Study on crack monitoring method of wind turbine blade based on AI model: Integration of classification, detection, segmentation and fault level evaluation

The stability of wind turbines is closely related to the economic benefits of wind energy. To improve the stability of the wind turbine, a comprehensive image diagnosis method based on artificial intelligence method called 'Multivariate Information Perception You Look Only Once' (MIP-YOLO) is proposed. MIP-YOLO is an improved algorithm based on YOLOv8 that can classify, detect, segment and evaluate the crack damage level, and can be utilized for monitoring surface cracks on wind turbine blades. To improve the detection capability of small, relatively weak targets such as cracks, Multivariate Information Perception and C2TR modules are put forward. Aim at enhancing the ability of extracting objects with edge features, the Haar wavelet attention (HWA) and C2fGhost modules are proposed. In order to make the model extract features better, C2CBAM module is put forward in this paper. For purpose of solving the problem that some samples in the dataset may have poor quality, wise-IOU is introduced into the model. The detection performance of the proposed method is tested using wind turbine's blade images with cracks taken by Unmanned Aerial Vehicles (UAV). The experiment shows that MIP-YOLO can realize the fault diagnosis of blade effectively and improve the economic benefit of wind energy.