Date of Publication :17th March 2024

Abstract:Wind turbines can become damaged during operation, and wind turbine blades are especially susceptible. Machine learning algorithms are often used to classify images, and these images are commonly processed prior to their use. One of these preprocessing methods is edge detection, which isolates the areas of an image that contain high-frequency information, such as edges. This paper explores the effect of applying edge detection as a preprocessing method for machine learning algorithms trained to classify defects in images of wind turbine blades. Specifically, edge detection is applied to the Xception and VGG19 convolutional neural networks. Conclusions as to the efficacy of edge detection as a preprocessing method for this type of data are drawn by comparing the classical performance of the selected machine learning algorithms to the performance of the same algorithms after implementing edge detection. If found to be successful, this technique can be used to improve automated detection of faulty wind turbines, which has implications for the reduction of energy and revenue loss at wind farms due to wind turbine downtime.

Reference :