Locating and Extracting Wind Turbine Blade Cracks Using Haar-like Features and Clustering

Abstract

Wind turbine blades can sustain damage during operation that can jeopardize the reliability of the entire wind power generator. This damage can be difficult to detect using conventional methods and, if unaddressed, could eventually result in the failure of the wind turbine. In this paper, a method of detecting wind turbine blade cracks from images is investigated which utilizes Haar-like features to locate cracks and the Jaya K-Means algorithm to extract the image pixels containing cracks. A modified turbine blade crack detection methodology based on existing technology is presented and coded in Python. Initial results for a small-scale wind turbine prototype with faulty blades at Utah Valley University look promising. Finally, a direction for continuing this undergraduate research project is put forth.