deep learning, LSTM autoencoders, fault monitoring, vibration, one class classification, anomaly detection
Abstract
In the last few years, as energy demand has grown at the same rate as global environmental concerns, there has been rapid growth in both the quantity and size of wind turbines. However, ensuring reliable performance and safety remains a challenge, especially in the blades, where failure can be both catastrophic and difficult to detect through traditional inspection methods. Therefore, this paper proposes a Structural Health Monitoring (SHM) method to detect anomalies based on health data deviations from the LSTM autoencoder, which includes two LSTM layers, data reconstruction, and accelerometer data, resulting in 97.4% balanced accuracy.
