Innovative Methodology Utilizing Vibration Sensor Data for Early Detection and Prediction in Wind Turbine Systems

This research work presents an innovative approach for forecasting the state of high-speed bearings in wind turbines. The method employs raw data obtained from sensors that quantify vibrations. The procedure involves computing a Kurtosis Spectrogram for each of the 50 samples and employing a 1D-CNN to classify bearings into two categories: Medium-life Expectancy (over 15 days) or Short- life Expectancy (under 15 days). This study emphasizes the significance of early detection in predicting the Remaining- Useful Life (RUL) of machine components, which is crucial for preventing failures and optimizing maintenance decisions. The 1D-CNN, implemented using the Keras framework, exhibits impressive accuracy, reaching an overall accuracy of about 90% as measured by recall. This study highlights the potential use of this technology in alert systems, providing crucial support for technical decisions about component maintenance, ensuring the reliability and efficiency of wind turbine systems.