Method and system for detecting defects of wind turbine blades based on drones

Abstract

This invention provides a method and system for detecting wind turbine blade defects based on unmanned aerial vehicles (UAVs), belonging to the field of wind turbine blade defect detection technology. This invention simultaneously acquires wind turbine operating parameters and historical blade defect data, constructs transient mechanical stress and fatigue accumulation factors, and uses a nonlinear coupling model to conduct risk assessment to determine the risk index of each blade. It then selects target blades and wind turbines, plans the shortest spatial node sequence with the minimum total flight range of the UAV as a spatial constraint, and generates a UAV scheduling queue. It extracts the dominant anomaly parameters of high-risk blades and maps them to local high-risk areas, extracts the surface geometric center of these areas, and offsets it along the normal direction to generate key detection anchor points. Finally, using these anchor points as the center, it adjusts the UAV flight path spacing and image acquisition frequency according to the risk index to generate an acquisition path for defect detection, thereby significantly improving the overall scheduling efficiency and endurance utilization of the inspection, and greatly increasing the effective capture rate of minute defects.

Description

Technical Field

[0001] This invention relates to the field of wind turbine blade defect detection technology, specifically to a method and system for detecting wind turbine blade defects based on unmanned aerial vehicles (UAVs). Background Technology

[0002] Wind energy, as a clean and renewable energy source, has been widely used. As the core component for capturing wind energy, the structural integrity of wind turbine blades directly determines the power generation efficiency and operational safety of the unit. Wind turbine blades are subjected to complex alternating loads and harsh natural environments over long periods, making them highly susceptible to structural defects such as cracks and spalling. Traditional blade defect detection mainly relies on personnel suspended at height or ground-based telescope inspections. This method is not only risky and time-consuming, but also easily affected by human visual fatigue, resulting in significant blind spots and efficiency bottlenecks, making it difficult to meet the large-scale, routine operation and maintenance needs of modern wind farms.

[0003] Existing UAV-based methods for detecting defects in wind turbine blades mostly employ a pre-set fixed flight path and a globally uniform acquisition frequency for full-coverage static scanning. These methods typically assume that the health status of all wind turbines and their blades within the same wind farm is relatively consistent, or treat them merely as homogeneous structural components for uniform flight path planning and image acquisition. In single-blade inspection, they fail to distinguish between the overall surface and locally vulnerable areas.

[0004] However, in actual operation, significant individual risk differences exist between different wind turbines and even between different blades of the same turbine. Furthermore, fatigue damage within a single blade is not uniformly distributed; early micro-fatigue cracks are often highly localized. Existing static full-coverage data acquisition methods not only waste significant UAV resources in healthy areas, leading to low inspection efficiency, but also often miss high-risk local areas prone to micro-defects due to insufficient acquisition accuracy or angle, making it difficult to balance overall endurance utilization with the effective capture rate of local critical defects.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for detecting defects in wind turbine blades based on unmanned aerial vehicles (UAVs), in order to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for detecting defects in wind turbine blades based on unmanned aerial vehicles (UAVs), comprising the following steps: Step 1: Within a preset time window, synchronously acquire the main shaft torque variance, yaw position absolute error, and aerodynamic performance attenuation rate of each wind turbine blade, and determine the physical length of historical defects and the service life after repair of each wind turbine blade. Step 2: Obtain the transient mechanical stress corresponding to each wind turbine blade based on the main shaft torque variance and yaw position absolute error. Construct the fatigue accumulation factor of each blade using the physical length of the historical defect and the service time after repair. Perform risk assessment on the transient mechanical stress, fatigue accumulation factor and aerodynamic performance decay rate of each wind turbine blade through a nonlinear coupling model to determine the risk index of each wind turbine blade. Step 3: Select wind turbine blades with a risk index greater than the preset safety threshold as target blades, extract the target wind turbines to which the target blades belong and remove duplicate wind turbines, obtain the spatial coordinates of the selected target wind turbines, and use minimizing the total global flight distance of the UAV as a spatial constraint to plan the shortest spatial node sequence that traverses all target wind turbines in sequence to generate the UAV scheduling queue. Step 4: According to the arrangement order of the target wind turbines in the UAV scheduling queue, and traverse the target blades contained therein, extract the dominant abnormal parameters of the target blade, map the dominant abnormal parameters to the local high-risk area of ​​the target blade based on the force mapping relationship, extract the surface geometric center of the local high-risk area and offset it along the normal by a preset safe distance to generate key detection anchor points. Step 5: Using the key detection anchor point as the spatial reference center, adjust the flight path spacing and image acquisition frequency of the UAV according to the risk index of the target blade, and generate an adaptive image acquisition path to drive the UAV to perform defect detection along the path.

[0008] Furthermore, the preset time window is a window that traces back a preset time length from the current time as the endpoint, wherein the preset time length covers at least one standard statistical period of wind turbine operation data; The logic for synchronously acquiring the main shaft torque variance, yaw position absolute error, and aerodynamic performance attenuation rate of each wind turbine blade is as follows: Obtain the instantaneous spindle torque and its arithmetic mean within a preset time window, calculate the sum of squared deviations between the two, and divide the sum of squared deviations by the total number of sampling points within the preset time window to obtain the spindle torque variance characterizing the alternating pulsating load. Obtain the actual environmental wind direction angle and the actual yaw angle of the cabin within a preset time window, calculate the mean of the absolute values ​​of the differences between the two, and obtain the absolute error of the yaw position in the quantified asymmetric shear force state. Obtain the actual average power generation of the whole machine within a preset time window and the theoretical average power under the same wind speed, calculate the difference between the theoretical average power and the actual average power generation of the whole machine, and use the ratio of the difference to the theoretical average power as the macro deviation ratio. Extract the pitch angle execution deviation, blade root load bending moment, and pitch motor operating current of each wind turbine blade based on feedback from the independent pitch sensor, and perform weighted calculation to obtain the state allocation coefficient. Multiply the macroscopic deviation ratio with the state allocation coefficient of each wind turbine blade to obtain the aerodynamic performance attenuation rate characterizing the independent state of each wind turbine blade.

[0009] Furthermore, the step of obtaining the transient mechanical stress is as follows: Obtain the preset rated reference torque, calculate the ratio of the square root of the spindle torque variance to the rated reference torque, and calibrate it as the torque ratio. The hyperbolic cosine value of the absolute error of the yaw position is calculated and multiplied with the torque ratio to obtain the transient mechanical stress, which is then used as the transient mechanical stress shared by all blades on the corresponding wind turbine. The steps for obtaining the fatigue accumulation factor are as follows: Based on the preset reference length constant, reference time constant, and empirical expansion coefficient, calculate the first ratio of the physical length of the historical defect of the wind turbine blade to the reference length constant, and the second ratio of the service time after repair to the reference time constant. The first ratio, the second ratio, and the empirical expansion coefficient are multiplied to obtain the defect expansion product. The defect expansion product is then summed with the natural constant to construct a composite fatigue proper number term. The logarithm of the composite fatigue proper number term is calculated with the natural constant as the base to obtain the fatigue accumulation factor of each wind turbine blade.

[0010] Furthermore, the steps for obtaining the risk index are as follows: Calculate the difference between the preset constant and the aerodynamic performance decay rate to obtain the effective aerodynamic retention coefficient characterizing the remaining work capacity of the blade; Calculate the ratio of the transient mechanical stress to the effective aerodynamic holding coefficient and calibrate it as the attenuated reduced stress characterizing the force amplification effect; The risk index of each wind turbine blade is calculated by multiplying the attenuated reduced stress with the fatigue accumulation factor. The specific calculation formula is as follows: In the formula, For the first Typhoon machine Risk index of each leaf For the first Variance of the main shaft torque of a typhoon generator. This is the rated reference torque for this type of fan. For the first The absolute error of the yaw position of the typhoon generator. For the first Typhoon machine aerodynamic performance degradation rate of each blade For the first Typhoon machine The physical length of the historical defects of each blade For the first Typhoon machine The service life of each repaired blade. and These are the reference length constant and the reference time constant, respectively. This is a preset empirical expansion coefficient.

[0011] Furthermore, the steps for creating the drone scheduling queue are as follows: The wind turbine blades with risk indices greater than a preset safety threshold are grouped into a target blade set. The equipment number of each blade in the target blade set is parsed, the corresponding target wind turbine number is extracted, and redundant target wind turbine numbers in the target wind turbine numbers are removed to obtain a deduplicated target wind turbine set. Obtain the preset UAV takeoff anchor point and the three-dimensional spatial coordinates of each target wind turbine in the deduplicated target wind turbine set, and define the three-dimensional spatial coordinates as independent spatial nodes. Calculate the spatial straight-line distance between each pair of spatial nodes and construct a global distance matrix. Based on the global distance matrix, taking the spatial node corresponding to the UAV takeoff anchor point as the starting node, taking the minimization of the total global flight distance of the UAV as the spatial constraint, and taking traversing all target wind turbines in the deduplicated target wind turbine set once as the solution objective, iterative optimization is performed to calculate the shortest spatial node sequence, and the shortest spatial node sequence is used as the UAV scheduling queue.

[0012] Furthermore, the judgment logic for the dominant abnormal parameter is as follows: The transient mechanical stress, fatigue accumulation factor and aerodynamic performance attenuation rate of the target blade are obtained and used as risk parameters. The absolute difference between the actual value of each risk parameter and the preset standard health benchmark value is calculated. The absolute difference is divided by the corresponding standard health benchmark value to obtain the deviation of each risk parameter. The deviation of each risk parameter is numerically compared and the parameter with the largest value is extracted and determined as the dominant abnormal parameter of the target blade.

[0013] Furthermore, the specific logic for mapping the dominant abnormal parameters to the local high-risk area of ​​the target blade based on the force mapping relationship, and offsetting it along the normal direction by a preset safety distance to generate key detection anchor points is as follows: When the dominant abnormal parameter is transient mechanical stress, the local high-risk area is mapped to the preset stress concentration area at the root of the blade in the target blade spatial coordinate system; When the dominant abnormal parameter is the aerodynamic performance decay rate, the local high-risk area is mapped to the leading edge and suction surface area of ​​the blade from the middle to the tip in the target blade spatial coordinate system. When the dominant anomaly parameter is the fatigue accumulation factor, the local high-risk area is mapped to the historical defect in-situ coordinates of the target blade and the extended area of ​​its surrounding preset radius. Obtain the unit outward normal vector perpendicular to the tangent plane where the surface geometric center is located. Multiply the unit outward normal vector with the safety distance to obtain the normal offset vector. Add the normal offset vector with the three-dimensional spatial coordinates of the surface geometric center to calculate the key detection anchor point with absolute spatial coordinates.

[0014] Furthermore, the specific logic for adjusting the flight path spacing and image acquisition frequency of the UAV based on the risk index of the target blade is as follows: Calculate the index value with the natural constant as the base and the negative of the product of the risk index of the target blade and the preset spacing attenuation adjustment coefficient as the index. Then, multiply the index value with the preset basic flight path spacing to obtain the initial spacing. Compare the initial spacing with the preset minimum limit safety spacing and extract the maximum value of the two as the flight path spacing of the UAV. The risk index is multiplied by a preset frequency gain adjustment coefficient and summed with a constant. The sum is then multiplied by a preset base image acquisition frequency to obtain an initial frequency. The initial frequency is then compared with a preset maximum continuous shooting frequency, and the minimum value between the two is extracted as the image acquisition frequency.

[0015] Furthermore, the specific logic for generating the adaptive image acquisition path is as follows: Obtain the preset topology scanning rules and the initial timestamp of the UAV arriving at the key detection anchor point; Using the key detection anchor points as the three-dimensional space reference origin, and extending outward according to the topology scanning rules and the flight path spacing of the UAV, a serialized spatial relative offset vector is constructed. The spatial relative offset vectors and the three-dimensional spatial coordinates of the key detection anchor points are respectively subjected to spatial vector addition operations to generate serialized three-dimensional spatial flight waypoints. The reciprocal of the image acquisition frequency is calculated to obtain the time interval between adjacent shutter actions. Using the initial timestamp as a reference, the time intervals are sequentially accumulated to calculate the absolute timestamps corresponding to each 3D space flight waypoint. The 3D space flight waypoints are then sequentially bound to their corresponding absolute timestamps to construct a machine control command stream for generating an adaptive image acquisition path. The specific construction formula is as follows: In the formula, For the final obtained number Typhoon machine Adaptive image acquisition path for each leaf For the first in the path sequence There are 3D space waypoints, among which... For the step index in the path sequence, and The drones reached the first The first waypoint and the first Each waypoint and the absolute timestamp that triggers the camera shutter, where, This is the initial timestamp of the drone arriving at the key detection anchor point. For the first Typhoon machine Key inspection anchor points for each blade Based on the distance between routes The first one constructed with topology scan rules Step space relative offset vector, For the first Typhoon machine The spacing between the flight paths of each blade For the first Typhoon machine Image acquisition frequency per leaf.

[0016] The present invention also provides a drone-based wind turbine blade defect detection system, which is used to implement the above-mentioned drone-based wind turbine blade defect detection method, including: The data acquisition module is used to synchronously acquire the main shaft torque variance, yaw position absolute error and aerodynamic performance attenuation rate of each wind turbine blade within a preset time window, and determine the physical length of historical defects and the service life after repair of each wind turbine blade. The risk assessment module is used to obtain the transient mechanical stress corresponding to each wind turbine blade based on the main shaft torque variance and the absolute error of the yaw position. It constructs the fatigue accumulation factor of each blade using the physical length of the historical defect and the service time after repair. It performs risk assessment on the transient mechanical stress, fatigue accumulation factor and aerodynamic performance decay rate of each wind turbine blade through a nonlinear coupling model to determine the risk index of each wind turbine blade. The scheduling and planning module is used to select wind turbine blades with a risk index greater than a preset safety threshold as target blades, extract the target wind turbines to which the target blades belong and remove duplicate wind turbines, obtain the spatial coordinates of the selected target wind turbines, and plan the shortest spatial node sequence for traversing all target wind turbines in turn, with the minimum global flight range of the UAV as the spatial constraint, to generate the UAV scheduling queue. The anchor point generation module is used to extract the dominant abnormal parameters of the target blades according to the arrangement order of the target wind turbines in the UAV scheduling queue and to traverse the target blades contained therein. Based on the force mapping relationship, the dominant abnormal parameters are mapped to the local high-risk area of ​​the target blade. The surface geometric center of the local high-risk area is extracted and offset by a preset safety distance along the normal to generate key detection anchor points. The path generation module is used to adjust the flight path spacing and image acquisition frequency of the UAV based on the risk index of the target blade, using the key detection anchor point as the spatial reference center, and to generate an adaptive image acquisition path to drive the UAV to perform defect detection along the path.

[0017] Compared with the prior art, the beneficial effects of the present invention are: This invention obtains transient mechanical stress based on the variance of the main shaft torque and the absolute error of the yaw position of each wind turbine synchronously acquired within a preset time window. It also constructs a fatigue accumulation factor by combining the physical length of historical defects and the service time after repair. Through a nonlinear coupling model, risk assessment is conducted to determine the risk index of each wind turbine blade, thereby screening target blades and target wind turbines. The invention also generates a UAV scheduling queue by planning the shortest spatial node sequence with the minimum global flight range of the UAV as a spatial constraint. This effectively changes the existing technology's mode of static scanning with fixed routes and full coverage, avoids wasting UAV resources in healthy areas, achieves precise guidance for inspection, and significantly improves the overall endurance utilization and operational efficiency of global scheduling. This invention also extracts the dominant abnormal parameters of the target blade, maps them to a local high-risk area based on the force mapping relationship, extracts the geometric center of the surface of the area and offsets it by a preset safety distance along the normal to generate a key detection anchor point. Using this anchor point as a spatial reference center, the flight path spacing and image acquisition frequency of the UAV are adjusted according to the risk index to generate an adaptive image acquisition path for defect detection. It fully addresses the characteristics of uneven fatigue damage and highly localized micro-cracks, breaks the limitation of a globally uniform acquisition frequency, guides the UAV to perform adaptive matching scanning in high-risk areas, successfully eliminates the blind spots caused by insufficient accuracy or angle in traditional methods, and greatly improves the effective capture rate of micro-defects. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 A graph showing the relationship between the wind turbine blade risk index and the flight path spacing; Figure 3 This is a schematic diagram of the overall system structure of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0020] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0021] Example: Please see Figures 1-2 The present invention provides a technical solution: A method for detecting defects in wind turbine blades based on unmanned aerial vehicles (UAVs), comprising the following steps: Step 1: Within a preset time window, synchronously acquire the main shaft torque variance, yaw position absolute error, and aerodynamic performance attenuation rate of each wind turbine blade, and determine the physical length of historical defects and the service life after repair for each wind turbine blade.

[0022] Wind turbine blades operate continuously in complex natural wind fields, exhibiting significant dynamic characteristics in their stress and aerodynamic efficiency changes. This embodiment sets the preset time window to a period of 10 minutes prior to the current moment. This time length covers the standard statistical cycle of wind turbine operating data, effectively filtering out random spikes caused by transient extreme gusts and accurately reflecting the current stable operating status of the wind turbine. Simultaneously, the data sampling interval within this time window is set to 1 second (i.e., the sampling frequency is...). This high-frequency sampling setting ensures that a sufficient density of valid running data points is contained within a single time window.

[0023] To accurately quantify the destructive mechanical loads borne by the blades, the instantaneous torque of the main shaft and its arithmetic mean within a preset time window are obtained. The sum of squared deviations between the two is calculated and divided by the total number of sampling points to obtain the variance of the main shaft torque, which characterizes the alternating pulsating load. The frequent fluctuations in the instantaneous torque of the main shaft directly reflect the instability of the wind turbine's absorption of wind energy. This variance value accurately extracts the absolute intensity characterizing the alternating pulsating load. Simultaneously, the actual environmental wind direction angle and the actual yaw angle of the nacelle within the preset time window are obtained, and the mean of the absolute values ​​of the differences between the two is calculated to obtain the absolute error of the yaw position, which quantifies the asymmetric shear force state. The misalignment between the actual wind direction and the nacelle orientation forces the blades to bear continuous lateral thrust, and this error value directly reflects the severity of this off-center load.

[0024] The system obtains the difference between the theoretical average power and the actual average power generation of the entire unit at the same wind speed within a preset time window. The ratio of this difference to the theoretical average power is used as the macroscopic deviation ratio. The system extracts the pitch angle execution deviation, blade root load bending moment, and pitch motor operating current of each wind turbine blade based on feedback from independent pitch sensors, and performs weighted calculations to obtain the state allocation coefficient. The macroscopic deviation ratio is multiplied by the state allocation coefficient of each wind turbine blade to obtain the aerodynamic performance attenuation rate characterizing the independent state of each wind turbine blade. This logic extrapolates from the overall power loss of the unit downwards and allocates power proportionally based on the independent characteristics of each blade. This allows for the precise identification and mapping of the overall unit operating efficiency decline to specific individual blades, thereby transforming the imperceptible microscopic physical damage on the blade surface into quantifiable attenuation characteristics.

[0025] Furthermore, determining the physical length of historical defects and the service life after repair for each wind turbine blade is crucial because the material continuity within the repaired area has been disrupted, essentially representing a high-risk, weak point with stress concentration. The physical length of historical defects determines the initial structural impact span of that area, while the service life after repair accurately records the fatigue accumulation process under alternating loads after the blade is put back into operation.

[0026] Step 2: Obtain the transient mechanical stress corresponding to each wind turbine blade based on the main shaft torque variance and yaw position absolute error. Construct the fatigue accumulation factor of each blade using the physical length of the historical defect and the service time after repair. Perform risk assessment on the transient mechanical stress, fatigue accumulation factor and aerodynamic performance attenuation rate of each wind turbine blade through a nonlinear coupling model to determine the risk index of each wind turbine blade.

[0027] In this embodiment, to eliminate the differences in physical limitations between different models, the preset rated reference torque is set as the reference torque for the unit design. To dedimensionalize and standardize historical defects, the preset reference length constant is set as follows: The reference time constant is set to a standard full calendar year (i.e. In addition, to compensate for the objective influence of the complex on-site temperature and humidity on the microcrack propagation rate of glass fiber composite materials, the preset empirical propagation coefficient was set to 0.15.

[0028] Obtaining transient mechanical stress aims to quantify the external instantaneous destructive force borne by the wind turbine. The steps for obtaining the transient mechanical stress are as follows: The preset rated reference torque is obtained, and the ratio of the square root of the spindle torque variance to the rated reference torque is calculated and calibrated as the torque ratio. This calculation process extracts the absolute fluctuation amplitude of the pulsating load, eliminates model limitations, and thus unifies the relative intensity of alternating pulsating loads.

[0029] The transient mechanical stress is obtained by calculating the hyperbolic cosine value of the absolute error of the yaw position and multiplying it by the torque ratio. Because nacelle yaw misalignment causes the rotor's rotational plane to cut into an asymmetrical wind direction, resulting in extremely uneven stress distribution, the introduction of the hyperbolic cosine function accurately characterizes the extreme eccentric bending moment that amplifies nonlinearly under this yaw condition. This transient mechanical stress serves as the dynamic failure benchmark shared by all blades on the corresponding wind turbine.

[0030] This embodiment quantifies the degree of latent vulnerability of the blade's internal structure due to historical damage by obtaining a fatigue accumulation factor. The steps for obtaining the fatigue accumulation factor are as follows: Based on the preset reference length constant, reference time constant, and empirical expansion coefficient, the first ratio of the physical length of the historical defect corresponding to the wind turbine blade to the reference length constant and the second ratio of the service time after repair to the reference time constant are calculated respectively. The first ratio and the second ratio define the initial physical span of the repair area and the cumulative load time of re-service.

[0031] The first ratio, the second ratio, and the empirical expansion coefficient used to compensate for differences in environmental evolution are multiplied to obtain the defect expansion product. This defect expansion product is then summed with the natural constant to construct a composite fatigue proper number term. By introducing the natural constant, the basic weight of the healthy blades is ensured not to be cleared to zero. The logarithm of the composite fatigue proper number term is calculated with the natural constant as the base to obtain the fatigue accumulation factor of each wind turbine blade, which truly reflects the physical decay law of the nonlinear propagation of hidden microcracks.

[0032] In this embodiment, the risk index is determined by deeply integrating external physical loads, surface aerodynamic losses, and internal structural fatigue to provide quantitative decision-making values ​​for subsequent UAV scheduling. The steps for obtaining the risk index are as follows: The difference between the preset constant and the aerodynamic performance attenuation rate is calculated to obtain the effective aerodynamic retention coefficient characterizing the remaining work capacity of the blade. When minor physical damage occurs on the blade surface, the loss of its aerodynamic efficiency will directly disrupt the original force balance of the three blades of the wind turbine.

[0033] The ratio of the transient mechanical stress to the effective aerodynamic holding coefficient is calculated and calibrated as the attenuated reduced stress characterizing the force amplification effect. The decrease in the effective aerodynamic holding coefficient (the denominator becomes smaller) will, in turn, force the blade to bear more severe structural compensation forces, thereby producing a significant amplification effect of the overall transient force locally.

[0034] The risk index of each wind turbine blade is calculated by multiplying the attenuated reduced stress with the fatigue accumulation factor. The specific calculation formula is as follows: In the formula, For the first Typhoon machine Risk index of each leaf For transient mechanical stress, For the first Variance of the main shaft torque of a typhoon generator. This is the rated reference torque for this type of fan. For the first The absolute error of the yaw position of the typhoon generator. For the first Typhoon machine aerodynamic performance degradation rate of each blade As the fatigue accumulation factor, For the first Typhoon machine The physical length of the historical defects of each blade For the first Typhoon machine The service life of each repaired blade. and These are the reference length constant and the reference time constant, respectively. This is a preset empirical expansion coefficient.

[0035] In complex wind farm operating conditions, blade destructive forces originate from instantaneous loads and local compensatory forces caused by aerodynamic degradation. This method extracts the relative strength of alternating pulsating loads, eliminating model limitations, by calculating the ratio of the square root of the main shaft torque variance to the rated reference torque; it introduces the hyperbolic cosine value of the absolute error of yaw position and multiplies it to obtain transient mechanical stress, accurately characterizing the extreme off-center bending moment that amplifies nonlinearly when yaw distortion increases.

[0036] Meanwhile, aerodynamic degradation on the blade surface disrupts the wind turbine's force balance. An effective aerodynamic retention coefficient is obtained by calculating the difference between a preset constant and the aerodynamic performance degradation rate. Since the loss of aerodynamic efficiency forces damaged blades to bear more severe compensatory forces, thus amplifying the overall external force locally, the transient mechanical stress is divided by this coefficient and calibrated as the attenuated reduced stress.

[0037] Furthermore, the propagation of microcracks within composite materials exhibits nonlinear decay characteristics. This method calculates the ratio of the physical length of historical defects to a reference length constant, and the ratio of the post-repair service life to a reference time constant, and multiplies these ratios by an empirical propagation coefficient to obtain the defect propagation product. The natural constant is then added to this product, and a natural logarithmic operation is performed to obtain the fatigue accumulation factor. When there are no defects, the logarithmic result is one, ensuring that the stress on the healthy blade foundation is not falsely amplified and that the evaluation weights are not zeroed. As the defect size and time increase, the logarithmic function dynamically and accurately quantifies the internal fragility using a nonlinear curvature that closely matches the material's fatigue life. Finally, the attenuated equivalent stress is multiplied by the fatigue accumulation factor to calculate the risk index, deeply intertwining external compensating stress and internal material fragility, and mapping them uniformly into an intuitive engineering evaluation index.

[0038] Step 3: Select wind turbine blades with a risk index greater than the preset safety threshold as target blades, extract the target wind turbines to which the target blades belong and remove duplicate wind turbines, obtain the spatial coordinates of the selected target wind turbines, and use minimizing the total global flight distance of the UAV as the spatial constraint to plan the shortest spatial node sequence that traverses all target wind turbines in sequence to generate the UAV scheduling queue.

[0039] In this embodiment, due to the vast area of ​​actual wind farms and the limited endurance of drones, traditional indiscriminate full-coverage inspections would result in a significant waste of resources for relocation. To achieve targeted allocation of inspection resources, the steps for creating the drone scheduling queue are as follows: Based on the structural design standards and material limits of wind turbine units, this embodiment sets the preset safety threshold to 1.0. When the blade is in an absolutely healthy state with no historical defects and no aerodynamic losses, the logarithmic result of its fatigue accumulation factor and the effective aerodynamic holding coefficient both degenerate to 1.0. At this time, the risk index is determined only by the relative transient mechanical stress characterizing the external load, and this value is strictly less than 1.0 under normal safe operating conditions of the unit. However, when the risk index is greater than or equal to 1.0, it indicates that under the superposition of local force amplification effect and internal microscopic fragility, the combined destructive force borne by the blade has substantially exceeded the elastic safety redundancy of the original design.

[0040] Wind turbine blades with risk indices exceeding a preset safety threshold are grouped into a target blade set, thereby directly isolating healthy operational targets. The equipment number of each blade in the target blade set is analyzed to extract its corresponding target wind turbine number. Since multiple blades on the same wind turbine may simultaneously exceed the preset safety threshold, using a single high-risk blade as an independent navigation endpoint would inevitably lead to overlapping and chaotic flight paths. Therefore, redundant target wind turbine numbers are removed to obtain a deduplicated target wind turbine set. This deduplication logic ensures that the UAV uses the entire wind turbine as a macroscopic spatial dwell node for subsequent close-range operations.

[0041] In actual wind farm operation and maintenance, drones are typically deployed in fixed locations. Therefore, in this embodiment, the preset drone takeoff anchor point is specifically set as the helipad of the wind farm operation and maintenance center. The preset drone takeoff anchor point and the three-dimensional spatial coordinates of each target wind turbine in the deduplicated target wind turbine set are obtained, and these three-dimensional spatial coordinates are defined as independent spatial nodes, thereby mapping the virtual risk list to precise physical navigation anchor points. Based on this, the straight-line distance between each pair of spatial nodes is calculated and a global distance matrix is ​​constructed. This global distance matrix accurately quantifies the spatial physical span between any nodes and the expected relocation power consumption cost.

[0042] Based on the global distance matrix, taking the spatial node corresponding to the UAV takeoff anchor point as the starting node, since battery life is the biggest physical bottleneck limiting the operational efficiency of UAVs, the goal is to minimize the total global flight range of the UAV as the spatial constraint. An iterative optimization is performed with the objective of traversing all target wind turbines in the deduplicated target wind turbine set once. The shortest spatial node sequence is calculated through the above optimization and used as the UAV scheduling queue. This minimizes the power consumption of invalid flight segments during the UAV's transfer scheduling between different positions, thereby maximizing the conversion of limited carrier power into effective inspection and dwell time in high-risk target areas.

[0043] Step 4: According to the arrangement order of the target wind turbines in the UAV scheduling queue, and traverse the target blades contained therein, extract the dominant abnormal parameters of the target blade, map the dominant abnormal parameters to the local high-risk area of ​​the target blade based on the force mapping relationship, extract the surface geometric center of the local high-risk area and offset it along the normal by a preset safe distance to generate key detection anchor points.

[0044] In this embodiment, when the UAV arrives at the target wind turbine according to the scheduling queue, due to the huge size of the wind turbine blades, indiscriminate full-size uniform scanning would consume a lot of resources. Because the physical causes leading to increased risk vary, early microscopic damage is often highly concentrated in specific local areas. To isolate the driving source that forces the blades close to failure, the transient mechanical stress, fatigue accumulation factor, and aerodynamic performance degradation rate corresponding to the target blade are obtained and used as risk parameters.

[0045] Based on the standard operating conditions of the wind turbine, this embodiment sets the baseline values ​​of transient mechanical stress and fatigue accumulation factor under healthy conditions to 1.0. Considering the inherent roughness tolerance and initial drag of the new blade surface, the standard healthy baseline value of the aerodynamic performance degradation rate is set to 0.02. Subsequently, the absolute difference between the actual value of each risk parameter and the preset standard healthy baseline value is calculated, and the deviation of each risk parameter is obtained by dividing the absolute difference by the corresponding standard healthy baseline value. The deviation of each risk parameter is numerically compared, and the parameter with the largest value is extracted and determined as the dominant abnormal parameter of the target blade. This logic unifies risk factors with different physical dimensions into a relative deterioration scale, thereby accurately identifying the most destructive core cause.

[0046] In this embodiment, the specific logic for mapping the dominant abnormal parameters to the local high-risk area of ​​the target blade based on the force mapping relationship, and offsetting it along the normal direction by a preset safety distance to generate key detection anchor points is as follows: When the dominant abnormal parameter is transient mechanical stress, it indicates that the blade is experiencing extreme external physical bending moment. Based on the stress distribution characteristics of the cantilever beam, the local high-risk area is mapped to a preset stress concentration area at the blade root in the target blade spatial coordinate system, specifically defined as the area at the distance from the blade root. to The transition section between the cylinder and the maximum chord length within the extended range; When the dominant abnormal parameter is the aerodynamic performance degradation rate, it indicates a severe decrease in wind energy capture efficiency. The local high-risk area is mapped to the leading edge and suction surface area of ​​the blade from the middle to the tip in the target blade spatial coordinate system, specifically defined as the area from the blade root. to High linear velocity surfaces within the extended range are aerodynamically sensitive zones that are highly susceptible to sand hole spalling and leading edge erosion. When the dominant abnormal parameter is the fatigue accumulation factor, it indicates that the internal hidden crack propagation is the core threat. The local high-risk area is mapped to the historical defect in-situ coordinates of the target blade and the extension area of ​​the surrounding preset radius. According to the physical law of crack propagation in composite materials, this embodiment sets the preset radius to 1.0 meter.

[0047] Subsequently, the surface geometric center of the local high-risk area is extracted, and the unit outward normal vector perpendicular to the tangent plane containing the surface geometric center is obtained. The unit outward normal vector is multiplied by the safety distance to obtain the normal offset vector. In this embodiment, the preset safety distance is set to 10 meters. The normal offset vector is added to the three-dimensional spatial coordinates of the surface geometric center to calculate the key detection anchor point with absolute spatial coordinates. This ensures that the UAV's optical lens can directly face the most vulnerable high-risk surface from the best perpendicular and orthogonal perspective, while simultaneously forcibly reserving sufficient avoidance distance in physical space.

[0048] Step 5: Using the key detection anchor point as the spatial reference center, adjust the flight path spacing and image acquisition frequency of the UAV according to the risk index of the target blade, and generate an adaptive image acquisition path to drive the UAV to perform defect detection along the path.

[0049] In this embodiment, traditional UAV inspections typically employ globally uniform flight path spacing and image acquisition frequency. However, when dealing with wind turbine blades exhibiting significant individual risk variations, this fixed-pathway operation easily generates massive amounts of redundant image data in healthy areas, while in high-risk areas with a high risk of breakage, it misses minute cracks due to insufficient spatial resolution or sampling density. Therefore, dynamically adjusting the flight path spacing and image acquisition frequency based on a pre-quantified risk index aims to overcome the limitations of fixed-pathway inspections and achieve precise matching between local detection accuracy and the physical resources of the carrier aircraft.

[0050] In this embodiment, considering the actual flight performance of industrial-grade UAVs and camera specifications, the preset basic flight path spacing is set to 3.0 meters, the preset minimum safety spacing is set to 1.0 meter, and the preset spacing attenuation adjustment coefficient is set to 0.5; the preset basic image acquisition frequency is set to... (i.e., 1 frame per second), the preset maximum continuous shooting frequency is set to The preset frequency gain adjustment coefficient is set to 1.2.

[0051] The specific logic for adjusting the UAV's flight path spacing and image acquisition frequency based on the risk index of the target blade is as follows: The calculation uses a natural constant as the base, with the exponent value being the negative of the product of the target blade's risk index and a preset spacing attenuation adjustment coefficient. This exponent value is then multiplied by a preset basic flight path spacing to obtain the initial spacing. The initial spacing is compared with a preset minimum safety spacing, and the maximum value is extracted as the UAV's flight path spacing. A higher risk index for the target blade indicates denser and more subtle underlying microscopic physical damage, necessitating a reduction in the physical span between adjacent flight paths to improve image spatial overlap and detection resolution. Simultaneously, introducing a minimum safety spacing as a lower limit avoids the risk of physical collisions caused by gusts of wind when the UAV approaches a damaged surface.

[0052] The risk index is multiplied by a preset frequency gain adjustment coefficient and summed with a constant. The sum is then multiplied by a preset base image acquisition frequency to obtain an initial frequency. This initial frequency is then compared with a preset maximum continuous shooting frequency, and the minimum of the two is extracted as the image acquisition frequency. Since the flight path spacing in high-risk areas is significantly compressed, the UAV needs to perform more intensive spatial maneuvers. Increasing the image acquisition frequency ensures that a sufficient number of high-definition, distortion-free, overlapping frames are captured during complex maneuvers. Furthermore, applying a threshold constraint between the image acquisition frequency and the maximum continuous shooting frequency balances the data buffer limit and storage read / write performance of the electro-optical pod.

[0053] In this embodiment, considering the elongated shape of the wind turbine blades, the preset topology scanning rule is set as a bow-shaped reciprocating scanning strategy that alternates between the spanwise and chordwise directions of the blades. The specific logic for generating the adaptive image acquisition path is as follows: Obtain the preset topology scanning rules and the initial timestamp of the UAV arriving at the key detection anchor point; using the key detection anchor point as the three-dimensional spatial reference origin, extend outward according to the topology scanning rules and the flight path distance of the UAV to construct a serialized spatial relative offset vector, perform spatial vector addition operation on each spatial relative offset vector and the three-dimensional spatial coordinates of the key detection anchor point respectively to generate a serialized three-dimensional spatial flight waypoint, and expand the single stationing anchor point into a three-dimensional operation grid covering the local high-risk surface.

[0054] The reciprocal of the image acquisition frequency is calculated to obtain the time interval between adjacent shutter actions. Using the initial timestamp as a reference, the time intervals are sequentially accumulated to calculate the absolute timestamps corresponding to each 3D space flight waypoint. The 3D space flight waypoints are then sequentially bound to their corresponding absolute timestamps to construct a machine control command stream for generating an adaptive image acquisition path. The specific construction formula is as follows: In the formula, For the final obtained number Typhoon machine Adaptive image acquisition path for each leaf For the first in the path sequence There are 3D space waypoints, among which... For the step index in the path sequence, and The drones reached the first The first waypoint and the first Each waypoint and the absolute timestamp that triggers the camera shutter, where, This is the initial timestamp of the drone arriving at the key detection anchor point. For the first Typhoon machine Key inspection anchor points for each blade Based on the distance between routes The first one constructed with topology scan rules Step space relative offset vector, For the first Typhoon machine The spacing between the flight paths of each blade For the first Typhoon machine Image acquisition frequency per leaf.

[0055] In particular, during detailed local inspections, the distribution of fatigue damage within a single blade is extremely uneven. This method first calculates an exponent value with a base of the natural constant and an exponent equal to the negative of the product of the risk index and the spacing attenuation adjustment coefficient. This exponent is then multiplied by the basic flight path spacing to obtain the initial spacing. This nonlinear negative correlation calculation indicates that areas with higher risk require a reduction in physical span to improve image overlap. The value is then compared to the minimum safety spacing limit and extracted as the flight path spacing, thus forcibly setting a lower safety limit for obstacle avoidance and collision prevention.

[0056] The initial frequency is obtained by multiplying the risk index by the frequency gain adjustment coefficient and adding one, then multiplying it by the base image acquisition frequency. High-risk areas require intensive maneuvers, and this positively correlated gain ensures that enough high-definition overlapping frames are captured per unit time. The initial frequency is compared with the highest continuous shooting frequency, and the minimum value is taken as the image acquisition frequency, taking into account the cache write limit of the camera hardware.

[0057] After establishing the above parameters, the key detection anchor points are used as the three-dimensional spatial reference origin. Spatial relative offset vectors are constructed by extending outward according to the topology scanning rules and flight path spacing. Each offset vector is added to the key detection anchor points to generate serialized three-dimensional flight waypoints, thereby expanding a single dwelling point into a three-dimensional operational grid covering high-risk surfaces.

[0058] Finally, the reciprocal of the image acquisition frequency is calculated to obtain the time interval between adjacent shutter actions. This time interval is then accumulated based on the initial timestamp to calculate the absolute timestamp corresponding to each 3D spatial waypoint. The spatial waypoints are then bound to the timestamp sequence to generate an adaptive image acquisition path, driving the UAV to perform locally detailed scans with appropriate density.

[0059] Table 1 is an example table of drone inspection and scheduling data based on multi-dimensional risk parameters.

[0060] Table 1: Drone Inspection and Scheduling Data Based on Multidimensional Risk Parameters Table 1 shows the data demonstrating the ability of this method to quantitatively assess and control the potential risk levels of different blades under complex wind farm operating conditions. For healthy blades or blades with only mild aerodynamic performance degradation (such as those numbered...), the data is presented in Table 1. Its various risk-causing parameters are stable, and the output risk index is extremely low. Therefore, a wider spacing between flight paths is set. ) and image acquisition frequency ( This reflects maximizing inspection efficiency while ensuring safety redundancy; blades dominated by sudden increases in transient mechanical stress (such as those numbered...) ) exhibits a moderately high risk index ( This led to a gradual reduction in the spacing between flight routes. The steady frequency gain indicates that this method can accurately capture the local risks caused by alternating off-center loading; however, for blades with severe historical defect evolution or high fatigue accumulation factors (such as those numbered...), the method is more effective. Its risk index surged ( Even if the aerodynamic performance degradation rate is still within the normal range, the exponential amplification effect in the nonlinear coupling model triggers extremely stringent anti-missed detection scheduling, thus significantly compressing the flight path spacing (approaching) Extreme values) and the shutter frequency skyrockets (up to 1000 Hz). The table demonstrates that this method breaks through the limitations of traditional single-criteria judgment, comprehensively quantifying abnormal stress, surface wear, and historical damage into a true level of risk. This not only eliminates blind spots in the detection of high-risk hazards but also avoids wasting drone resources in safe areas.

[0061] Please see Figure 3 The present invention also provides a drone-based wind turbine blade defect detection system, which is used to implement the above-mentioned drone-based wind turbine blade defect detection method, including: The data acquisition module is used to synchronously acquire the main shaft torque variance, yaw position absolute error and aerodynamic performance attenuation rate of each wind turbine blade within a preset time window, and determine the physical length of historical defects and the service life after repair of each wind turbine blade. The risk assessment module is used to obtain the transient mechanical stress corresponding to each wind turbine blade based on the main shaft torque variance and the absolute error of the yaw position. It constructs the fatigue accumulation factor of each blade using the physical length of the historical defect and the service time after repair. It performs risk assessment on the transient mechanical stress, fatigue accumulation factor and aerodynamic performance decay rate of each wind turbine blade through a nonlinear coupling model to determine the risk index of each wind turbine blade. The scheduling and planning module is used to select wind turbine blades with a risk index greater than a preset safety threshold as target blades, extract the target wind turbines to which the target blades belong and remove duplicate wind turbines, obtain the spatial coordinates of the selected target wind turbines, and plan the shortest spatial node sequence for traversing all target wind turbines in turn, with the minimum global flight range of the UAV as the spatial constraint, to generate the UAV scheduling queue. The anchor point generation module is used to extract the dominant abnormal parameters of the target blades according to the arrangement order of the target wind turbines in the UAV scheduling queue and to traverse the target blades contained therein. Based on the force mapping relationship, the dominant abnormal parameters are mapped to the local high-risk area of ​​the target blade. The surface geometric center of the local high-risk area is extracted and offset by a preset safety distance along the normal to generate key detection anchor points. The path generation module is used to adjust the flight path spacing and image acquisition frequency of the UAV based on the risk index of the target blade, using the key detection anchor point as the spatial reference center, and to generate an adaptive image acquisition path to drive the UAV to perform defect detection along the path.

[0062] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0063] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0064] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0065] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for detecting defects in wind turbine blades based on unmanned aerial vehicles (UAVs), characterized in that, The specific steps include: Step 1: Within a preset time window, synchronously acquire the main shaft torque variance, yaw position absolute error, and aerodynamic performance attenuation rate of each wind turbine blade, and determine the physical length of historical defects and the service life after repair of each wind turbine blade. Step 2: Obtain the transient mechanical stress corresponding to each wind turbine blade based on the main shaft torque variance and yaw position absolute error. Construct the fatigue accumulation factor of each blade using the physical length of the historical defect and the service time after repair. Perform risk assessment on the transient mechanical stress, fatigue accumulation factor and aerodynamic performance decay rate of each wind turbine blade through a nonlinear coupling model to determine the risk index of each wind turbine blade. Step 3: Select wind turbine blades with a risk index greater than the preset safety threshold as target blades, extract the target wind turbines to which the target blades belong and remove duplicate wind turbines, obtain the spatial coordinates of the selected target wind turbines, and use minimizing the total global flight distance of the UAV as a spatial constraint to plan the shortest spatial node sequence that traverses all target wind turbines in sequence to generate the UAV scheduling queue. Step 4: According to the arrangement order of the target wind turbines in the UAV scheduling queue, and traverse the target blades contained therein, extract the dominant abnormal parameters of the target blade, map the dominant abnormal parameters to the local high-risk area of ​​the target blade based on the force mapping relationship, extract the surface geometric center of the local high-risk area and offset it along the normal by a preset safe distance to generate key detection anchor points. Step 5: Using the key detection anchor point as the spatial reference center, adjust the flight path spacing and image acquisition frequency of the UAV according to the risk index of the target blade, and generate an adaptive image acquisition path to drive the UAV to perform defect detection along the path.

2. The method for detecting defects in wind turbine blades based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: The preset time window is a window that traces back a preset time length from the current time as the end point, wherein the preset time length covers at least one standard statistical period of wind turbine operation data; The logic for synchronously acquiring the main shaft torque variance, yaw position absolute error, and aerodynamic performance attenuation rate of each wind turbine blade is as follows: Obtain the instantaneous spindle torque and its arithmetic mean within a preset time window, calculate the sum of squared deviations between the two, and divide the sum of squared deviations by the total number of sampling points within the preset time window to obtain the spindle torque variance characterizing the alternating pulsating load. Obtain the actual environmental wind direction angle and the actual yaw angle of the cabin within a preset time window, calculate the mean of the absolute values ​​of the differences between the two, and obtain the absolute error of the yaw position in the quantified asymmetric shear force state. Obtain the actual average power generation of the whole machine within a preset time window and the theoretical average power under the same wind speed, calculate the difference between the theoretical average power and the actual average power generation of the whole machine, and use the ratio of the difference to the theoretical average power as the macro deviation ratio. Extract the pitch angle execution deviation, blade root load bending moment, and pitch motor operating current of each wind turbine blade based on feedback from the independent pitch sensor, and perform weighted calculation to obtain the state allocation coefficient. Multiply the macroscopic deviation ratio with the state allocation coefficient of each wind turbine blade to obtain the aerodynamic performance attenuation rate characterizing the independent state of each wind turbine blade.

3. The method for detecting defects in wind turbine blades based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: The steps for obtaining the transient mechanical stress are as follows: Obtain the preset rated reference torque, calculate the ratio of the square root of the spindle torque variance to the rated reference torque, and calibrate it as the torque ratio. The hyperbolic cosine value of the absolute error of the yaw position is calculated and multiplied with the torque ratio to obtain the transient mechanical stress, which is then used as the transient mechanical stress shared by all blades on the corresponding wind turbine. The steps for obtaining the fatigue accumulation factor are as follows: Based on the preset reference length constant, reference time constant, and empirical expansion coefficient, calculate the first ratio of the physical length of the historical defect of the wind turbine blade to the reference length constant, and the second ratio of the service time after repair to the reference time constant. The first ratio, the second ratio, and the empirical expansion coefficient are multiplied to obtain the defect expansion product. The defect expansion product is then summed with the natural constant to construct a composite fatigue proper number term. The logarithm of the composite fatigue proper number term is calculated with the natural constant as the base to obtain the fatigue accumulation factor of each wind turbine blade.

4. The method for detecting defects in wind turbine blades based on unmanned aerial vehicles (UAVs) according to claim 3, characterized in that: The steps for obtaining the risk index are as follows: Calculate the difference between the preset constant and the aerodynamic performance decay rate to obtain the effective aerodynamic retention coefficient characterizing the remaining work capacity of the blade; Calculate the ratio of the transient mechanical stress to the effective aerodynamic holding coefficient and calibrate it as the attenuated reduced stress characterizing the force amplification effect; The risk index of each wind turbine blade is calculated by multiplying the attenuated reduced stress with the fatigue accumulation factor. The specific calculation formula is as follows: In the formula, For the first Typhoon machine Risk index of each leaf For the first Variance of the main shaft torque of a typhoon generator. This is the rated reference torque for this type of fan. For the first The absolute error of the typhoon's yaw position. For the first Typhoon machine aerodynamic performance degradation rate of each blade For the first Typhoon machine The physical length of the historical defects of each blade For the first Typhoon machine The service life of each repaired blade. and These are the reference length constant and the reference time constant, respectively. This is a preset empirical expansion coefficient.

5. The method for detecting defects in wind turbine blades based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: The steps for creating the drone scheduling queue are as follows: The wind turbine blades with risk indices greater than a preset safety threshold are grouped into a target blade set. The equipment number of each blade in the target blade set is parsed, the corresponding target wind turbine number is extracted, and redundant target wind turbine numbers in the target wind turbine numbers are removed to obtain a deduplicated target wind turbine set. Obtain the preset UAV takeoff anchor point and the three-dimensional spatial coordinates of each target wind turbine in the deduplicated target wind turbine set, and define the three-dimensional spatial coordinates as independent spatial nodes. Calculate the spatial straight-line distance between each pair of spatial nodes and construct a global distance matrix. Based on the global distance matrix, taking the spatial node corresponding to the UAV takeoff anchor point as the starting node, taking the minimization of the total global flight distance of the UAV as the spatial constraint, and taking traversing all target wind turbines in the deduplicated target wind turbine set once as the solution objective, iterative optimization is performed to calculate the shortest spatial node sequence, and the shortest spatial node sequence is used as the UAV scheduling queue.

6. The method for detecting defects in wind turbine blades based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: The judgment logic for the dominant abnormal parameter is as follows: The transient mechanical stress, fatigue accumulation factor and aerodynamic performance attenuation rate of the target blade are obtained and used as risk parameters. The absolute difference between the actual value of each risk parameter and the preset standard health benchmark value is calculated. The absolute difference is divided by the corresponding standard health benchmark value to obtain the deviation of each risk parameter. The deviation of each risk parameter is numerically compared and the parameter with the largest value is extracted and determined as the dominant abnormal parameter of the target blade.

7. The method for detecting defects in wind turbine blades based on unmanned aerial vehicles (UAVs) according to claim 6, characterized in that: The specific logic for mapping the dominant abnormal parameters to the local high-risk area of ​​the target blade based on the force mapping relationship, and offsetting them along the normal direction by a preset safety distance to generate key detection anchor points is as follows: When the dominant abnormal parameter is transient mechanical stress, the local high-risk area is mapped to the preset stress concentration area at the root of the blade in the target blade spatial coordinate system; When the dominant abnormal parameter is the aerodynamic performance decay rate, the local high-risk area is mapped to the leading edge and suction surface area of ​​the blade from the middle to the tip in the target blade spatial coordinate system. When the dominant anomaly parameter is the fatigue accumulation factor, the local high-risk area is mapped to the historical defect in-situ coordinates of the target blade and the extended area of ​​its surrounding preset radius. Obtain the unit outward normal vector perpendicular to the tangent plane where the surface geometric center is located. Multiply the unit outward normal vector with the safety distance to obtain the normal offset vector. Add the normal offset vector with the three-dimensional spatial coordinates of the surface geometric center to calculate the key detection anchor point with absolute spatial coordinates.

8. The method for detecting defects in wind turbine blades based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: The specific logic for adjusting the UAV's flight path spacing and image acquisition frequency based on the risk index of the target blade is as follows: Calculate the index value with the natural constant as the base and the negative of the product of the risk index of the target blade and the preset spacing attenuation adjustment coefficient as the index. Then, multiply the index value with the preset basic flight path spacing to obtain the initial spacing. Compare the initial spacing with the preset minimum limit safety spacing and extract the maximum value of the two as the flight path spacing of the UAV. The risk index is multiplied by a preset frequency gain adjustment coefficient and summed with a constant. The sum is then multiplied by a preset base image acquisition frequency to obtain an initial frequency. The initial frequency is then compared with a preset maximum continuous shooting frequency, and the minimum value between the two is extracted as the image acquisition frequency.

9. The method for detecting defects in wind turbine blades based on unmanned aerial vehicles (UAVs) according to claim 8, characterized in that: The specific logic for generating the adaptive image acquisition path is as follows: Obtain the preset topology scanning rules and the initial timestamp of the UAV arriving at the key detection anchor point; Using the key detection anchor points as the three-dimensional space reference origin, and extending outward according to the topology scanning rules and the flight path spacing of the UAV, a serialized spatial relative offset vector is constructed. The spatial relative offset vectors and the three-dimensional spatial coordinates of the key detection anchor points are respectively subjected to spatial vector addition operations to generate serialized three-dimensional spatial flight waypoints. The reciprocal of the image acquisition frequency is calculated to obtain the time interval between adjacent shutter actions. Using the initial timestamp as a reference, the time intervals are sequentially accumulated to calculate the absolute timestamps corresponding to each 3D space flight waypoint. The 3D space flight waypoints are then sequentially bound to their corresponding absolute timestamps to construct a machine control command stream for generating an adaptive image acquisition path. The specific construction formula is as follows: In the formula, For the final obtained number Typhoon machine Adaptive image acquisition path for each leaf For the first in the path sequence There are 3D space waypoints, among which... For the step index in the path sequence, and The drones reached the first The first waypoint and the first Each waypoint and the absolute timestamp that triggers the camera shutter, where, This is the initial timestamp of the drone arriving at the key detection anchor point. For the first Typhoon machine Key inspection anchor points for each blade Based on the distance between routes The first one constructed with topology scan rules Step space relative offset vector, For the first Typhoon machine The spacing between the flight paths of each blade For the first Typhoon machine Image acquisition frequency per leaf.

10. A wind turbine blade defect detection system based on unmanned aerial vehicles (UAVs), characterized in that: The UAV-based wind turbine blade defect detection system is used to implement the UAV-based wind turbine blade defect detection method according to any one of claims 1-9, comprising: The data acquisition module is used to synchronously acquire the main shaft torque variance, yaw position absolute error and aerodynamic performance attenuation rate of each wind turbine blade within a preset time window, and determine the physical length of historical defects and the service life after repair of each wind turbine blade. The risk assessment module is used to obtain the transient mechanical stress corresponding to each wind turbine blade based on the main shaft torque variance and the absolute error of the yaw position. It constructs the fatigue accumulation factor of each blade using the physical length of the historical defect and the service time after repair. It performs risk assessment on the transient mechanical stress, fatigue accumulation factor and aerodynamic performance decay rate of each wind turbine blade through a nonlinear coupling model to determine the risk index of each wind turbine blade. The scheduling and planning module is used to select wind turbine blades with a risk index greater than a preset safety threshold as target blades, extract the target wind turbines to which the target blades belong and remove duplicate wind turbines, obtain the spatial coordinates of the selected target wind turbines, and plan the shortest spatial node sequence for traversing all target wind turbines in turn, with the minimum global flight range of the UAV as the spatial constraint, to generate the UAV scheduling queue. The anchor point generation module is used to extract the dominant abnormal parameters of the target blades according to the arrangement order of the target wind turbines in the UAV scheduling queue and to traverse the target blades contained therein. Based on the force mapping relationship, the dominant abnormal parameters are mapped to the local high-risk area of ​​the target blade. The surface geometric center of the local high-risk area is extracted and offset by a preset safety distance along the normal to generate key detection anchor points. The path generation module is used to adjust the flight path spacing and image acquisition frequency of the UAV based on the risk index of the target blade, using the key detection anchor point as the spatial reference center, and to generate an adaptive image acquisition path to drive the UAV to perform defect detection along the path.

Images