AI‐Driven Deep Learning Framework for Detecting Subtle Surface Defects on Wind Turbine Blades

ABSTRACT Wind turbine blade surface defect detection is of great significance in ensuring the safety and operational efficiency of wind power systems. However, accurately detecting subtle and small‐scale defects remains challenging under complex imaging conditions. In particular, progressive downsampling in deep layers tends to cause severe feature loss for subtle and small‐scale defects, whereas insufficient cross‐layer feature compensation further weakens spatial detail preservation and multiscale discrimination. To tackle these challenges, this paper introduces an enhanced defect detection algorithm based on You Only Look Once version 8 nano (YOLOv8n). The main improvements include (1) introducing the C2f_EMBC module in the backbone network to replace the original C2f module to enhance the feature extraction capability; (2) adopting the Gold‐YOLO architecture in the neck network and utilizing the innovative gather‐and‐distribute (GD) mechanism to optimize the multiscale feature fusion through low‐GD and high‐GD branches. Compared with the original YOLOv8n algorithm, the proposed method improves the recall rate by 6.7% on the wind turbine defect dataset and increases the mean Average Precision at Intersection over Union (IoU) threshold 0.5 ([email protected]) by 2.2%. The accompanying efficiency metrics, including inference speed in frames per second (FPS), the number of parameters (Params), and computational complexity in giga floating‐point operations (GFLOPs), further suggest that these gains are achieved with acceptable computational overhead, indicating suitability for real‐time UAV‐based inspection and potential deployment on edge platforms.