Vibration Analysis for Fault Detection of Wind Turbine: New Methodology of Supervised Machine Learning Techniques

Abstract

The implementation of supervised machine learning techniques classifiers is changing wind turbine maintenance. This automatic and autonomous learning methodology allows one to predict, detect, and anticipate the degeneration of any electrical and mechanical components present in a wind turbine. In this paper, two different failure states are simulated due to bearing vibrations, comparing frequency analysis and some machine learning classifiers. With the implementation of the KNN and SVM algorithms, we can evaluate different methodologies for supervision, monitoring, and fault diagnosis in a wind turbine. With the implementation of these techniques, it reduces downtime, anticipates potential breakdowns, and aspect import if they are offshore.