A New Health Indicator Based on the Energy and Distribution of Frequency Domain Fault Information for Rolling Bearing Condition Monitoring

Health condition monitoring of rolling bearings is of significance to ensure the safety and reliability of modern industry, which depends on a robust health indicator (HI). However, most existing HIs still struggle with accurate incipient fault detection and monotonic degradation assessment. To address this issue, a new HI based on the energy and distribution of frequency domain fault information (EDFF-HI) is proposed. Firstly, the improved clustering envelope spectrum (ICES) that developed from the recently proposed clustering weighted envelope spectrum (CWES) is used to identify the frequency-domain fault information. Through the ICES algorithm, the informative spectral frequency slices (SFSs) in the spectral coherence (SCoh) can be identified, and the potential fault characteristic frequencies (PFCFs) of each SFS can also be determined, which is also called its label. Secondly, the signal-noise ratio of each informative SFS is calculated according to its label to reflect the fault energy. Meanwhile, the sum of the signal-noise ratio of all informative SFSs is calculated to construct the HI, which considers the fault information distribution. To further enhance the performance of HI, the square root operation and moving average algorithm are introduced to obtain the final HI. This method only requires a rough estimate of FCFs rather than accurate values. Two bearing run-to-failure datasets are used to demonstrate the effectiveness of the proposed HI. The results indicate that the proposed EDFF-HI can detect incipient faults earlier, and the incipient fault sensitivity (IFS) index is usually more than 1.5 times that of other popular HIs.