Bearing Fault Diagnosis using Singular Spectrum Analysis-Based Envelope Detection

Abstract

Rolling element bearings are critical components in rotating machinery and it is important to monitor their health and detect their faults in early stage during their operations. The vibration energy generated by the faults in rolling element bearings is usually small comparing to that of other rotating components such as rotors/shafts and gears in mechanical systems. Envelope analysis is a widely used method in bearing fault diagnosis. The major challenge in envelope analysis is the identification of resonance frequency band for narrow-band filtering before applying envelope operation. In this paper, a method combining singular spectrum analysis (SSA) and envelope analysis is proposed for diagnosing the rolling element bearing faults. The SSA is utilized to decompose the bearing vibration signal into a set of eigentriples (principle components), and then a subset of the eigentriples that encompass the dominant variation in the original signal is selected for signal reconstruction. Envelope analysis is then applied to the reconstructed signal to extract the modulation information that caused by the bearing faults. The proposed SSA-based envelope analysis is applied to the data sets of the bearing data center at Case Western Reserve University (CWRU), and the results are compared with that of the widely used Kurtogram-based envelope analysis. The results show that the SSA-based envelope analysis is more effective than the Kurtogrambased envelope analysis in bearing fault diagnosis.