Research on a wind turbine blade surface defect detection method based on an improved YOLO11 model

Wind turbine blades are continuously exposed to environmental factors such as wind-blown sand, humidity, heat, and ultraviolet radiation. As service time increases, surface defects—including cracks, stains, and material peeling—gradually emerge. If not detected at an early stage, these defects may worsen over time and compromise the safety and operational stability of the turbine. In current operation and maintenance practices, blade inspection still relies heavily on manual methods, which not only suffer from limited efficiency but also involve risks associated with high-altitude work. Moreover, inspection results are often influenced by the experience of individual inspectors. In response to these practical challenges, this study focuses on the identification of surface defects on wind turbine blades and introduces targeted improvements within a deep learning–based detection framework. A single-stage object detection approach is adopted for model construction, with two key enhancements. First, a DySample dynamic upsampling module is incorporated into the feature fusion process to improve information transfer across features of different scales. Second, the FocalUIoU loss function is introduced in the bounding box regression stage to enhance the model's localization capability for small-scale defects and to improve optimization during training. Experimental results show that the improved model achieves a Precision of 0.874, a Recall of 0.793, and mAP@50 and mAP@50–95 of 0.856 and 0.544, respectively, on the test set. Compared with the baseline YOLO11 model, Recall is increased by approximately 15.6%, while mAP@50 and mAP@50–95 are improved by 6.5% and 7.9%, respectively. The model contains only 2.60 million parameters, indicating minimal change in overall scale while retaining the characteristics of a lightweight architecture. Under the premise of acceptable detection efficiency, the proposed method improves the recognition performance for multiple categories of blade defects and provides technical support for wind power equipment inspection and subsequent maintenance.