Reconstruction-based Fault Prognosis for Bearings with Principal Component Analysis

Bearings play an important role in mechanical systems, and it is of great significance to monitor the health status of the bearing in real time for ensuring long-term stable operation of mechanical system. This paper proposes a new reconstruction-based fault prognosis method to process the time domain and frequency domain characteristics of vibration signals. Firstly, the principal component analysis (PCA) model is established to detect the fault factors in current measurement data. And then, the fault directions are selected according to its fault correlation and the robustness of the corresponding fault magnitude evolution process. Finally, the vector autocorrelation (VAR) model is used to predict the fault degradation process to achieve complete fault prognosis. The effectiveness of the developed method is validated by PRONOSTIA data.