Improved Energy Spectrum via Spectral Correntropy-Based Coherence-Gram for Bearing Fault Diagnosis

Abstract

The key steps of gram methods for bearing fault diagnosis are using an indicator to identify the optimal frequency band for demodulation and obtaining the signal’s squared envelope spectrum (SES). Using cyclostationary-based indicators for optimal band determination is one of the most widely used methods. However, the effectiveness of these methods may be compromised when the signal is non-Gaussian distributed. Compared with cyclostationary analysis, correntropy has the potential capability to detect cyclostationarity under non-Gaussian conditions. Therefore, this study proposes a novel tool called the spectral correntropy-based coherence gram (IEESCCgram) as an optimal band determination and fault diagnosis method. First, a novel bi-frequency map, the spectral correntropy-based coherence (SCECoh), is proposed, which is more robust to impulsive noise and can reveal hidden cyclic periodicities. Then, by represented in a 1/3 binary tree, the IEESCCgram uses the proposed cyclic band spectral negentropy (SNc) to select the optimal band. Finally, the improved energy spectrum (IEES) is obtained for bearing fault diagnosis. Vibration signals collected from experimental test rigs and real-life applications are used to verify the effectiveness and robustness of the IEESCCgram for bearing fault diagnosis when the signal is contaminated by impulsive noise and cyclostationary interference.