Bearing performance degradation evaluation using recurrence quantification analysis and auto-regression model

This paper presents an integrated approach which combines recurrence quantification analysis (RQA) with the auto-regression (AR) model, for evaluating bearing performance degradation, including degradation monitoring and state prediction. RQA is applied to extracting recurrence plot (RP) entropy feature from vibration signals for both monitoring and predicting the bearing degradation through an AR model. The experimental results indicate that the RP entropy can be used as an effective indictor for bearing degradation monitoring. Furthermore, the AR model built upon the RP entropy can predict the bearing failure one hour in advance.