Remote Health Monitoring of Wind Turbines Employing Vibroacoustic Transducers and Autoencoders

Implementation of remote monitoring technology for real wind turbine structures designed to detect potential sources of failure is described. An innovative multi-axis contactless acoustic sensor measuring acoustic intensity as well as previously known accelerometers were used for this purpose. Signal processing methods were proposed, including feature extraction and data analysis. Two strategies were examined: Mel Frequency Cepstral Coefficients pruned with principal component analysis and autoencoder-based feature extraction. The scientific experiment resulted in data gathering and analysis to predict potential wind turbine mechanism failures.