In allusion to the issue of degradation feature extraction and degradation phase division, a logistic chaotic map is used to study the variation pattern of spectral entropy, and a technique based on Gath–Geva fuzzy clustering is proposed. The degradation features include spectral entropy, root mean square, and “curved time,” which are more in line with the performance degradation process than degradation time. Gath–Geva fuzzy clustering is introduced to divide different phases in the degradation process. The rolling bearing lifetime vibration signal from the intelligent maintenance systems (IMS) bearing test center was introduced for instance analysis. The results show that spectral entropy is able to effectively describe the complexity variation pattern in the performance degradation process and has some advantages in sensitivity and calculation speed. The introduced “curved time” is able to reflect the agglomeration character of the degradation condition on a time scale, which is more in line with the performance degradation pattern of mechanical equipment. Gath–Geva fuzzy clustering is able to divide the degradation phase of mechanical equipment such as bearings accurately.
Fault diagnosis of rolling bearing based on de-noising manifold learning algorithm
