Wind driven generator blade surface defect identification method, system, equipment and medium

Abstract

The embodiment of the invention provides a wind driven generator blade surface defect identification method, system and device and a medium. The method comprises the steps of obtaining a wind turbine blade surface image; inputting the surface image of the wind driven generator blade into a pre-trained fan blade surface defect identification model, and outputting positioning, classification and severity information of defects in the surface image of the wind driven generator blade; and generating a defect identification report according to the positioning, classification and severity information. According to the embodiment of the invention, through an image recognition technology based on YOLOv8, the blade defect detection time is remarkably shortened through an automatic detection process; omission detection and false detection can be effectively avoided due to the high-precision recognition capability of the YOLOv8 model; the dependence of manual inspection is reduced, the labor cost is reduced, and meanwhile, the sudden maintenance requirement is reduced through early warning; and hidden dangers can be timely found through regular monitoring, and safety accidents caused by blade faults are avoided.

Description

Technical Field

[0001] The embodiments of the present disclosure belong to the technical field of wind power equipment maintenance, and specifically relate to a method, system, equipment and medium for identifying surface defects of wind turbine blades. Background Art

[0002] As an important part of renewable energy, wind power generation is widely used around the world. However, wind turbine blades are exposed to harsh environments for a long time and are prone to various surface defects. If these defects are not discovered and handled in time, they will seriously affect the power generation efficiency and equipment life. Traditional manual visual inspection methods are time-consuming and labor-intensive, and are greatly affected by human factors, making it difficult to meet the efficient operation and maintenance needs of large-scale wind farms. Summary of the invention

[0003] The embodiments of the present disclosure aim to solve at least one of the technical problems existing in the prior art and provide a method, system, device and medium for identifying surface defects of wind turbine blades.

[0004] One aspect of the present disclosure provides a method for identifying surface defects of a wind turbine blade, the method comprising:

[0005] Acquire the surface image of wind turbine blades;

[0006] Inputting the wind turbine blade surface image into a pre-trained wind turbine blade surface defect recognition model, and outputting location, classification, and severity information of defects in the wind turbine blade surface image;

[0007] A defect identification report is generated based on the positioning, classification, and severity information.

[0008] Furthermore, the fan blade surface defect recognition model is implemented through the following steps:

[0009] Training:

[0010] Acquire multiple blade surface defect images, and annotate defect locations and categories in the multiple blade surface defect images to obtain a training image set;

[0011] The AdamW optimizer was selected and the YOLOv8 model was trained using the training image set to obtain a wind turbine blade surface defect recognition model.

[0012] Optionally, after marking the defect positions and categories in the plurality of blade surface defect images, the method further includes:

[0013] The multiple blade surface defect images are enhanced.

[0014] Optionally, the training of the YOLOv8 model includes:

[0015] Monitor the training process, track loss changes and mAP;

[0016] Save the model periodically and evaluate model performance.

[0017] Optionally, the training of the YOLOv8 model includes:

[0018] Tune the learning rate, batch size, and IOU threshold.

[0019] Furthermore, the outputting of the location, classification, and severity information of defects in the surface image of the wind turbine blade includes:

[0020] Outputting a defect bounding box; wherein the bounding box has a category probability distribution and a confidence score;

[0021] Determining the defect type corresponding to the bounding box according to the highest category probability corresponding to the bounding box;

[0022] Feature extraction is performed on the bounding box to determine the severity of the defect.

[0023] Optionally, after outputting the defect boundary box, the method further includes:

[0024] filtering the bounding box according to the confidence score and a preset confidence threshold; and / or,

[0025] The overlapping bounding boxes are eliminated using the NMS algorithm.

[0026] Another aspect of the present disclosure provides a system for identifying surface defects of a wind turbine blade, the system comprising:

[0027] An acquisition module, used for acquiring a surface image of a wind turbine blade;

[0028] A recognition module, used for inputting the wind turbine blade surface image into a pre-trained wind turbine blade surface defect recognition model, and outputting location, classification, and severity information of defects in the wind turbine blade surface image;

[0029] A reporting module is used to generate a defect identification report based on the location, classification, and severity information.

[0030] Another aspect of the present disclosure provides an electronic device, comprising:

[0031] at least one processor; and,

[0032] The memory communicatively connected to the at least one processor is used to store one or more programs, which, when executed by the at least one processor, enable the at least one processor to implement the above-mentioned method for identifying surface defects of wind turbine blades.

[0033] Another aspect of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, can implement the above-mentioned method for identifying surface defects of a wind turbine blade.

[0034] The present invention discloses a method, system, device and medium for identifying surface defects of a wind turbine blade based on YOLOv8 (You Only Look Once version 8).

[0035] 8) image recognition technology has achieved rapid and accurate identification of surface defects on wind turbine blades, improved detection efficiency and accuracy, reduced labor costs, and ensured the safe and stable operation of wind power equipment. The automated detection process significantly shortens the time for blade defect detection; the high-precision recognition capability of the YOLOv8 model can effectively avoid missed detections and false detections; it reduces reliance on manual inspections and reduces labor costs, while providing early warnings to reduce sudden maintenance needs; it can be regularly monitored to ensure timely detection of hidden dangers and avoid safety accidents caused by blade failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A schematic flow chart of a method for identifying surface defects of a wind turbine blade according to an embodiment of the present disclosure;

[0037] Figure 2 This is a schematic structural diagram of a wind turbine blade surface defect recognition system according to another embodiment of the present disclosure;

[0038] Figure 3 The figure is a schematic diagram of the structure of an electronic device according to another embodiment of the present disclosure. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present disclosure.

[0040] In addition, the described features, structures or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known methods, devices, implementations or operations are not shown or described in detail to avoid blurring the various aspects of the present disclosure.

[0041] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0042] It should be understood that although the terms first, second, third, etc. may be used in this disclosure to describe various components, these components should not be limited by these terms. These terms are used to distinguish one component from another component. Therefore, the first component discussed below can be referred to as the second component without departing from the teachings of the concepts of the present disclosure. As used in this disclosure, the term "and / or" includes any one of the associated listed items and all combinations of one or more.

[0043] Those skilled in the art will appreciate that the drawings are merely schematic diagrams of example embodiments, and the modules or processes in the drawings are not necessarily necessary for implementing the present disclosure, and therefore cannot be used to limit the protection scope of the present disclosure.

[0044] like Figure 1 As shown, an embodiment of the present disclosure provides a method for identifying surface defects of a wind turbine blade, the method comprising:

[0045] Step S1, obtaining a surface image of a wind turbine blade.

[0046] Specifically, the surface of the wind turbine blade is photographed in all directions by a drone or a high-definition camera installed on the tower to obtain a high-definition image. In order to optimize the image quality in the subsequent recognition process, the image can also be pre-processed, including denoising, correction, and enhancement.

[0047] Step S2: input the wind turbine blade surface image into a pre-trained wind turbine blade surface defect recognition model, and output location, classification, and severity information of defects in the wind turbine blade surface image.

[0048] Specifically, a large number of labeled wind turbine blade defect image datasets are used in advance to train the YOLOv8 model. The YOLOv8 model is an object detection and image segmentation model, known for its efficient object detection capabilities, and can simultaneously predict the location and category of the target. By adjusting the network structure and hyperparameters, the model's recognition accuracy for typical defects such as blade cracks, corrosion, and wear is optimized. The training process includes but is not limited to:

[0049] 1. Prepare data and collect a large amount of image data of surface defects of wind turbine blades, including but not limited to different types of defects such as cracks, corrosion, and wear, to ensure that the data covers various lighting conditions, viewing angles, and defect levels; use professional annotation tools to accurately annotate images, including the location (bounding box) and category labels of defects; apply image enhancement techniques (such as rotation, flipping, scaling, brightness adjustment, etc.) to increase data diversity, reduce the risk of overfitting, and improve model generalization capabilities;

[0050] 2. Make targeted adjustments to the YOLOv8 model structure. For the blade defect detection task, it may be necessary to adjust the depth or width of the network, as well as the feature fusion strategy of the Neck part; considering the characteristics of blade defects, it may be necessary to fine-tune the last few layers of the model or add a specific attention mechanism to improve the recognition ability of small cracks or special shape defects;

[0051] 3. Set training parameters, adjust the weight of YOLOv8's loss function according to task requirements, balance position error, classification error and confidence loss, and ensure that the model achieves the best balance between detection accuracy and recall rate; select optimizer and learning rate, commonly use AdamW optimizer, combined with cosine annealing or multi-stage learning rate decay strategy, fast decline in the early stage, and gentle adjustment in the later stage to accelerate convergence and avoid falling into local optimality too early; adjust the batch size according to available computing resources, usually between 2-16. The number of training iterations (epochs) depends on the amount of data and model complexity, and may range from hundreds to thousands;

[0052] 4. Configure the hardware and environment to ensure that the training environment has correctly configured CUDA and cuDNN to fully utilize the GPU to accelerate the training process; use a virtual environment to isolate project dependencies, and install the Ultralytics library and other necessary libraries, such as PyTorch, OpenCV, etc.; CUDA is a parallel computing platform and programming model, cuDNN is a GPU-accelerated deep neural network primitive library, and Ultralytics is a Python library created to perform advanced computer vision tasks;

[0053] 5. Manage the training process, use TensorBoard or other recorders to monitor the training process, track key indicators such as loss (loss factor) changes and mAP (average of AP in each category) for easy tuning; save the model regularly during training and evaluate the performance on the validation set. Use indicators such as COCO mAP, Recall, Precision, etc. to measure model performance; TensorBoard is a set of visualization tools that can help developers easily understand, debug, and optimize TensorFlow programs;

[0054] 6. Perform hyperparameter tuning, using grid search or random search to tune hyperparameters such as learning rate, batch size, IOU threshold, etc., and use automated tools such as Optuna or Ray Tune to find the optimal parameter combination;

[0055] 7. Test and evaluate the model. Perform comprehensive tests on an independent test set to ensure the generalization ability of the model and evaluate the accuracy and stability of the model in practical applications.

[0056] Before inputting the image into the trained model for prediction, the leaf image to be input can be normalized to ensure that the pixel value range meets the standard during model training (usually 0-1 range), which helps the model to stabilize the output. According to the input requirements of the model, adjust the image size to a fixed resolution. The YOLO series models prefer square input, such as 640x640 pixels.

[0057] First, the preprocessed image is input into the trained YOLOv8 model, which will output a series of bounding boxes, each with a category probability distribution and an object confidence score. Appropriate confidence thresholds and category probability thresholds are set to filter out low-confidence predictions and reduce false positives. The NMS algorithm is used to eliminate overlapping bounding boxes and retain the predictions that are most likely to represent real defects, further improving recognition accuracy and robustness.

[0058] Then, the type of defect (such as crack, corrosion, wear, etc.) is determined based on the highest class probability corresponding to each bounding box. This step is based on the class distinction ability learned during model training. If the model is designed for multi-class classification, ensure that appropriate post-processing logic is applied to each defect box, including but not limited to multi-label classification or single-label maximum probability classification.

[0059] Finally, the severity of the defects is evaluated. Through feature extraction, for each detected defect, its shape, texture, color and other features can be further extracted, which are related to the severity of the defect. Using predefined rules or training an auxiliary model, a severity score or grade classification is given to each defect based on the extracted features, such as minor, moderate, severe, etc.

[0060] After obtaining the above results, the results can also be visualized: the bounding box of the defect is superimposed on the original image, and the defect type and severity are annotated to generate a visual report that is easy to understand and analyze; a user-friendly interface is developed to allow operators to view the inspection results, zoom in to see details, adjust thresholds for re-analysis, or directly export the inspection report.

[0061] Finally, post-processing optimization can also be performed: in actual applications, the differences between the model prediction results and the manual review are continuously collected for model iterative optimization and adjustment of the post-processing logic, such as updating NMS parameters, adjusting classification thresholds, etc.; a performance monitoring system is established to track the performance of the model in different environments and time to ensure long-term stability and accuracy.

[0062] Step S3: Generate a defect identification report based on the positioning, classification, and severity information.

[0063] Specifically, the system organizes the identified defect information, including defect location, type, severity assessment, etc., and automatically generates a test report to provide a basis for maintenance decisions of wind turbines. According to the severity of the defect, the system automatically triggers an early warning signal and provides the maintenance team with immediate maintenance suggestions, including priority sorting, recommended repair measures, etc.

[0064] A method for identifying surface defects of wind turbine blades in an embodiment of the present disclosure realizes rapid and accurate identification of surface defects of wind turbine blades through image recognition technology based on YOLOv8 (YouOnly Look Once version 8), improves detection efficiency and accuracy, reduces labor costs, and ensures safe and stable operation of wind power equipment. The automated detection process significantly shortens the time for blade defect detection; the high-precision recognition capability of the YOLOv8 model can effectively avoid missed detection and false detection; it reduces reliance on manual inspections, reduces labor costs, and provides early warning to reduce sudden maintenance needs; it can monitor regularly to ensure timely detection of hidden dangers and avoid safety accidents caused by blade failures.

[0065] like Figure 2 As shown, another embodiment of the present disclosure provides a wind turbine blade surface defect recognition system, the system comprising:

[0066] An acquisition module 210 is used to acquire a surface image of a wind turbine blade;

[0067] The recognition module 220 is used to input the wind turbine blade surface image into a pre-trained wind turbine blade surface defect recognition model, and output the location, classification, and severity information of the defects in the wind turbine blade surface image;

[0068] The reporting module 230 is used to generate a defect identification report based on the location, classification, and severity information.

[0069] A wind turbine blade surface defect recognition system according to an embodiment of the present disclosure is used to implement the wind turbine blade surface defect recognition method described above. The specific implementation process has been described in detail above and will not be repeated here.

[0070] like Figure 3 As shown, another embodiment of the present disclosure provides an electronic device, including:

[0071] At least one processor 301; and a memory 302 in communication with the at least one processor 301, for storing one or more programs, which, when executed by the at least one processor 301, enable the at least one processor 301 to implement the wind turbine blade surface defect identification method described above.

[0072] The memory 302 and the processor 301 are connected in a bus manner, and the bus may include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors 301 and the memory 302 together. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be one element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. The data processed by the processor 301 is transmitted on a wireless medium through an antenna, and further, the antenna also receives data and transmits the data to the processor 301.

[0073] The processor 301 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management and other control functions. The memory 302 can be used to store data used by the processor 301 when performing operations.

[0074] Yet another embodiment of the present disclosure provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the above-mentioned method for identifying surface defects of a wind turbine blade.

[0075] The computer-readable storage medium may be included in the system or electronic device of the present disclosure, or may exist independently.

[0076] Computer-readable storage media may be any tangible media that contains or stores a program, which may be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device. More specific examples include, but are not limited to, an electrical connection with one or more wires, a portable computer disk, a hard disk, optical fiber, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0077] The computer-readable storage medium may also include a data signal propagated in baseband or as part of a carrier wave, in which the computer-readable program code is carried. Specific examples include but are not limited to electromagnetic signals, optical signals, or any suitable combination thereof.

[0078] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present disclosure, but the present disclosure is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and substance of the present disclosure, and these modifications and improvements are also considered to be within the scope of protection of the present disclosure.

Claims

1. A method for identifying surface defects of a wind turbine blade, characterized in that: The method comprises: Acquire the surface image of wind turbine blades; Inputting the wind turbine blade surface image into a pre-trained wind turbine blade surface defect recognition model, and outputting location, classification, and severity information of defects in the wind turbine blade surface image; A defect identification report is generated based on the positioning, classification, and severity information.

2. The method according to claim 1, characterized in that The wind turbine blade surface defect recognition model is trained through the following steps: Acquire multiple blade surface defect images, and annotate defect locations and categories in the multiple blade surface defect images to obtain a training image set; The AdamW optimizer was selected and the YOLOv8 model was trained using the training image set to obtain a wind turbine blade surface defect recognition model.

3. The method according to claim 2, characterized in that After marking the defect positions and categories in the plurality of blade surface defect images, the method further includes: The multiple blade surface defect images are enhanced.

4. The method according to claim 2, characterized in that: The training of the YOLOv8 model includes: Monitor the training process, track loss changes and mAP; Save the model periodically and evaluate model performance.

5. The method according to claim 2, characterized in that: The training of the YOLOv8 model includes: Tune the learning rate, batch size, and IOU threshold.

6. The method according to any one of claims 1 to 5, characterized in that: The outputting of the location, classification, and severity information of defects in the surface image of the wind turbine blade includes: Outputting a defect bounding box; wherein the bounding box has a category probability distribution and a confidence score; Determining the defect type corresponding to the bounding box according to the highest category probability corresponding to the bounding box; Feature extraction is performed on the bounding box to determine the severity of the defect.

7. The method according to claim 6, characterized in that After outputting the defect boundary box, the method further includes: filtering the bounding box according to the confidence score and a preset confidence threshold; and / or, The overlapping bounding boxes are eliminated using the NMS algorithm.

8. A wind turbine blade surface defect recognition system, characterized in that: The system comprises: An acquisition module, used for acquiring a surface image of a wind turbine blade; A recognition module, used for inputting the wind turbine blade surface image into a pre-trained wind turbine blade surface defect recognition model, and outputting location, classification, and severity information of defects in the wind turbine blade surface image; A reporting module is used to generate a defect identification report based on the location, classification, and severity information.

9. An electronic device, characterized in that: include: at least one processor; as well as, A memory communicatively connected to the at least one processor, for storing one or more programs, which, when executed by the at least one processor, enables the at least one processor to implement the method for identifying surface defects of a wind turbine blade as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it can implement the method for identifying surface defects of a wind turbine blade as described in any one of claims 1 to 7.

Images