DAMAGE LOCALIZATION ON WIND TURBINE BLADES USING ACOUSTIC EMISSION (AE) SIGNALS AND GRAPH NEURAL NETWORK (GNN)

Accurately localizing the source of early structural damage on wind turbine blades poses a challenge for constructing acoustic emission (AE) based structural health monitoring (SHM) systems. Existing damage localization methods struggle to extract and utilize the intricate connections inherent in the AE signals. A novel method for localizing structural damage zone on wind turbine blades based on AE and graph neural networks (GNN) is proposed in this paper. First, the AE signals are converted into graph structure data in non-Euclidean space combined with time-frequency analysis. Then, the Euclidean distance between the features of each pair of nodes is calculated to determine the connectivity of the graph. The information of the neighboring nodes in the graph is aggregated using the message passing paradigm, which can not only make effective use of the node features, but also excavate the deeper intrinsic connection of the AE signals. The proposed method can easily localize structural damage on wind turbine blades with the assistance of the established graph. Another novelty of the proposed method is its ability to locate damage for wind turbine blade with high accuracy using only one AE sensor. This not only simplifies the sensor setup but also leads to significant cost reduction. The effectiveness of the proposed method is validated using an experimental dataset collected from a segment of a wind turbine blade. The results demonstrate the superior performance and high accuracy of the proposed method.