Blade defect detection method and system for wind generating set

Abstract

The invention discloses a wind generating set blade defect detection method and system, and relates to the technical field of blade defect detection. If the wind generating set is in the shutdown state, the unmanned aerial vehicle detection equipment obtains coordinate information and image information of blades of the wind generating set; planning a detection path of unmanned aerial vehicle detection equipment based on the coordinate information, and preliminarily analyzing the defect condition of the blade based on the image information; the unmanned aerial vehicle detection equipment moves along the detection path, and ultrasonic data are acquired by using an ultrasonic sensor; comprehensively analyzing the defect condition of the blade based on the ultrasonic data and the initial defect condition of the blade; the method is used for solving the problems that a traditional blade defect detection result is inaccurate, and the detection operation risk is high.

Description

Technical Field

[0001] This invention relates to the field of blade defect detection technology, specifically to a method and system for detecting defects in wind turbine blades. Background Technology

[0002] With the booming development of the new energy industry, my country's wind power sector has experienced rapid growth, with the power output and rotor size of individual wind turbines continuously increasing, and blade length jumping from the early 30-40 meter range to the 60-70 meter level. As the core component of wind turbine generators, blades bear the important mission of converting wind kinetic energy into mechanical energy. However, due to the complex structural design, precision manufacturing process, and long-term exposure to harsh conditions such as wind and rain erosion and salt spray corrosion, the formation of various damage defects is unavoidable. If these defects are not identified and addressed in a timely manner, they may lead to major safety accidents such as blade breakage, seriously threatening the stable operation of the wind power system.

[0003] In terms of inspection technology, traditional operating methods mainly rely on visual inspection with telescopes or manual sling inspection. This high-altitude operation not only poses personnel safety risks but also has inherent drawbacks such as long inspection cycles, high inspection costs, and high labor intensity. Although new inspection technologies based on machine vision and image analysis have begun to be applied in recent years, the single-dimensional analysis method that relies solely on surface image data is difficult to comprehensively and accurately identify internal structural defects in blades, thus limiting the reliability of inspection results.

[0004] Therefore, we propose a method that can reduce operational risks and improve detection accuracy. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for detecting defects in wind turbine blades, which solves the problems of inaccurate results and high risks in traditional blade defect detection.

[0006] This invention is achieved through the following technical solution:

[0007] A method for detecting defects in wind turbine blades, specifically including:

[0008] Periodically acquire the operating status of wind turbine generators;

[0009] If the wind turbine is in a shutdown state, the coordinate information and image information of the wind turbine blades will be obtained by the drone detection equipment;

[0010] The detection path of the UAV detection equipment is planned based on coordinate information, and the defect situation of the blade is preliminarily analyzed based on image information.

[0011] The drone inspection equipment moves along the inspection path while using ultrasonic sensors to acquire ultrasonic data;

[0012] Based on ultrasonic data and the initial defects of the blade, the defect situation of the blade is comprehensively analyzed.

[0013] Furthermore, the specific steps for the UAV detection equipment to acquire image information of wind turbine blades are as follows:

[0014] LiDAR point cloud data acquired by the lidar of the drone inspection equipment

[0015] Based on LiDAR point cloud data, the iterative nearest point algorithm is used in conjunction with the CAD model of the blade to calculate the actual pose of the blade.

[0016] Multispectral images of the leaves were acquired using a multispectral camera on a drone inspection device.

[0017] Furthermore, the iterative nearest-point algorithm, combined with the CAD model of the blade, is used to calculate the actual pose of the blade, and its objective function is:

[0018]

[0019] Where N is the total number of matching point pairs in the point cloud, i is the index of the current point pair, R is the rotation matrix, t is the translation vector, and p i For point cloud data, q i These are the corresponding points in the model.

[0020] Furthermore, the process of planning the detection path for the UAV inspection equipment based on coordinate information and conducting a preliminary analysis of the blade defects based on image information specifically includes:

[0021] The U-Net deep learning model is used to segment the leaf region. Based on multispectral images, the output leaf mask M(x,y)∈{0,1} is generated, where x and y are the horizontal and vertical coordinates of the image, respectively.

[0022] Extracting multispectral features f from pixels within the leaf mask i (x,y);

[0023] Based on historical defect data and multispectral features, a defect probability heatmap P(x,y) is generated.

[0024] Based on the defect probability heatmap P(x,y), the ant colony algorithm is used to plan the detection path.

[0025] Furthermore, the formula for calculating the defect probability heatmap P(x,y) is as follows:

[0026]

[0027] Where N is the number of multispectral features, i is the index of the current feature, and ω iThese are the weighting coefficients.

[0028] Furthermore, the specific steps for planning the detection path using the ant colony algorithm are as follows:

[0029] Define the objective function as:

[0030]

[0031] Where N is the total number of path segments that the UAV needs to detect, i is the sequence number of the current path segment, and d i Let λ1 and λ2 be the flight distance of path segment i, and P(x) be the weighting coefficients. i ,y i ) represents the defect probability value at the location covered by the i-th path segment;

[0032] The pheromone update rule is determined as follows:

[0033]

[0034] Where, τ ij (t) represents the pheromone concentration of the path segment i→j at time t, ρ represents the pheromone volatility coefficient, m represents the total number of ants, and k represents the current ant's sequence number. Let $\frac{k}{j}$ be the amount of pheromone released by the $k$-th ant on the path segment $i→j$, and its calculation formula is as follows:

[0035]

[0036] Q is a constant, L k Let be the total path length of the k-th ant.

[0037] Furthermore, the specific steps for acquiring ultrasonic data using an ultrasonic sensor are as follows:

[0038] Acquire echo data;

[0039] Preprocess the echo data and calculate the defect depth based on the echo data;

[0040] Based on the preprocessed echo data, echo energy characteristics are extracted, and correlation coefficients are calculated.

[0041] Furthermore, wavelet transform is used to filter out environmental noise, completing the preprocessing of the echo data.

[0042] Furthermore, the specific steps for comprehensively analyzing the blade's defect situation based on ultrasonic data and the initial defect status of the blade are as follows:

[0043] Update the defect thermal map based on echo energy characteristics;

[0044] The input feature vector is constructed based on echo energy characteristics, defect depth, and correlation coefficient.

[0045] Construct a decision function and calculate the defect category based on the input feature vector.

[0046] A wind turbine blade defect detection system includes a drone detection device, a detection path planning unit, a data acquisition unit, and a defect analysis unit;

[0047] The drone detection equipment includes the drone itself, as well as a lidar, a multispectral camera, and an ultrasonic sensor installed on the drone itself.

[0048] The detection path planning and analysis unit is used to plan the detection path of the UAV detection equipment based on coordinate information and to perform preliminary analysis of the blade defects based on image information.

[0049] The data acquisition unit is used to obtain the initial defect status of the blade and ultrasonic data;

[0050] The defect analysis unit is used to comprehensively analyze the defect status of the blade based on ultrasonic data and the defect status of the blade.

[0051] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0052] This invention discloses a method and system for detecting defects in wind turbine blades. By using a drone inspection device, it is possible to automatically complete high-altitude blade defect detection operations. Workers only need to maintain the drone equipment on the ground, which improves the safety of workers.

[0053] By combining image information and ultrasonic data to determine the defects in the blades, the accuracy of the detection results can be improved. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of a method flow of the present invention;

[0055] Figure 2 This is a schematic diagram of a system structure according to the present invention;

[0056] Figure 3 This is a schematic diagram of the electronic device in this invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0058] Example 1

[0059] like Figure 1 The method for detecting defects in wind turbine blades, as shown, specifically includes:

[0060] Periodically acquire the operating status of wind turbine generators;

[0061] Since wind turbines rely on wind power for operation, but wind direction can change with seasons or climate, and blade defects may only appear after a period of operation, the acquisition cycle is generally two months. If the last test ends two months later, the operation of the wind turbine will be tested again. If the wind turbine is still running, it is still in the preparation stage of the test cycle. If the wind turbine is shut down, the test stage of the test cycle will begin, which is the next step of this method.

[0062] If the wind turbine is in a shutdown state, the UAV inspection equipment will acquire the coordinate information and image information of the wind turbine blades. The coordinate information is acquired by the lidar on the UAV inspection equipment, and the image information is acquired by the multispectral camera on the UAV inspection equipment. In addition, when the UAV inspection equipment takes off and is a certain distance away from the wind turbine blades, it can activate the lidar and multispectral camera to complete the acquisition of coordinate information and image information, without the UAV equipment needing to do any extra flight.

[0063] The detection path of the UAV detection equipment is planned based on coordinate information, and the defect situation of the blade is preliminarily analyzed based on image information.

[0064] The drone inspection equipment moves along the inspection path while using ultrasonic sensors to acquire ultrasonic data;

[0065] Based on ultrasonic data and the initial defect status of the blade, the defect status of the blade is comprehensively analyzed. By combining image information and ultrasonic data to judge the defect status of the blade, the accuracy of the detection results can be improved.

[0066] In addition, after comprehensively analyzing the defects of the blades, the period until the wind turbine is restarted is the recovery phase of the inspection cycle. Once the wind turbine is restarted, a new cycle begins.

[0067] It is important to note that regardless of whether the blades are defective, the wind turbine will inevitably be shut down for a short period of time. Once the wind turbine is restarted, it indicates that there are no defects or that the defects have been resolved. Therefore, wait another two months before starting the next inspection cycle.

[0068] Example 2

[0069] As one embodiment, the specific steps for the UAV detection equipment to acquire image information of wind turbine blades are as follows:

[0070] LiDAR point cloud data acquired by the lidar of the drone inspection equipment

[0071] Based on LiDAR point cloud data, the iterative nearest point algorithm is used in conjunction with the CAD model of the blade to calculate the actual pose of the blade.

[0072] The Iterative Nearest Point Algorithm is used to align two point cloud datasets, typically a source point cloud (LiDAR point cloud data) and a target point cloud (CAD model of the blade). The goal is to align the source and target point clouds by rotation (represented by the rotation matrix R) and translation (represented by the translation vector t), minimizing the distance between corresponding points in the two point clouds. Therefore, the objective function for calculating the actual pose of the blade is:

[0073]

[0074] Where N is the total number of matching point pairs in the point cloud, i is the index of the current point pair, R is the rotation matrix, t is the translation vector, and p i For point cloud data, q i The corresponding point in the model; and the objective function terminates after reaching the maximum number of iterations;

[0075] Multispectral images of the leaves were acquired using a multispectral camera on a drone inspection device.

[0076] Furthermore, the planning of the detection path for the UAV detection equipment based on coordinate information and the preliminary analysis of blade defects based on image information specifically include:

[0077] The U-Net deep learning model is used to segment the blade region. Based on multispectral images, the output blade mask M(x,y)∈{0,1} is generated, where x and y are the horizontal and vertical coordinates of the image, respectively. The blade mask is a binary image segmented by the deep learning model. Its core function is to accurately identify the blade region. M(x,y)=1 represents the blade region, and M(x,y=0) represents the background region. This can avoid the interference of background noise (such as clouds and shadows) on defect detection and reduce the false detection rate.

[0078] Extracting multispectral features f from pixels within the leaf mask i (x,y), where the multispectral features include the pixel values ​​of each band within the leaf mask, namely visible light, near-infrared, and short-wave infrared, as well as the ratio of near-infrared to red light and the contrast ratio:

[0079] Based on historical defect data and multispectral features, a defect probability heatmap P(x,y) is generated.

[0080] Furthermore, the formula for calculating the defect probability heatmap P(x,y) is as follows:

[0081]

[0082] Where N is the number of multispectral features, i is the index of the current feature, and ω i These are the weighting coefficients.

[0083] Based on the defect probability heatmap P(x,y), the ant colony algorithm is used to plan the detection path.

[0084] As needed, the specific steps for planning the detection path using the ant colony algorithm are as follows:

[0085] Define the objective function as:

[0086]

[0087] Where N is the total number of path segments that the UAV needs to detect, and i is the sequence number of the current path segment. If the blade is divided into 10 segments, then N = 10, i = 1, 2, ..., 10, d i Let λ1 and λ2 be the flight distance of path segment i, and P(x) be the weighting coefficients. i ,y i Let be the defect probability value of the location covered by the i-th path segment; the optimal detection path is generated by balancing the flight distance and the risk coverage.

[0088] In this method, an ant may represent a candidate path for the drone from the root of the leaf to the tip of the leaf.

[0089] The pheromone update rule is determined as follows:

[0090]

[0091] Where, τ ij (t) represents the pheromone concentration of the path segment i→j at time t, ρ represents the pheromone volatility coefficient, m represents the total number of ants, and k represents the current ant's sequence number. Let $\frac{k}{j}$ be the amount of pheromone released by the $k$-th ant on the path segment $i→j$, and its calculation formula is as follows:

[0092]

[0093] Q is a constant used to control the baseline amount of pheromone released by ants along their path, and L... k Let be the total path length of the k-th ant;

[0094] Ants release pheromones along their paths. The higher the quality of the path (e.g., shorter distance, more coverage of high-risk areas), the higher the pheromone concentration. Ants then select paths based on the pheromone concentration, forming a positive feedback mechanism that gradually converges to the globally optimal path. Finally, ants iterate and update their path selection strategy multiple times, adaptively adjusting their flight routes to balance path length and detection coverage.

[0095] In this detection path planning method, "ant" is the core abstract entity of the ant colony optimization algorithm. By simulating the foraging behavior of ant colonies, it achieves efficient path exploration and optimization in complex environments. Its essence is an optimization strategy based on swarm intelligence. It is decoupled from the UAV detection equipment, but the final generated path can guide actual detection operations.

[0096] Example 3

[0097] As one embodiment, the specific steps for acquiring ultrasonic data using an ultrasonic sensor are as follows:

[0098] Obtain the echo data s(t);

[0099] Preprocess the echo data and calculate the defect depth d based on the echo data;

[0100]

[0101] Where v is the propagation speed of ultrasound in the blade material, and Δt is the echo time difference;

[0102] In particular, wavelet transform is used to filter out environmental noise and complete the preprocessing of the echo data;

[0103] Based on the preprocessed echo data, echo energy characteristics are extracted, and correlation coefficients are calculated.

[0104] The echo energy characteristic is used to quantify signal strength to assess defect size, and the specific calculation formula is as follows:

[0105]

[0106] Where t1 and t2 are echo windows;

[0107] The correlation coefficient is used to identify anomalous reflections, and the specific calculation formula is as follows:

[0108]

[0109] Among them, s ref (t) represents the defect-free sample signal.

[0110] Example 4

[0111] like Figure 2In one embodiment shown, the step of comprehensively analyzing the blade's defect status based on ultrasonic data and the initial defect status of the blade includes the following steps:

[0112] Update the defect thermal map based on echo energy characteristics;

[0113] The specific calculation formula is as follows:

[0114]

[0115] Among them, P updated (x,y) represents the updated defect heatmap, β is the fusion coefficient, and E is the echo energy characteristic.

[0116] The input feature vector x = [E, d, C] is constructed based on the echo energy characteristics, defect depth, and correlation coefficient.

[0117] Construct a decision function and calculate the defect category based on the input feature vector;

[0118] The decision function calculates a score or probability based on the feature vector x, which is used to determine the category of a sample. The specific calculation formula is as follows:

[0119]

[0120] Where N is the total number of input feature vectors, i is the i-th input feature vector, and α i For the weights of the support vectors, y i Let b be the label of the training sample, b be the bias term, and K(x) be the bias term. i ,x) is a linear kernel function, specifically using a Gaussian kernel (RBF) to capture complex spectral reflectance modes. The sensitivity and generalization ability of the model are balanced by adjusting the value of γ in the Gaussian kernel.

[0121] Where f(x) = 1 indicates the presence of cracks, f(x) = -1 indicates the presence of delamination, and f(x) = 0 indicates the absence of defects, the defect status of the blade is obtained by comprehensively analyzing the information. When cracks or delamination defects occur, the UAV transmits the information to the ground base station via the communication antenna. Once the staff is aware of this, they will organize the repair of the blade.

[0122] Example 5

[0123] A wind turbine blade defect detection system includes a drone detection device, a detection path planning unit, a data acquisition unit, and a defect analysis unit;

[0124] The drone detection equipment includes the drone itself, as well as a lidar, a multispectral camera, and an ultrasonic sensor installed on the drone itself.

[0125] The detection path planning and analysis unit is used to plan the detection path of the UAV detection equipment based on coordinate information and to perform preliminary analysis of the blade defects based on image information.

[0126] The data acquisition unit is used to obtain the initial defect status of the blade and ultrasonic data;

[0127] The defect analysis unit is used to comprehensively analyze the defect status of the blade based on ultrasonic data and the defect status of the blade.

[0128] Example 6

[0129] As attached Figure 3 An electronic device is shown, characterized in that it comprises:

[0130] Processor, memory, communication interface;

[0131] The memory is used to store the executable instructions of the processor;

[0132] The processor is configured to execute the aforementioned wind turbine blade defect detection method by executing the executable instructions.

[0133] A readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the above-described method for detecting defects in wind turbine blades.

[0134] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting defects in wind turbine blades, characterized in that, Specifically, it includes: Periodically acquire the operating status of wind turbine generators; If the wind turbine is in a shutdown state, the coordinate information and image information of the wind turbine blades will be obtained by the drone detection equipment; The detection path of the UAV detection equipment is planned based on coordinate information, and the defect situation of the blade is preliminarily analyzed based on image information. The drone inspection equipment moves along the inspection path while using ultrasonic sensors to acquire ultrasonic data; Based on ultrasonic data and the initial defects of the blade, the defect situation of the blade is comprehensively analyzed.

2. The method for detecting defects in wind turbine blades according to claim 1, characterized in that: The specific steps for the UAV detection equipment to acquire image information of wind turbine blades are as follows: LiDAR point cloud data acquired by the lidar of the drone inspection equipment Based on LiDAR point cloud data, the iterative nearest point algorithm is used in conjunction with the CAD model of the blade to calculate the actual pose of the blade. Multispectral images of the leaves were acquired using a multispectral camera on a drone inspection device.

3. The method for detecting defects in wind turbine blades according to claim 2, characterized in that: The iterative nearest-point algorithm, combined with the CAD model of the blade, is used to calculate the actual pose of the blade. The objective function is: Where N is the total number of matching point pairs in the point cloud, i is the index of the current point pair, R is the rotation matrix, t is the translation vector, and p i For point cloud data, q i These are the corresponding points in the model.

4. The method for detecting defects in wind turbine blades according to claim 3, characterized in that: The method of planning the detection path of the UAV detection equipment based on coordinate information and conducting preliminary analysis of the blade defects based on image information specifically includes: The U-Net deep learning model is used to segment the leaf region. Based on multispectral images, the output leaf mask M(x,y)∈{0,1} is generated, where x and y are the horizontal and vertical coordinates of the image, respectively. Extracting multispectral features f from pixels within the leaf mask i (x,y); Based on historical defect data and multispectral features, a defect probability heatmap P(x,y) is generated. Based on the defect probability heatmap P(x,y), the ant colony algorithm is used to plan the detection path.

5. The method for detecting defects in wind turbine blades according to claim 4, characterized in that: The formula for calculating the defect probability heatmap P(x,y) is as follows: Where N is the number of multispectral features, i is the index of the current feature, and ω i These are the weighting coefficients.

6. The method for detecting defects in wind turbine blades according to claim 5, characterized in that: The specific steps for planning the detection path using the ant colony algorithm are as follows: Define the objective function as: Where N is the total number of path segments that the UAV needs to detect, i is the sequence number of the current path segment, and d i Let λ1 and λ2 be the flight distance of path segment i, and P(x) be the weighting coefficients. i ,y i ) represents the defect probability value at the location covered by the i-th path segment; The pheromone update rule is determined as follows: Where, τ ij (t) represents the pheromone concentration of the path segment i→j at time t, ρ represents the pheromone volatility coefficient, m represents the total number of ants, and j represents the current ant's sequence number. Let $\frac{k}{j}$ be the amount of pheromone released by the $k$-th ant on the path segment $i→j$, and its calculation formula is as follows: Q is a constant, L k Let be the total path length of the k-th ant.

7. The method for detecting defects in wind turbine blades according to claim 6, characterized in that: The specific steps for acquiring ultrasonic data using an ultrasonic sensor are as follows: Acquire echo data; Preprocess the echo data and calculate the defect depth based on the echo data; Based on the preprocessed echo data, echo energy characteristics are extracted, and correlation coefficients are calculated.

8. The method for detecting defects in wind turbine blades according to claim 7, characterized in that: Wavelet transform is used to filter out environmental noise and complete the preprocessing of echo data.

9. The method for detecting defects in wind turbine blades according to claim 7, characterized in that: The specific steps for comprehensively analyzing the defect situation of the blade based on ultrasonic data and the initial defect situation of the blade are as follows: Update the defect thermal map based on echo energy characteristics; The input feature vector is constructed based on echo energy characteristics, defect depth, and correlation coefficient. Construct a decision function and calculate the defect category based on the input feature vector.

10. A wind turbine blade defect detection system for implementing the wind turbine blade defect detection method as described in any one of claims 1-9, comprising a UAV detection device, a detection path planning unit, a data acquisition unit, and a defect analysis unit; The drone detection equipment includes the drone itself, as well as a lidar, a multispectral camera, and an ultrasonic sensor installed on the drone itself. The detection path planning and analysis unit is used to plan the detection path of the UAV detection equipment based on coordinate information and to perform preliminary analysis of the blade defects based on image information. The data acquisition unit is used to obtain the initial defect status of the blade and ultrasonic data; The defect analysis unit is used to comprehensively analyze the defect status of the blade based on ultrasonic data and the defect status of the blade.

Images