Rolling bearing fault detection by short-time statistical features

Rolling element bearing fault diagnosis is a very important part of condition-based maintenance. In this article, a new method for detection of rolling element bearing defects is proposed. The method is based on the concept of the cyclostationarity of the vibration signal to find periodicity in statistical features of the vibration signal. Various statistical features are examined to find the best choice. Several case studies including inner race, outer race, and rolling element defects are investigated in this article. Comparison with the envelope analysis showed that the proposed method benefits from clearer defect frequency identification.