A novel rolling bearing fault diagnosis method based on continuous hierarchical fractional range entropy

Abstract

Recognizing rolling bearing faults heavily relies on effective feature extraction methods. In practical engineering, the collected vibration data contains significant levels of noise and amplitude fluctuations, impeding the successful extraction of fault features. To address these challenges, a fault extraction method known as continuous hierarchical fractional range entropy (CHFRE) is proposed. First, range entropy (RE) is introduced and derived to the fractional domain to quantify the fault characteristics. This can overcome the effects of high noise and amplitude fluctuations. Second, the fault features hidden in the low- and high-frequency components are extracted in the multiscale domain by improved hierarchical analysis. Finally, a model for rolling bearing faults identification is established. This paper explores the CHFRE parameters, and a crucial finding is presented: the fractional order contains several invalid intervals within its defined range. The validity of the method and the existence of the invalid intervals are verified by two experiments.