[3A1] Computer vision and deep learning for structural health monitoring of wind turbines
M Shafiee
University of Surrey, UK
The offshore wind energy infrastructure constitutes the backbone of modern societies and, hence, their security is of utmost importance. The use of computer vision (CV) and deep learning (DL) technology in the automated analysis of condition monitoring (CM) data for wind turbines inspection and maintenance has attracted considerable attention in recent years. The CV and DL methods offer significant efficiency and comprehensive advantages over human operators in the CM domain, such as improved accuracy, faster data interpretation and automatic reporting of results to clients. Several CV and DL techniques, such as convolutional neural networks (CNNs), YOLO damage detection and ResNet50 image classification, have been applied to structural health monitoring (SHM) systems. The aim of this paper is to provide a state-of-the-art review of CV and DL techniques and their applications to fault detection in materials and structural components of wind turbines. We explore various types of data yielded by CM techniques from wind turbines and then provide a brief overview of the data-driven techniques that can be used to extract damage-sensitive features. A classification of the literature that has reported the deployment of CV and DL technology for the automatic acquisition and interpretation of large-volume CM data from various sources is proposed. This paper also provides insights into overcoming the typical challenges associated with implementing CV and DL technology in the inspection and monitoring of wind turbines for further research directions.
Keywords: condition monitoring (CM), wind turbines (WTs), structural health monitoring (SHM), computer vision (CV), deep learning (DL).
Keywords: condition monitoring (CM), wind turbines (WTs), structural health monitoring (SHM), computer vision (CV), deep learning (DL).
