IESMGCFFOgram: A new method for multicomponent vibration signal demodulation and rolling bearing fault diagnosis

Abstract

Resonance demodulation of vibration signals is an essential strategy for rolling bearing fault diagnosis. However, most resonance demodulation methods focus on searching for only one optimal informative frequency band, which is only valid for single-component signal analysis. In industrial applications, many interfering components are included in the monitored signal because of the complex equipment composition and running environment. Therefore, multicomponent vibration signal demodulation methods are more practical for bearing fault diagnosis. Although decomposition methods are commonly used to extract multiple components from a mixed signal, they are not accurate enough because they are implemented based on empirical statistical indicators rather than fault-related information with physical meaning. To achieve accurate multicomponent vibration signal demodulation, the improved envelope spectrums via multigroup candidate fault frequency optimization-gram (IESMGCFFOgram) is proposed. Firstly, the candidate fault frequencies (CFFs) are identified based on the local maxima distribution of the Spectral coherence (SCoh). Secondly, multigroup CFFs (MGCFFs) are extracted from the original CFFs according to the multiple relationship. Thirdly, the MGCFFs are optimized by redundant group elimination, significance ranking, and homologous group fusion. Finally, several IESMGCFFOgrams and corresponding improved envelope spectrums (IESs) are respectively obtained by using different diagnostic indicators and the classical 1/3-binary tree filter structure, where the diagnostic indicators are adaptively designed based on each group of CFFs in the final MGCFFs. It is worth noting that the noise benchmark of all IESs will be unified by subtracting the minimum before constructing each IESMGCFFOgram. The effectiveness and superiority of the proposed method are validated by simulated and experimental data. Results show that the proposed IESMGCFFOgram has higher demodulation accuracy and computational efficiency compared with three state-of-the-art methods.