A hierarchical adaptive selection neural network for dynamic impact localization of wind turbine blades

Abstract

Abstract This paper focuses on the localization problem of dynamic impacts that can lead to significant damages on wind turbine blades (WTBs). Localization of dynamic impacts on WTBs is essential for wind turbines due to their vulnerability to dynamic impacts such as birds, stones, hails. The proposed deep learning methodology contributes to accurately locate the impacted blade and specific position using the measurements from a limited number of sensors. In particular, a novel hierarchical adaptive selection neural network is proposed, which integrates a classification subnetwork and a regression subnetwork. Specifically, an adaptive blade selection mechanism is designed to determine the impacted blade for classification while an adaptive window selection mechanism is developed to highlight the representative time period for regression. By deploying a limited number of sensors to acquire measured vibration data, the proposed method can accurately identify the collision locations of transient impacts loaded on WTBs. In both simulated and real-world experiments, the proposed method achieves the mean absolute error of 0.189 centimeter and 1.088 centimeter for impact localization. The experimental results demonstrate the superior performance of the proposed model in comparison with the existing methods for localizing impulsive loads on WTBs.