Unmanned aerial vehicle intelligent identification system for fan blade defects

Abstract

The invention provides an unmanned aerial vehicle intelligent identification system for fan blade defects. The system comprises an inspection tracking module, a data acquisition module, a defect identification module and a data integration module. The inspection tracking module is used for planning and tracking an inspection route of the unmanned aerial vehicle; the data acquisition module is used for acquiring image information of fan blades; the defect identification module is used for analyzing and identifying defect information of the fan blade; the data integration module is used for integrating the defect information of the fan blade to generate a visual result and a maintenance suggestion; according to the method, efficient detection, accurate identification and visual analysis of the fan blade defects are realized, and the inspection efficiency and the scientificity of maintenance decision are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation equipment detection systems, and in particular to an unmanned aerial vehicle intelligent recognition system for defects in wind turbine blades. Background Art

[0002] With the rapid development of the wind power industry, wind turbine blades are key components of wind power generation equipment, and their operating status is directly related to the safety and efficiency of wind turbines. However, due to the long-term exposure of wind turbine blades to complex natural environments, they are affected by multiple factors such as strong winds, dust, rain, snow, and ultraviolet rays, and are prone to cracks, corrosion, and shedding. If these defects are not discovered and handled in time, they may lead to reduced blade performance or even structural damage, which in turn affects the operational reliability of the entire wind turbine. Traditional wind turbine blade inspection methods mainly rely on manual inspections or image acquisition with the help of ground equipment. These methods have problems such as low efficiency, limited coverage, and reliance on manual experience, and are difficult to meet the large-scale, multi-frequency inspection needs of wind farms. At the same time, the analysis of inspection data usually relies on manual judgment, lacking efficient, unified, and standardized processing methods, which further increases the difficulty and cost of maintenance.

[0003] According to the related disclosed technical solutions, the technology with the publication number CN118188349A proposes a method, device and system for detecting defects of wind turbine blades based on drones, wherein the drone is equipped with a camera and a spectrometer, and the detection method includes: obtaining a wind turbine blade image obtained by photographing the wind turbine blade with a camera and a wind turbine blade spectrum obtained by scanning the wind turbine blade with a spectrometer; identifying the wind turbine blade image and the wind turbine blade spectrum by a deep learning-based defect detection model to obtain surface defect information of the wind turbine blade; using a deep learning-based defect detection model to identify the wind turbine blade image and the wind turbine blade spectrum, the surface defect information of the wind turbine blade can be obtained quickly and accurately; the problem that the current wind turbine blade detection technology based on drones mainly relies on manual analysis of images and lacks effective automatic recognition and analysis is solved; although the scheme realizes the automatic recognition of surface defects of wind turbine blades, it only relies on image and spectral data for detection, does not involve the severity assessment and cause analysis of defects, and the analysis of the criticality and impact range of defects is relatively vague; in addition, the scheme does not provide the functions of defect priority sorting and maintenance suggestion generation, and it is difficult to support subsequent maintenance optimization and resource allocation. Summary of the invention

[0004] The purpose of the present invention is to propose a UAV intelligent identification system for fan blade defects in view of the current shortcomings.

[0005] The present invention adopts the following technical solution:

[0006] A UAV intelligent identification system for fan blade defects, the system comprising an inspection tracking module, a data acquisition module, a defect identification module and a data integration module; the inspection tracking module is used to plan and track the inspection route of the UAV; the data acquisition module is used to collect image information of the fan blades; the defect identification module is used to analyze and identify the defect information of the fan blades; the data integration module is used to integrate the defect information of the fan blades to generate visualization results and maintenance suggestions.

[0007] The inspection and tracking module includes a route planning unit and a track tracking unit; the route planning unit is used to generate an inspection route according to the wind turbine distribution and inspection requirements; the track tracking unit is used to monitor the flight status of the UAV in real time and control the UAV to navigate along the inspection route.

[0008] The data acquisition module includes an image acquisition unit and a data transmission unit; the image acquisition unit is used to acquire image information of the wind turbine blades, and the data transmission unit is used to transmit the image information to a subsequent processing module on the ground side through wireless communication technology.

[0009] The defect recognition module and the data integration module are arranged on the ground side; the defect recognition module includes an image preprocessing unit and a defect recognition unit; the image preprocessing unit is used to preprocess the original wind blade image information collected by the UAV; the defect recognition unit is used to analyze the preprocessed wind blade image information to identify the specific defect type and defect position of the wind blade.

[0010] Furthermore, the data integration module includes a data storage unit, a maintenance suggestion generation unit, a defect cause analysis unit and a display unit; the data storage unit is used to store image information and defect information of the wind turbine blades; the maintenance suggestion generation unit is used to generate maintenance suggestions for the wind turbine blades; the defect cause analysis unit is used to analyze the cause pattern of defects in the wind turbine blades; and the display unit is used to visualize the content collected and analyzed by the display system.

[0011] Furthermore, the defect recognition unit completes the classification of wind turbine blade defects and the sorting of defect priorities through a multi-task learning model; the framework of the multi-task learning model includes:

[0012] Shared feature layer: used to obtain input image information and extract global features related to wind blade defect classification and priority sorting from the input image information;

[0013] Defect classification branch layer: used to analyze global features, obtain the specific types of fan blade defects and output the corresponding category probabilities;

[0014] Priority sorting branch layer: used to evaluate the severity of defects and output a priority score based on the severity of the defect type. The priority score means the urgency of defect repair. A higher score means that it needs to be handled first.

[0015] Furthermore, the multi-task learning model uses a joint loss function to optimize the defect classification and priority sorting tasks; the joint loss function is expressed as follows:

[0016] L=α·L cls +β·L bank ;

[0017] Among them, L is the joint loss function; α is the weight of the defect classification task, β is the weight of the priority sorting task, and L cls is the loss function value of the defect classification task, L bank is the loss function value of the priority sorting task; for α and β, it satisfies:

[0018]

[0019] Furthermore, the defect recognition unit inputs the fan blade image information corresponding to each fan blade position into the multi-task learning model, thereby outputting the specific defect type of the fan blade at that position and the priority score corresponding to the defect type.

[0020] Beneficial effects achieved by the present invention:

[0021] By combining a multi-task learning model, this solution can achieve accurate classification and priority assessment of wind turbine blade defects, provide a quantitative analysis basis for the urgency of repair, significantly improve inspection efficiency, and reduce maintenance costs; by introducing a dynamic weight allocation mechanism in the joint loss function design in the multi-task learning model, the weights of classification and sorting tasks are automatically adjusted according to task difficulty, achieving balanced optimization of tasks, further improving the overall performance of the model and the reliability of output results; in addition, the solution also includes generating maintenance recommendations for wind turbine blades and analyzing the cause patterns of defects, thereby optimizing maintenance strategies and improving the reliability and operation efficiency of wind turbines. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The present invention can be further understood from the following description in conjunction with the accompanying drawings. The components in the figures are not necessarily drawn to scale, but the emphasis is placed on illustrating the principles of the embodiments. In different views, the same reference numerals designate corresponding parts.

[0023] Figure 1 It is a schematic diagram of the overall module of the present invention.

[0024] Figure 2 Schematic diagram of the multi-task learning model framework of the present invention.

[0025] Figure 3 Schematic diagram of the working process of the defect cause analysis unit of the present invention. DETAILED DESCRIPTION

[0026] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with its embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention; for those skilled in the art, other systems, methods and / or features of the present embodiment will become apparent after reviewing the following detailed description; it is intended that all such additional systems, methods, features and advantages are included in this specification; included within the scope of the present invention and protected by the appended claims; additional features of the disclosed embodiments are described in the following detailed description, and these features will be apparent from the following detailed description.

[0027] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right" and the like indicate directions or positional relationships based on the directions or positional relationships shown in the drawings, it is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or component referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limitations on this patent. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0028] Embodiment 1:

[0029] like Figure 1 As shown, this embodiment provides a UAV intelligent identification system for fan blade defects, the system comprising an inspection tracking module, a data acquisition module, a defect identification module and a data integration module; the inspection tracking module is used to plan and track the inspection route of the UAV; the data acquisition module is used to collect image information of the fan blade; the defect identification module is used to analyze and identify the defect information of the fan blade; the data integration module is used to integrate the defect information of the fan blade to generate visualization results and maintenance suggestions;

[0030] The inspection and tracking module includes a route planning unit and a track tracking unit; the route planning unit is used to generate an inspection route according to the wind turbine distribution and inspection requirements; the track tracking unit is used to monitor the flight status of the UAV in real time and control the UAV to navigate along the inspection route;

[0031] The data acquisition module includes an image acquisition unit and a data transmission unit; the image acquisition unit is used to acquire image information of the wind turbine blades, and the data transmission unit is used to transmit the image information to a subsequent processing module on the ground side through wireless communication technology;

[0032] The defect recognition module and the data integration module are arranged on the ground side; the defect recognition module includes an image preprocessing unit and a defect recognition unit; the image preprocessing unit is used to preprocess the original wind blade image information collected by the drone; the defect recognition unit is used to analyze the preprocessed wind blade image information to identify the specific defect type and defect location of the wind blade;

[0033] Furthermore, the data integration module includes a data storage unit, a maintenance suggestion generation unit, a defect cause analysis unit and a display unit; the data storage unit is used to store image information and defect information of the fan blade; the maintenance suggestion generation unit is used to generate maintenance suggestions for the fan blade; the defect cause analysis unit is used to analyze the cause mode of the fan blade defect; the display unit is used to visualize the content collected and analyzed by the system;

[0034] Furthermore, the image preprocessing unit preprocesses the original wind turbine blade image information by:

[0035] Image denoising: remove clutter introduced by sensor noise or environmental interference during image capture;

[0036] Image enhancement: Improve the brightness and contrast of the image to make the defects on the surface of the wind turbine blades clearer;

[0037] Image geometry correction: Correct image deformation caused by the drone shooting angle or lens distortion, so that the wind turbine blades appear in the correct proportion in the image;

[0038] Image segmentation and region extraction: extract the main area of ​​the wind turbine blade in the image, remove background information, and reduce the complexity of subsequent processing;

[0039] Image normalization: normalize the image size;

[0040] Furthermore, the defect recognition unit completes the classification of wind turbine blade defects and the sorting of defect priorities through a multi-task learning model; Figure 2 As shown, the framework of the multi-task learning model includes:

[0041] Shared feature layer: used to obtain input image information and extract global features related to fan blade defect classification and priority sorting from the input image information; its form is expressed as follows:

[0042] F = CNN(X; θ s );

[0043] Among them, F is the extracted global feature, X is the input image information, that is, the fan blade image information preprocessed by the image preprocessing unit; θ s is the model parameter of the shared feature layer; CNN() is the operation function of the deep convolutional neural network;

[0044] Defect classification branch layer: used to analyze global features, obtain the specific types of fan blade defects and output the corresponding category probabilities; its form is expressed as follows:

[0045]

[0046] in, is the predicted probability distribution of defect type; W cls and b cls is the model parameter of the defect branch classification layer; Softmax() is the model activation function;

[0047] Priority sorting branch layer: used to evaluate the severity of defects and output a priority score based on the severity of the defect type. The priority score means the urgency of defect repair. A higher score indicates that it needs to be handled first. Its form is expressed as follows:

[0048]

[0049] in, W is the defect priority score, which takes a continuous value; rank and b rank Model parameters of branch layers for prioritization;

[0050] Furthermore, the multi-task learning model uses a joint loss function to optimize the defect classification and priority sorting tasks; the joint loss function is expressed as follows:

[0051] L=α·L cls +β·L bank ;

[0052] Among them, L is the joint loss function; α is the weight of the defect classification task, β is the weight of the priority sorting task, and L cls is the loss function value of the defect classification task, L bank is the loss function value of the priority sorting task; for α and β, it satisfies:

[0053]

[0054] Furthermore, the loss function of the defect classification task can be obtained through cross entropy calculation;

[0055] Furthermore, the loss function of the priority sorting task can be obtained by a mean square error calculation;

[0056] Furthermore, the defect recognition unit inputs the fan blade image information corresponding to each fan blade position into the multi-task learning model, thereby outputting the specific defect type of the fan blade at the position and the priority score corresponding to the defect type;

[0057] This solution combines a multi-task learning model to complete the classification of wind turbine blade defects and the evaluation of defect priorities, thereby achieving accurate identification of defect types and quantitative analysis of repair urgency, significantly improving inspection efficiency and reducing inspection and maintenance costs. By setting a shared feature layer in the multi-task learning model, the model shares global features in classification and sorting tasks, greatly reducing computational redundancy and ensuring efficient collaboration between tasks. By dynamically allocating the weights of defect classification tasks and priority sorting tasks according to task difficulty in the multi-task joint loss function, balanced optimization of classification and sorting tasks is achieved, further improving the overall performance of the multi-task learning model and the reliability of the output results.

[0058] Embodiment 2:

[0059] This embodiment should be understood to include at least all the features of any of the above embodiments, and further improve upon them;

[0060] This embodiment provides a UAV intelligent identification system for fan blade defects, the system comprising an inspection tracking module, a data acquisition module, a defect identification module and a data integration module; the inspection tracking module is used to plan and track the inspection route of the UAV; the data acquisition module is used to collect image information of the fan blade; the defect identification module is used to analyze and identify the defect information of the fan blade; the data integration module is used to integrate the defect information of the fan blade to generate visualization results and maintenance suggestions;

[0061] Furthermore, the data integration module includes a data storage unit, a maintenance suggestion generation unit, a defect cause analysis unit and a display unit; the data storage unit is used to store image information and defect information of the fan blades; the maintenance suggestion generation unit is used to generate maintenance suggestions for the fan blades; the defect cause analysis unit is used to analyze the cause mode of the fan blade defects; the display unit is used to visualize the system acquisition and analysis content;

[0062] Further, the fan blade image information stored in the data storage unit includes image information corresponding to each position of the fan blade, and the defect information includes a defect type corresponding to each position of the fan blade and a priority score corresponding to the defect type;

[0063] Furthermore, the maintenance suggestion generating unit generates maintenance suggestions sorted by priority scores according to the defect types and priority scores of the current wind turbine blades in combination with a third-party expert knowledge base, and provides a detailed maintenance operation plan;

[0064] Further, such as Figure 3 As shown, the defect cause analysis unit completes the analysis of the cause mode of the fan blade defect in the following manner:

[0065] S1: Divide the fan blades into multiple areas according to their structural characteristics and functional distribution;

[0066] S2: Setting a fixed monitoring period and obtaining defect information at each wind turbine blade area position within the monitoring period;

[0067] S3: For each area position, calculate the frequency index of each defect type at the area position:

[0068]

[0069] Among them, I k is the frequency index of the k-th defect type at the location in the region, and the frequency index reflects the occurrence frequency and severity of the defect type at the location in the region; S priority,i is the priority score of the k-th defect type when it occurs for the i-th time, which is obtained by the defect recognition unit in the historical acquisition process; γ is a pre-set adjustment coefficient, which is used to control the influence of the defect occurrence time on the frequency index, and is set by pre-experiment; t end is the end time of the monitoring period, t total,i is the time point when the k-th defect type occurs for the i-th time, T total is the total length of the monitoring period;

[0070] S4: For each regional location, obtain the frequency and weight indexes of all defect types corresponding to it, and sort them from high to low, extract the top L defect types in the sorting, and use them as the key defect types of the regional location. L can be set according to user needs; the key defect type combines the frequency and severity of the defect, and reflects the multiple defects that have the greatest impact on the reliability of wind turbine operation during the monitoring period;

[0071] S5: For the corresponding key defect type in each area, combined with the defect cause records in the historical database, match the cause pattern of the same defect type;

[0072] S6: Verify the matched causal patterns in combination with the operating data and environmental data in the current monitoring period, and filter out causal patterns that do not meet the current operating conditions or environmental conditions;

[0073] S7: Sort the filtered cause patterns according to the strength of their association with the key defect types to clarify the priority of each cause pattern:

[0074] For a certain filtered causal pattern j:

[0075]

[0076] Among them, R j is the priority coefficient of the jth causal mode, M k,j is the strength of association between the kth defect type and the jth cause pattern in the key defect types, which is pre-set according to the matching degree between the defect cause records and the defect types in the historical database;

[0077] S8: For each regional location, obtain the corresponding cause mode arranged from large to small according to the cause priority coefficient as the cause report of the defect generated in the regional location;

[0078] Furthermore, part of the running code of the defect cause analysis unit is as follows:

[0079]

[0080]

[0081]

[0082]

[0083] This solution conducts a comprehensive assessment of the frequency and severity of defects, first obtaining key defect types and clarifying their impact on wind turbine operation reliability; then, combining the correlation between key defect types and cause patterns, matching and verifying cause patterns, and generating defect cause analysis reports according to priority, providing a scientific basis for optimizing maintenance strategies and resource allocation;

[0084] Furthermore, the content that the display unit visually displays to the user specifically includes:

[0085] Construct a three-dimensional display model based on the geographic information system to display the defect distribution of each area of ​​each wind turbine in the wind farm; and insert the defect information and image information into the corresponding position in the three-dimensional display model;

[0086] Display maintenance suggestions for each area of ​​each wind turbine in the wind farm;

[0087] The key defect types and their cause patterns in each area of ​​each wind turbine in the wind farm are listed and sorted by priority.

[0088] The contents disclosed above are only preferred feasible embodiments of the present invention, and do not limit the protection scope of the present invention. Therefore, all equivalent technical changes made using the contents of the present invention specification and drawings are included in the protection scope of the present invention. In addition, the elements therein can be updated as technology develops.

Claims

1. An unmanned aerial vehicle intelligent identification system for fan blade defects, characterized in that: The system includes an inspection and tracking module, a data acquisition module, a defect recognition module and a data integration module; the inspection and tracking module is used to plan and track the inspection route of the drone; the data acquisition module is used to collect image information of the wind turbine blades; The defect recognition module is used to analyze and identify the defect information of the fan blades; the data integration module is used to integrate the defect information of the fan blades to generate visualization results and maintenance suggestions; The inspection and tracking module includes a route planning unit and a track tracking unit; the route planning unit is used to generate an inspection route according to the wind turbine distribution and inspection requirements; The track tracking unit is used to monitor the flight status of the UAV in real time and control the UAV to navigate along the inspection route; The data acquisition module includes an image acquisition unit and a data transmission unit; the image acquisition unit is used to acquire image information of the wind turbine blades, and the data transmission unit is used to transmit the image information to a subsequent processing module on the ground side through wireless communication technology; The defect recognition module and the data integration module are arranged on the ground side; The defect recognition module includes an image preprocessing unit and a defect recognition unit; the image preprocessing unit is used to preprocess the original wind turbine blade image information collected by the drone; The defect recognition unit is used to analyze the preprocessed fan blade image information to identify the specific defect type and defect position of the fan blade.

2. The UAV intelligent identification system for wind turbine blade defects according to claim 1 is characterized in that: The data integration module includes a data storage unit, a maintenance suggestion generation unit, a defect cause analysis unit and a display unit; the data storage unit is used to store image information and defect information of the wind turbine blades; the maintenance suggestion generation unit is used to generate maintenance suggestions for the wind turbine blades; The defect cause analysis unit is used to analyze the cause pattern of fan blade defects; the display unit is used to visualize the content collected and analyzed by the display system.

3. The UAV intelligent identification system for wind turbine blade defects according to claim 2 is characterized in that: The defect recognition unit completes the classification of fan blade defects and the sorting of defect priorities through a multi-task learning model; The framework of the multi-task learning model includes: Shared feature layer: used to obtain input image information and extract global features related to wind blade defect classification and priority sorting from the input image information; Defect classification branch layer: used to analyze global features, obtain the specific types of fan blade defects and output the corresponding category probabilities; Priority sorting branch layer: used to evaluate the severity of defects and output a priority score based on the severity of the defect type. The priority score means the urgency of defect repair. A higher score means that it needs to be handled first.

4. The UAV intelligent identification system for wind turbine blade defects according to claim 3 is characterized in that: The multi-task learning model uses a joint loss function to optimize the defect classification and prioritization tasks; the joint loss function is expressed as follows: L=α·L cls +β·L bank ; Among them, L is the joint loss function; α is the weight of the defect classification task, β is the weight of the priority sorting task, and L cls is the loss function value of the defect classification task, L bank is the loss function value of the priority sorting task; for α and β, it satisfies:

5. The UAV intelligent identification system for wind turbine blade defects according to claim 4 is characterized in that: The defect recognition unit inputs the fan blade image information corresponding to each fan blade position into the multi-task learning model, thereby outputting the specific defect type of the fan blade at that position and the priority score corresponding to the defect type.

Images