Wind Turbine Blade Positioning and Defect Recognition via SCNs and YOLOv5

The maintenance of blades is essential for the stable and safe operation of wind turbines to ensure power generation efficiency. However, the low positioning efficiency of wind turbine blades and the difficulty in feature extraction of blade surface defects make the detection and recognition of wind turbine blade defects challenging. Therefore, combined with artificial intelligence transfer learning, YOLOv5 is proposed to extract the blade area in a complex background, and then the fine-tuned ResNet18 is used to extract the feature of the blade area to obtain the feature vector with a length of 512. Finally, Stochastic Configuration Networks (SCNs) is used to classify the data obtained by dimensionality reduction. The experimental results show that the AP: 50 of YOLOv5 reaches 99.5%, and the defect classification accuracy of SCNs reaches 97.2%. This method can accurately and quickly locate the position of wind turbine blades and identify the type of blade defects.