A neural network approach for prediction of bearing performance degradation tendency

Accurate bearing performance degradation indicator and prediction of bearing performance degradation tendency is the foundation of bearing remaining useful life prediction. Artificial neural network (ANN) methods have been widely used in prediction. However, the input of most ANN methods only utilize one condition monitoring data. Since the complexity of the bearing failure situation, we cannot measure the bearing performance degradation only by one indicator. In this paper, 35 performance degradation indicators are established in time domain, frequency domain and time - frequency domain, respectively. And a comprehensive characterization indicator is obtained by PCA method. ANN methods are used to predict the comprehensive characterization indicator which makes a good foundation for bearing remaining useful life prediction.