A fan non-stop intelligent inspection method, system, device and medium

Abstract

The application relates to a fan non-stop intelligent inspection method, system, device and medium, and belongs to the technical field of fan inspection. The intelligent inspection method comprises the following steps: acquiring an environmental wind speed, a blade stiffness coefficient, a blade length and a fan center point coordinate; calculating a blade bending offset according to the environmental wind speed and the blade stiffness coefficient; determining a safe hovering distance of a UAV according to the blade length, and controlling the UAV to fly directly above the fan center; acquiring a yaw angle of the UAV in a vertical direction and a tower drum, and predicting yaw changes according to the yaw angle and a rate change of the environmental wind speed; calculating a photographing point coordinate according to the blade bending offset, the yaw angle, the safe hovering distance and the predicted yaw changes; controlling the UAV to reach the photographing point coordinate to collect a blade surface image, so as to identify a blade defect, and generating a defect report after the defect is determined. The application has the beneficial effects of realizing the inspection of the blade when the fan is not stopped and improving the inspection efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fan inspection, in particular to a fan non-stop intelligent inspection method, system, device and medium. BACKGROUND

[0002] At present, the new energy fan blade is subjected to strong wind, salt spray and other erosions for a long time, and is prone to cracks, corrosion and other defects. Regular inspection can timely find hidden troubles, avoid blade rupture, shedding and other serious accidents, protect equipment and personnel safety, improve power generation efficiency and economic benefits, and reduce downtime loss.

[0003] The traditional fan inspection method mainly includes manual climbing inspection and ordinary unmanned aerial vehicle inspection. When manually climbing inspection, the wind turbine generator can reach hundreds of meters high, and the operation and maintenance personnel need to climb to a high place for detection. High-intensity inspection is easy to cause visual fatigue, affecting the detection result and efficiency, and the safety hidden danger cannot be ignored. When ordinary unmanned aerial vehicle inspection, the fan usually needs to be stopped to ensure the comprehensiveness and clarity of the shooting, but this will cause large loss of power generation, and the single inspection time is too long. In addition, part of the unmanned aerial vehicle inspection depends on the technical level and judgment ability of the operation hand, the actual participation scene is less, and there are problems such as short battery endurance, high mechanical failure rate and low efficiency. SUMMARY

[0004] In order to realize the inspection of the blade when the fan is not stopped and improve the inspection efficiency, the present application provides a fan non-stop intelligent inspection method, system, device and medium.

[0005] In the first aspect, the present application provides a fan non-stop intelligent inspection method, which adopts the following technical scheme:

[0006] A fan non-stop intelligent inspection method, comprising:

[0007] Obtaining an environmental wind speed, a blade stiffness coefficient, a blade length and a fan center point coordinate;

[0008] According to the environmental wind speed and the blade stiffness coefficient, the blade bending offset is calculated;

[0009] According to the blade length, the safe hovering distance of the unmanned aerial vehicle is determined, and the unmanned aerial vehicle is controlled to fly to the center of the fan;

[0010] Obtaining the yaw angle of the vertical direction of the unmanned aerial vehicle hovering and the tower drum, and predicting the yaw change according to the current yaw angle and the rate change of the environmental wind speed;

[0011] According to the blade bending offset, the yaw angle, the safe hovering distance and the predicted yaw change, the space coordinates of the shooting point are calculated;

[0012] The unmanned aerial vehicle reaches the photographing point space coordinate to collect the blade surface image;

[0013] According to the blade surface image, a blade defect is identified, and a defect report is generated after the defect is determined.

[0014] By using the above technical solution, by acquiring environmental wind speed, blade stiffness coefficient and other parameters, the blade bending offset, safe hovering distance, predicted yaw change and the like are calculated, the photographing point space coordinate can be quickly and accurately calculated, the unmanned aerial vehicle can quickly reach the specified position to collect images, the time waste in the inspection process is reduced, and the fan does not need to be stopped. Without manual close-range operation, the risk of high-altitude operation brought by the traditional inspection method is avoided, and the working strength and operation risk of the operation and maintenance personnel are significantly reduced. According to the collected blade surface image, the blade defect is automatically identified, and a defect report is generated, which provides detailed defect information for the operation and maintenance personnel, and facilitates the development of maintenance plans and decisions.

[0015] Optionally, the step of acquiring the yaw angle of the unmanned aerial vehicle in the vertical direction and the tower tube includes:

[0016] When the unmanned aerial vehicle hovers, a tower tube image and original point cloud data of the tower tube are collected;

[0017] According to the tower tube image, a visual yaw angle is obtained, the visual yaw angle being an included angle between a tower tube edge and a vertical direction of the unmanned aerial vehicle;

[0018] According to the original point cloud data, a tower tube center is fitted, and a radar yaw angle is calculated in combination with current attitude data of the unmanned aerial vehicle;

[0019] The radar yaw angle and the visual yaw angle are fused to generate a final yaw angle.

[0020] By using the above technical solution, the tower tube image can directly reflect the included angle between the tower tube edge and the vertical direction of the unmanned aerial vehicle, and provide visual yaw information; the original point cloud data can be used to fit the tower tube center, and the radar yaw angle is calculated in combination with the attitude data of the unmanned aerial vehicle. The fusion of the two kinds of data makes the obtained yaw angle information more comprehensive and accurate, and makes up for the limitations that may exist in a single data source, thereby improving the accuracy of the final yaw angle.

[0021] Optionally, the step of calculating the photographing point space coordinate according to the blade bending offset, the yaw angle, the safe hovering distance and the predicted yaw change includes:

[0022] A three-dimensional space coordinate system with a fan center point as an origin is established;

[0023] Based on the safe hovering distance and the yaw angle, initial position coordinates of the unmanned aerial vehicle in a horizontal plane are calculated;

[0024] correct the initial position coordinates according to the blade bending offset;

[0025] adjust a horizontal position of the initial position coordinates based on the predicted yaw change to generate original photographing point spatial coordinates.

[0026] Optionally, the step after calculating the photographing point spatial coordinates comprises:

[0027] retrieve camera parameters and calculate a photographing distance of the unmanned aerial vehicle from the blade in combination with a target ground sampling distance;

[0028] obtain a distance difference between the photographing distance and the safe hovering distance;

[0029] when the distance difference is less than a difference threshold, correct the photographing point spatial coordinates according to the photographing distance.

[0030] By using the above technical solution, the distance most suitable for obtaining a clear and accurate image can be determined according to the camera performance and actual photographing requirements. When the difference between the photographing distance and the safe hovering distance is less than the difference threshold, the photographing point spatial coordinates are corrected according to the photographing distance. This enables the photographing position of the unmanned aerial vehicle to be flexibly adjusted according to the actual photographing distance requirements, avoids the problems of blurred and incomplete blade images caused by an unsuitable distance, and further guarantees the photographing quality. In addition, it can also monitor whether the distance between the unmanned aerial vehicle and the blade is within the safe range at all times. Even when the photographing position is adjusted according to the photographing requirements, the unmanned aerial vehicle can be ensured not to be too close to the blade, thereby avoiding collision accidents and guaranteeing the safety of the unmanned aerial vehicle equipment, and also avoiding damage to the fan blade.

[0031] Optionally, the step after correcting the photographing point spatial coordinates comprises:

[0032] obtain a fan rotating speed and calculate a linear speed of a blade tip according to the fan rotating speed;

[0033] calculate a distance advance amount for eliminating motion blur according to the linear speed and a camera exposure time in the camera parameters;

[0034] calculate actual photographing point spatial coordinates of the unmanned aerial vehicle for photographing and hovering according to the distance advance amount and the corrected photographing point spatial coordinates.

[0035] By adopting the technical scheme, the fan is in a high-speed rotating state during operation. The fan rotating speed is acquired and the linear speed of the blade tip is calculated, and the motion characteristics of the blade are fully considered. In combination with the camera exposure time in the camera parameters, the distance advance for eliminating motion blur is calculated, and the motion of the blade can be compensated when the unmanned aerial vehicle takes a picture. When the unmanned aerial vehicle hovers and takes a picture according to the actual photographing point space coordinates considering the distance advance, the image blur problem caused by the motion of the blade can be effectively avoided, and the collected blade surface image is more clear and accurate.

[0036] Optionally, according to the blade surface image, the step of identifying the blade defect comprises:

[0037] The blade surface image is processed in zones, the morphological features, texture features, polarization features and three-dimensional features of each zone are extracted and analyzed;

[0038] According to the analysis result, the abnormal area is located;

[0039] The historical feature data matched with the analysis result is searched from the historical defect database of the same type of fan, and the defect type associated with the searched historical feature data is taken as the current defect of the blade;

[0040] According to the abnormal area and the current defect, the defect is labeled at the same position of the blade of the pre-constructed three-dimensional model of the fan.

[0041] By adopting the technical scheme, compared with single feature analysis, the multi-feature combination greatly improves the accuracy of abnormal area positioning. Even if some features are disturbed by environmental factors, other features can still provide effective information, reducing the probability of misjudgment and omission. The historical data contains a large number of actual cases, and by comparing the features of the current blade with the historical feature data, the defect type can be more accurately judged. The defect labeling on the three-dimensional model provides clear guidance for maintenance personnel, so that the operation and maintenance personnel can intuitively see the position and type of the defect on the three-dimensional model, and quickly understand the overall condition of the blade.

[0042] Optionally, the intelligent inspection method further comprises:

[0043] Acquiring real-time operation data and historical maintenance logs of the fan;

[0044] Inputting the real-time operation data, historical maintenance logs, current blade defect identification result and environmental parameters of the fan into a pre-trained degradation prediction model to obtain blade potential degradation area coordinates and residual life probability distribution;

[0045] Based on the blade potential degradation area coordinates and residual life probability distribution, a dynamic inspection priority atlas is generated;

[0046] According to the dynamic inspection priority map, the subsequent inspection path and the photographing point sequence of the unmanned aerial vehicle are dynamically optimized.

[0047] By adopting the technical scheme, after the dynamic inspection priority map is generated, the unmanned aerial vehicle can preferentially cover the high-risk area, avoid redundant inspection of healthy leaves, and further improve the inspection efficiency. Meanwhile, the adaptive path optimization can dynamically adjust the inspection period according to the aging trend of the fan, realize preventive maintenance, avoid the risk of failure downtime in advance, and prolong the overall service life of the fan.

[0048] In a second aspect, the application provides a fan non-stop intelligent inspection system, which adopts the following technical scheme:

[0049] A fan non-stop intelligent inspection system comprises:

[0050] A parameter acquisition module is configured to acquire an environmental wind speed, a blade stiffness coefficient, a blade length, and a fan center point coordinate.

[0051] A data processing module is configured to calculate a blade bending offset based on the environmental wind speed and the blade stiffness coefficient, determine a safe hovering distance of an unmanned aerial vehicle based on the blade length, acquire a yaw angle of the unmanned aerial vehicle in a vertical direction with respect to a tower drum based on the blade bending offset, and predict a yaw change based on a current yaw angle and a rate of change of the environmental wind speed.

[0052] A control module is configured to control flight of the unmanned aerial vehicle.

[0053] An image acquisition module is configured to acquire a blade surface image when the unmanned aerial vehicle reaches the photographing point space coordinate.

[0054] A defect identification module is configured to identify a blade defect based on the blade surface image, and generate a defect report after the defect is determined.

[0055] In a third aspect, the application provides a computer device, which adopts the following technical scheme:

[0056] A computer device comprises a memory, a processor, and a computer program stored in the memory, and the processor executes the computer program to implement the fan non-stop intelligent inspection method of the first aspect.

[0057] In a fourth aspect, the application provides a computer readable storage medium, which adopts the following technical scheme:

[0058] A computer readable storage medium stores a computer program capable of being loaded and executed by a processor to implement the fan non-stop intelligent inspection method of the first aspect.

[0059] In summary, this application includes at least one of the following beneficial technical effects:

[0060] By acquiring parameters such as ambient wind speed and blade stiffness coefficient, the system calculates blade bending offset, safe hovering distance, and predicts yaw changes, enabling rapid and accurate calculation of the spatial coordinates of the image capture point. This allows the drone to quickly reach the designated location to collect images, reducing time wasted during inspections and eliminating the need to shut down the wind turbine. The elimination of close-range manual operation avoids the high-altitude work risks associated with traditional inspection methods, significantly reducing the workload and operational risks for maintenance personnel. Based on the collected blade surface images, the system automatically identifies blade defects and generates defect reports, providing maintenance personnel with detailed defect information to facilitate maintenance planning and decision-making. Attached Figure Description

[0061] Figure 1 This is a first flowchart of an embodiment of the method of this application;

[0062] Figure 2 This is a second flowchart of an embodiment of the method of this application;

[0063] Figure 3 This is a third flowchart of an embodiment of the method of this application;

[0064] Figure 4 This is the fourth flowchart of an embodiment of the method of this application;

[0065] Figure 5 This is the fifth flowchart of an embodiment of the method of this application;

[0066] Figure 6 This is the sixth flowchart of an embodiment of the method of this application. Detailed Implementation

[0067] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figures 1-6 The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.

[0068] The first embodiment of this application discloses a method for intelligent inspection of a wind turbine without shutting it down. (Refer to...) Figure 1 The intelligent inspection method includes S110-S170:

[0069] S110, obtain ambient wind speed, blade stiffness coefficient, blade length and wind turbine center point coordinates;

[0070] S120, calculate the blade bending offset based on the ambient wind speed and blade stiffness coefficient;

[0071] S130 determines the safe hovering distance of the drone based on the blade length and controls the drone to fly directly above the center of the wind turbine.

[0072] S140: Obtain the yaw angle of the UAV in the vertical direction relative to the tower when hovering, and predict the yaw change based on the current yaw angle and the rate of change of the ambient wind speed;

[0073] S150, calculates the spatial coordinates of the photo-taking point based on the blade bending offset, yaw angle, safe hovering distance, and predicted yaw change;

[0074] S160 controls the drone to reach the spatial coordinates of the shooting point to collect images of the blade surface;

[0075] S170 identifies blade defects based on blade surface images and generates a defect report after the defects are identified.

[0076] Specifically, in step S110, the ambient wind speed, blade stiffness coefficient, blade length, and wind turbine center point coordinates are obtained. This is achieved by: monitoring the ambient wind speed in real time using an ultrasonic anemometer installed on top of the wind turbine tower and transmitting the data to the onboard computer; retrieving the blade stiffness coefficient from the wind turbine specification database pre-stored in the data center (which contains parameters such as the elastic modulus of the blade material); obtaining the blade length by querying the wind turbine configuration file for the same wind farm; and using a high-precision RTK-GPS module to locate and obtain the wind turbine center point coordinates before the UAV takes off, with the coordinate data synchronously uploaded to the system platform.

[0077] In step S120, the blade bending offset is calculated based on the ambient wind speed and blade stiffness coefficient. Specifically, the offset model is derived using fluid dynamics formulas: Blade bending offset = (drag coefficient * air density * ambient wind speed squared * blade length cubed) / (m * blade stiffness coefficient * blade section moment of inertia), where the blade section moment of inertia is preloaded from the database, and m is a set constant term. The calculation is performed by an embedded algorithm on the airborne computer, and the result is used to compensate for the impact of dynamic blade deformation on UAV positioning, ensuring safe inspection.

[0078] The specific implementation of step S130 to determine the safe hovering distance is as follows: Based on the blade length L and the safety factor k (usually taken as 1.5), the safe hovering distance is determined as follows: Safe hovering distance = k × L + redundancy value. After the safe hovering distance is calculated, it is input into the UAV autopilot. The autopilot, combined with the coordinates (longitude, latitude, and elevation) of the wind turbine center point located by RTK-GPS, generates a straight flight path. The GPS / IMU sensors on the UAV correct the position in real time, and combined with the SLAM obstacle avoidance module, it flies to hover directly above the wind turbine center.

[0079] Reference Figure 2, in S140, the step of obtaining the yaw angle of the vertical direction of the unmanned aerial vehicle hovering and the tower drum includes S210-S240:

[0080] S210, when the unmanned aerial vehicle hovers, collecting tower drum image and original point cloud data of the tower drum;

[0081] S220, according to the tower drum image, the visual yaw angle is obtained, the visual yaw angle is the included angle between the edge of the tower drum and the vertical direction of the unmanned aerial vehicle;

[0082] S230, according to the original point cloud data, the center of the tower drum is fitted, and the radar yaw angle is calculated combined with the current attitude data of the unmanned aerial vehicle;

[0083] S240, fusing the radar yaw angle and the visual yaw angle to generate the final yaw angle.

[0084] Specifically, in the hovering state, the 2000 million pixel wide-angle camera carried by the unmanned aerial vehicle shoots the tower drum panoramic image, and then uses OpenCV library for edge detection (such as Canny operator) to extract the tower drum contour line; the included angle between the contour and the vertical axis of the unmanned aerial vehicle (provided by the IMU sensor) is analyzed by Hough transform, and the visual yaw angle is calculated, which represents the deviation of the tower drum relative to the attitude of the unmanned aerial vehicle, and is used for preliminary estimation of the direction of the wind turbine.

[0085] Synchronously activate the on-board solid-state laser radar (such as Livox Mid-40) to scan the tower drum to obtain the 360° point cloud of the tower drum, fit the cylindrical model by RANSAC algorithm to obtain the center coordinates of the tower drum, project the direction vector of the tower drum axis to the horizontal plane, solve the azimuth angle of the tower drum axis in the radar horizontal plane, and then fuse the IMU attitude data (pitch, roll) of the unmanned aerial vehicle to convert the azimuth angle to the global coordinate system of the unmanned aerial vehicle to obtain the radar yaw angle, which represents the yaw deviation based on spatial geometry, and the result is stored in the business layer database.

[0086] The radar yaw angle and the visual yaw angle are fused by using Kalman filter: the fusion weight is dynamically set by the weather sensor (such as on-board hygrometer) (for example, the visual weight is 0.7 and the radar weight is 0.3 in sunny day; the opposite in rainy and foggy day); the final yaw angle = visual weight * visual yaw angle + radar weight * radar yaw angle, visual weight + radar weight = 1, to ensure robustness and reduce environmental interference.

[0087] Time series model (such as ARIMA) is used for yaw change prediction, specifically: the wind speed rate change is obtained by differentiating the historical data of the wind speed sensor, the time series model inputs the final yaw angle and the wind speed derivative, and outputs the yaw angle change in the future n seconds (for example, 5 seconds), which is used to adjust the position of the unmanned aerial vehicle in advance. Model parameters are obtained by training historical running data and deployed on the on-board computer.

[0088] ReferenceFigure 3 In S150, the step of calculating the photographing point spatial coordinates according to the blade bending offset, the yaw angle, the safe hover distance and the predicted yaw change includes S310-S340:

[0089] S310, a three-dimensional spatial coordinate system is established with the center point of the wind turbine as the origin;

[0090] S320, the initial position coordinates of the unmanned aerial vehicle in the horizontal plane are calculated based on the safe hover distance and the yaw angle;

[0091] S330, the initial position coordinates are corrected according to the blade bending offset;

[0092] S340, the horizontal position of the initial position coordinates is adjusted based on the predicted yaw change to generate the original photographing point spatial coordinates.

[0093] Specifically, a three-dimensional spatial coordinate system is established with the center point of the wind turbine as the origin, which is realized by a coordinate transformation algorithm: the origin is defined as the GPS coordinates of the center point of the wind turbine in the on-board computer, the Z-axis is vertically upward, the positive north direction is the positive direction of the Y-axis, and the positive east direction is the positive direction of the X-axis, and a three-dimensional rectangular coordinate system is established.

[0094] Based on the safe hover distance and the current yaw angle, the initial position coordinates (X1, Y1, Z1) of the unmanned aerial vehicle in the horizontal plane are calculated:

[0095] X1= safe hover distance*sin(current yaw angle), Y1= safe hover distance*cos(current yaw angle), Z1=H0, H0 is a preset safe hover height, higher than the highest point of the blade.

[0096] According to the blade bending offset (including direction vector), the initial position is corrected, and the offset direction is in the opposite direction of the blade bending under the wind; the corrected coordinates (X2, Y2, Z2) are X2=X1+ blade bending offset component along the x-axis, Y2=Y1+ blade bending offset component along the y-axis, and Z2=Z1+ blade bending offset component along the z-axis.

[0097] Based on the yaw angle change predicted by the environmental wind speed change rate, the horizontal position of the corrected coordinates is adjusted, and the predicted yaw angle=current yaw angle+yaw angle change. The final photographing point coordinates (Xp, Yp, Zp) are:

[0098] Xp=X2*cos(yaw angle change)-Y2*sin(yaw angle change), Yp=X2*sin(yaw angle change)+Y2*cos(yaw angle change), Zp=Z2.

[0099] Referring to Figure 4 The steps after generating the original photographing point spatial coordinates include S410-S460:

[0100] S410, camera parameters are called and combined with the target ground sampling distance to calculate the shooting distance of the UAV from the blade;

[0101] S420, the distance difference between the shooting distance and the safe hovering distance is obtained;

[0102] S430, when the distance difference is less than the difference threshold, the shooting point spatial coordinates are corrected according to the shooting distance;

[0103] S440, the fan speed is obtained, and the linear speed of the blade tip is calculated according to the fan speed;

[0104] S450, the distance advance for eliminating motion blur is calculated according to the linear speed and the camera exposure time in the camera parameters;

[0105] S460, the actual shooting point spatial coordinates of the UAV shooting hovering are calculated according to the distance advance and the corrected shooting point spatial coordinates.

[0106] Specifically, the camera parameters are called and combined with the target ground sampling distance to calculate the shooting distance of the UAV from the blade, which is specifically implemented as follows: the camera parameters (such as focal length f, pixel size p) are loaded from the database, and the target ground sampling distance GSD (such as 5mm / pixel) is preset; the shooting distance = (GSD * f) / p, which is calculated by the on-board algorithm to ensure the image resolution.

[0107] The absolute value of the distance difference between the shooting distance and the safe hovering distance is obtained for safety verification. When the absolute value is less than the difference threshold (such as 0.5 meters), the shooting point spatial coordinates are corrected according to the shooting distance, which is specifically implemented as follows: the x and y coordinate values in the corrected shooting point spatial coordinates remain unchanged, and the z coordinate value is the shooting distance. When the distance difference is greater than or equal to the difference threshold, the z coordinate value of the shooting point spatial coordinates is the safe hovering distance.

[0108] The fan speed is provided by the fan SCADA system in real time, and the tip linear speed = 2π*L*ω, where L is the blade length and ω is the fan speed (rad / s). The distance advance = tip linear speed * (camera exposure time / 2) is used to offset the blade motion blur. The distance difference between the corrected shooting point spatial coordinates and the actual shooting point position of the UAV is the distance advance, i.e. assuming that the corrected shooting point position is p1 and the distance advance is 0.0147m, the UAV should hover 0.0147m in front of the p1 point and wait for the blade to enter the field of view.

[0109] After obtaining the actual photographing point spatial coordinates, a coordinate sequence containing 12 photographing points is generated (4 photographing angles for each blade: blade root, middle, tip, leading edge) in combination with the unmanned aerial vehicle inspection path (e.g., preset to fly around the wind turbine in a clockwise direction, with an interval of 60° for photographing one blade). Through the trajectory tracking module (PID control, position loop bandwidth 5 Hz) of the unmanned aerial vehicle flight control system, the unmanned aerial vehicle is controlled to reach each photographing point in turn at a speed of 3 m / s and hover. After the unmanned aerial vehicle hovers, the laser ranging sensor carried on the unmanned aerial vehicle measures the actual distance from the unmanned aerial vehicle to the blade in real time. When the error between the actual distance and the photographing distance is ≤0.1 m, the camera is triggered to collect the blade surface image. A single image is stored in TIFF format, and the file size is about 20 MB.

[0110] Referring to Figure 5 In S170, the step of identifying blade defects according to the blade surface image includes S510-S540:

[0111] S510, the blade surface image is processed by partitioning, the morphological features, texture features, polarization features and three-dimensional features of each region are extracted and analyzed;

[0112] S520, according to the analysis result, the abnormal area is located;

[0113] S530, the historical feature data matching the analysis result is searched from the historical defect database of the same type of wind turbine, and the defect type associated with the searched historical feature data is taken as the current defect of the blade;

[0114] S540, according to the abnormal area and the current defect, the defect is labeled at the same position of the blade of the pre-constructed wind turbine three-dimensional model.

[0115] Specifically, Mask R-CNN is used to segment the blade area, and 10×10 grid partitions are divided. Each partition extracts:

[0116] Morphological features: edge damage rate, hole area ratio (morphological operation);

[0117] Texture features: local binary pattern (LBP) entropy value, gray level co-occurrence matrix contrast;

[0118] Polarization features: multi-angle polarization camera obtains surface stress distribution;

[0119] Three-dimensional features: laser point cloud reconstructs surface curvature anomaly degree;

[0120] After analyzing the extracted features, the analysis results are combined into a feature vector, and the feature vector is input into a pre-trained ResNet-50 classification model to output abnormal region coordinates. By correlating the historical defect database (stored in the data center) of the same type of fan, the historical feature data is matched through cosine similarity to determine the defect type (such as lightning carbonization and leading edge corrosion). In the fan digital twin model built in the Unity engine, the abnormal region coordinates are located according to the blade coordinate system, and the defect range is highlighted in red. A PDF report containing positioning maps, defect quantification data, and maintenance recommendations is automatically generated.

[0121] Further, with reference to Figure 6 The intelligent inspection method can further include S610-S640:

[0122] S610, obtaining fan real-time running data and historical maintenance logs;

[0123] S620, inputting the fan real-time running data, historical maintenance logs, current blade defect identification results, and environmental parameters into a pre-trained degradation prediction model to obtain blade potential degradation region coordinates and residual life probability distribution;

[0124] S630, generating a dynamic inspection priority map based on the blade potential degradation region coordinates and the residual life probability distribution;

[0125] S640, dynamically optimizing the subsequent inspection path and photographing point sequence of the unmanned aerial vehicle according to the dynamic inspection priority map.

[0126] Specifically, access the vibration, power curve, and other real-time running data of the fan SCADA system, parse the repair parts and frequency in the historical maintenance logs, and input the current blade defect features, environmental temperature and humidity, and other parameters into a pre-trained LSTM-Transformer hybrid model (trained based on 200 fan full-life cycle data). The model output includes: ① Heat map coordinates of potential degradation regions (regions with a failure probability of >30% in the next 3 months) ② Residual life Weibull distribution curve (95% confidence interval of residual life distribution) According to the prediction results, a color-coded priority map is generated: red area (high-risk defect development area) is given 5 times sampling weight, yellow area (moderate degradation) is given 2 times weight. Combined with the endurance time of the unmanned aerial vehicle and the wind speed constraint, an improved ant colony algorithm is used to recalculate the inspection path, and 45° oblique photographing points are added in the key areas. The photographing point sequence is sorted by time optimization principle, and is updated to the flight control system in real time through the 4G module.

[0127] Based on the above method embodiments, the second embodiment of the present application discloses a fan non-stop intelligent inspection system. The fan non-stop intelligent inspection system of the embodiment of the present application can implement any one of the above fan non-stop intelligent inspection methods, and the specific working processes of each module in the fan non-stop intelligent inspection system can refer to the corresponding processes in the above method embodiments.

[0128] For ease of understanding, the following is an example: a fan non-stop intelligent inspection system, comprising:

[0129] a parameter acquisition module, configured to acquire an environmental wind speed, a blade stiffness coefficient, a blade length, and a fan center point coordinate;

[0130] a data processing module, configured to calculate a blade bending offset based on the environmental wind speed and the blade stiffness coefficient, determine a safe hovering distance of the unmanned aerial vehicle based on the blade length, acquire a yaw angle between a vertical direction of the unmanned aerial vehicle in hovering and the tower drum, and predict a yaw change based on a current yaw angle and a rate change of the environmental wind speed, and calculate a photographing point spatial coordinate based on the blade bending offset, the yaw angle, the safe hovering distance, and the predicted yaw change;

[0131] a control module, configured to control flight of the unmanned aerial vehicle;

[0132] an image acquisition module, configured to acquire a blade surface image when the unmanned aerial vehicle reaches the photographing point spatial coordinate;

[0133] a defect identification module, configured to identify a blade defect based on the blade surface image, and generate a defect report after determining the defect.

[0134] The third embodiment of the present application provides a computer device, which can include a memory, a processor, and a computer program stored in the memory, and the processor executes the computer program to implement a fan non-stop intelligent inspection method.

[0135] The memory can be in communication connection with the processor through a communication bus, and the communication bus can be an address bus, a data bus, a control bus, etc.

[0136] In addition, the memory can include a random access memory (RAM) and can also include a non-volatile memory (NVM), such as at least one disk memory.

[0137] The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc., and can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc.

[0138] The fourth embodiment of the present application provides a computer readable storage medium, which stores a computer program capable of being loaded and executed by a processor to implement the fan non-stop intelligent inspection method.

[0139] The computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus or device; the program code contained in the computer readable medium can be transmitted by any appropriate medium, including but not limited to wireless, wire, optical cable, RF, etc., or any appropriate combination of the above.

[0140] It should be noted that the computer device and the storage medium of the embodiments of the present application are electronic devices and storage media that apply the fan non-stop intelligent inspection method described above, and all embodiments of the fan non-stop intelligent inspection method described above are applicable to the computer device and the storage medium, and can achieve the same or similar beneficial effects. For the computer device / storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts are referred to the part of the method embodiments.

[0141] Although the present application is described herein in conjunction with various embodiments, other variations of the disclosed embodiments can be understood and implemented by those skilled in the art with reference to the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "one" does not exclude a plurality. A single processor or other unit can implement several functions listed in the claims. Some measures described in mutually different dependent claims can be combined and produce a good result.

[0142] The above are preferred embodiments of the present application, which do not limit the protection scope of the present application, any feature disclosed in the specification (including the abstract and the drawings) can be replaced by other equivalent or similar features unless specifically described, that is, each feature is only an example of a series of equivalent or similar features.

Claims

1. A method for intelligent inspection of a fan without stopping, characterized in that, The method comprises the following steps: Obtain the environmental wind speed, blade stiffness coefficient, blade length, and wind turbine center point coordinates; Calculate the blade bending offset based on the environmental wind speed and the blade stiffness coefficient; Determine the safe hovering distance of the unmanned aerial vehicle based on the blade length, and control the unmanned aerial vehicle to fly directly above the wind turbine center; Obtain the yaw angle between the vertical direction of the unmanned aerial vehicle hovering and the tower, and predict the yaw change based on the current yaw angle and the rate of change of the environmental wind speed; Calculate the photographing point space coordinates based on the blade bending offset, the yaw angle, the safe hovering distance, and the predicted yaw change; Control the unmanned aerial vehicle to reach the photographing point space coordinates to collect blade surface images; Identify blade defects based on the blade surface images, and generate a defect report after determining the defects; The step of calculating the photographing point space coordinates based on the blade bending offset, the yaw angle, the safe hovering distance, and the predicted yaw change comprises: Establish a three-dimensional coordinate system with the wind turbine center point as the origin; Calculate the initial position coordinates of the unmanned aerial vehicle in the horizontal plane based on the safe hovering distance and the yaw angle; Correct the initial position coordinates based on the blade bending offset; Adjust the horizontal position of the initial position coordinates based on the predicted yaw change to generate the original photographing point space coordinates.

2. The method of claim 1, wherein, The step of obtaining the yaw angle between the vertical direction of the unmanned aerial vehicle hovering and the tower comprises: Collect tower images and original point cloud data of the tower while the unmanned aerial vehicle is hovering; Obtain the visual yaw angle based on the tower images, which is the angle between the tower edge and the vertical direction of the unmanned aerial vehicle; Fit the tower center based on the original point cloud data, and calculate the radar yaw angle in combination with the current attitude data of the unmanned aerial vehicle; Fuse the radar yaw angle and the visual yaw angle to generate the final yaw angle.

3. The method of claim 1, wherein, The steps after calculating the photographing point space coordinates comprise: Retrieve the camera parameters and calculate the shooting distance of the unmanned aerial vehicle from the blade in combination with the target ground sampling distance; Obtain the distance difference between the shooting distance and the safe hovering distance; When the distance difference is less than the difference threshold, correct the photographing point space coordinates based on the shooting distance.

4. The method of claim 3, wherein, The steps after correcting the photographing point space coordinates comprise: Obtain the wind turbine speed and calculate the linear speed of the blade tip based on the wind turbine speed; Calculate the distance advance for eliminating motion blur based on the linear speed and the camera exposure time in the camera parameters; Calculate the actual photographing point space coordinates for the unmanned aerial vehicle photographing hovering based on the distance advance and the corrected photographing point space coordinates.

5. The method of claim 1, wherein, The step of identifying blade defects based on the blade surface images comprises: Divide the blade surface images into regions, extract the morphological features, texture features, polarization features, and three-dimensional features of each region, and analyze them; Locate abnormal regions based on the analysis results; Find historical feature data matching the analysis results from the historical defect database of the same type of wind turbine, and associate the defect type of the found historical feature data as the current defect of the blade; According to the abnormal area and the current defect, defect labeling is performed on the same position of a pre-constructed three-dimensional model blade of the fan.

6. The method of claim 1, wherein, The intelligent inspection method further comprises: acquiring real-time operation data and historical maintenance logs of the fan; inputting the real-time operation data, the historical maintenance logs, a current blade defect identification result, and environmental parameters into a pre-trained degradation prediction model to obtain blade potential degradation area coordinates and a residual life probability distribution; generating a dynamic inspection priority atlas based on the blade potential degradation area coordinates and the residual life probability distribution; dynamically optimizing a subsequent inspection path and a photographing point sequence of the unmanned aerial vehicle according to the dynamic inspection priority atlas.

7. A fan non-stop intelligent inspection system, characterized in that, The fan non-stop intelligent inspection method according to any one of claims 1 to 6 comprises: a parameter acquisition module configured to acquire environmental wind speed, blade stiffness coefficient, blade length, and fan center point coordinates; a data processing module configured to calculate blade bending offset according to the environmental wind speed and the blade stiffness coefficient, determine a safe hovering distance of the unmanned aerial vehicle according to the blade length, acquire a yaw angle of the unmanned aerial vehicle in a vertical direction with respect to the tower, and predict yaw change according to a current yaw angle and a rate of change of the environmental wind speed, and calculate photographing point spatial coordinates according to the blade bending offset, the yaw angle, the safe hovering distance, and the predicted yaw change; the step of calculating the photographing point spatial coordinates according to the blade bending offset, the yaw angle, the safe hovering distance, and the predicted yaw change comprises: establishing a three-dimensional coordinate system with the fan center point as the origin; calculating initial position coordinates of the unmanned aerial vehicle in a horizontal plane based on the safe hovering distance and the yaw angle; correcting the initial position coordinates according to the blade bending offset; adjusting a horizontal position of the initial position coordinates based on the predicted yaw change to generate original photographing point spatial coordinates; a control module configured to control the unmanned aerial vehicle to fly; an image acquisition module configured to acquire a blade surface image when the unmanned aerial vehicle reaches the photographing point spatial coordinates; a defect identification module configured to identify a blade defect according to the blade surface image and generate a defect report after the defect is determined.

8. A computer device, comprising: A computer program product comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the fan non-stop intelligent inspection method according to any one of claims 1 to 6 when executing the program.

9. A computer-readable storage medium, characterized in that, A computer program product comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the fan non-stop intelligent inspection method according to any one of claims 1 to 6 when executing the program.

Images