Early and effective detection of wind turbine blade (WTB) surface defects can avoid complex and expensive repair problems and serious safety hazards. The traditional target detection methods have the problems of insufficient detection capability, long model inference time and low recognition accuracy for small targets and long strip defects in WTB datasets. This paper proposes a high‐precision model SOD‐YOLO for WTB surface defect detection based on UAVs image analysis of YOLOv5. First, the WTB images are preprocessed by foreground segmentation and Hough transform to build the WTB defect dataset. Then, a micro‐scale detection layer is added to the original YOLOv5, and the K‐means algorithm is used to re‐cluster anchors and add the CBAM attention mechanism to each feature fusion layer to reduce the loss of feature information for small target defects and other defects. In addition, to improve the detection efficiency, the channel pruning algorithm is used to reduce the model size. The experimental results show that the average accuracy (mAP) of the SOD‐YOLO algorithm on the WTB dataset reaches 95.1%, which is 7.82% better than YOLOv5, and the FPS is 28.3% better. Therefore, SOD‐YOLO is able to detect small target defects and other defects quickly and effectively.

In order to realize the automatic detection of surface defects of lithium battery pole piece, a method for detection and identification of surface defects of lithium battery pole piece based on multifeature fusion and PSO-SVM was proposed in this paper. Firstly, image subtraction and contrast adjustment were used to preprocess the defect image to weaken the influence of non-defective areas and enhance the defect features. Then, Canny algorithm and the AND logical operation were used to extract the image of defect area. Next, the texture feature, edge feature, and HOG feature were combined to extract the feature of the defect area image. Finally, the support vector machine (SVM) optimized by particle swarm optimization (PSO) was used to automatically identify and classify defect images. The experimental results show that the proposed method in this paper can effectively detect surface multiple types defects of lithium battery pole piece, and the average recognition rate of defects reaches 98.3%, which is an effective and feasible automatic defect detection and identification method.

As a key component of wind turbines (WTs), the blade conditions are related to the WT normal operation and the WT blade inspection is a significant task. Most studies of WT blade inspection focus attention on acquired sensor signal processing; however, there exist problems of stability, sensor installation, and data storage and processing. Onsite visual surface inspection is still the most common inspection method, but it is inefficient and requires a long downtime. Aimed at solving the above issues, a novel blade inspection method based on deep learning and unmanned aerial vehicles is proposed. Since common defect types are visible, the inspection problem is regarded as an image recognition problem. Three convolutional neural networks are trained by using the constructed dataset for image recognition, and the F1-score is applied to evaluate the models. The VGG-11 model is chosen for the final model due to its best performance. Then, the alternating direction method of multipliers algorithm is employed to compress the model to reduce the requirements on hardware devices. The blind area of the WT can be reduced, the efficiency of subsequent maintenance can be improved, maintenance costs can be reduced, and the economic performance can be increased. Finally, a comparison experiment of different inspection methods is carried out to demonstrate the proposed advantages.

As a key part of the wind turbines (WTs), the blade has a direct influence on the efficiency of WT. Because the defect detection technology of WT blade is not widely used, and the robustness of traditional detection methods is poor, this paper proposes a multi-feature fusion residual network combined with transfer learning. In this paper, the WT blade image dataset is enhanced and constructed to train the convolutional network. Two residual structures of multi-feature fusion (two feature fusion and three feature fusion) are proposed and compared. At the same time, transfer learning is used to improve training process and accelerate convergence. Compared with several convolutional neural networks based on indices include training loss and testing accuracy, f1-score and confusion matrix, the method proposed greatly reduces the time while achieving accurate detection. K E Y W O R D S defect detection, multi-feature fusion, residual network, transfer learning, wind turbine blade How to cite this article: Zhu J, Wen C, Liu J. Defect identification of wind turbine blade based on multi-feature fusion residual network and transfer learning.

No abstract