ARTIFICIAL INTELLIGENCE IN PREDICTION OF THE REMAINING USEFUL LIFE OF WIND TURBINE SHAFT BEARINGS

Abstract

Long-term periodic rotation and unstable load changes in wind turbines can cause unexpected damage to high-speed shaft bearings (HSSBs). In this study, after preprocessing of the HSSB vibration signal, four different models for predicting bearing degradation in terms of remaining useful life (RUL) in days were investigated: support vector regression (SVR), convolutional neural networks (CNN), long short-term memory (LSTM), and CNN-LSTM. The experimental results revealed that the CNN achieved the best mean absolute error (MAE), at 0.44 days, based on frequency response plot using the fast Fourier transform (FFT), while that of the CNN-LSTM model predicted using the amplitude profile in frequency response was 1.24 days. Meanwhile, the MAE of the SVR that extracted a total of 15 features for prediction was 2.31 days, while that of the LSTM predicted with the original time-domain data was 14.93 days, which was the worst. The experimental results demonstrated that, compared to the traditional time-domain vibration characteristics, the FFT-based method can predict the degradation trend of HSSBs more accurately.