The study on fault diagnosis of key components in wind turbine

Abstract
This paper studies on monitoring vibrational signal of bearing and gearbox in wind turbine to diagnose its condition, and build up the relationship between fault and spectrum by using experimental data.
To simulate three kinds of failure conditions of gears, including imbalance gear, tooth breakage and unparallel shaft, the gear-rotor system is built up for the gear fault experiment. Fast Fourier Transform will transform time domain signal into spectrum which is frequency domain signal, and extract features of spectrum as the basis of diagnosis.
After features extraction, K-means algorithm and Bayesian Network are used to analyze features of spectrum. It is shown that Bayesian Network has higher precision as compared with K-means algorithm, and the average precision of Bayesian Network is up to 90 percent and above.