as a low-risk flight area.

5. The method for determining the detection distance of a bridge inspection UAV based on wind field pressure distribution as described in claim 1, characterized in that, In S5, the performance parameters include: volume parameters, mass parameters, and wind resistance performance parameters; The safe detection distance is determined by comparing the wind resistance capability of the drone with a preset benchmark threshold. If the wind resistance capability is less than the preset benchmark threshold, the drone's flight altitude is limited to the stable zone above the bridge; otherwise, the drone's flight altitude is limited to the medium-risk zone and the stable zone above the bridge. When the drone flies under the bridge, the vertical distance between the drone and the bottom of the bridge shall not be less than H; The wind resistance capability is determined based on the maximum wind speed parameter or the maximum wind pressure parameter that the UAV can withstand, and the preset benchmark threshold is determined based on the wind field pressure parameter of the corresponding risk zone in the bridge space.

6. A path planning method for a bridge inspection drone, which applies the bridge inspection drone detection distance determination method based on wind field pressure distribution as described in any one of claims 1-5, characterized in that, include: Step 1: Simulate the wind pressure distribution characteristics under heavy-duty vehicle driving conditions using a coupled flow field model; Step 2: The discrete wind pressure data output by the coupled flow field model is converted into a continuous cost field that can be directly queried in path planning using spatial interpolation techniques. Step 3, design the improved A * -RRT * For bridge detection, a fusion algorithm using a 3D grid A is employed. * The algorithm generates a global guiding path from the continuous cost field as the initial tree node of the RRT* algorithm, and obtains the three-dimensional trunk path I through iterative optimization; For under-bridge detection, a deterministic traversal method is used to generate a sequence of two-dimensional grid points covering the entire under-bridge area from a continuous cost field, and three-dimensional attributes are assigned through height optimization to obtain the three-dimensional backbone path II; Among them, the three-dimensional backbone path I and three-dimensional backbone path II include the safe detection distance of the UAV around the bridge obtained from S5; Step 4: Sample the main path at equal intervals to generate a path that continuously covers the entire width of the bridge deck; In step four, the main path is divided into several equally spaced sub-segments based on adjacent points, and RRT is run within a 2.0 m cube local area around each sub-segment. * The algorithm finds a line segment with a lower wind pressure integral than the corresponding sub-segment path, and replaces the corresponding sub-segment to complete the local optimization of the segmentation. After splicing all the optimized segments, perform cubic B-spline smoothing to obtain a continuous and smooth path graph.

7. The path planning method for a bridge inspection UAV as described in claim 6, characterized in that, In step three, the global guiding path is obtained by sequentially adding the point sequence from the initial path obtained by the A* algorithm into the tree and accumulating the cost. The RRT * The algorithm iterates around the global guiding path, continuously optimizing local connections by re-completing lines to obtain the three-dimensional main path I; In a continuous cost field, the cost C(p1,p2) of the line segment A formed by points p1 and p2 is defined by the following formula: In the above formula, m is the midpoint of line segment A, and c wind (∙) represents the normalized wind pressure cost function. α As weight, and In the above formula, clip (∙) represents the clamping function. The pressure value at the query point. P min , P max These represent the minimum and maximum global pressure values, respectively. NaN This indicates an invalid value.

Images