Mehpsogram: a robust adaptive approach for demodulation frequency band identification in rolling bearing diagnostics under severe interference

Abstract

Envelope demodulation remains the most widely used technique for diagnosing rolling bearing faults, but its performance depends on the demodulation frequency band (DFB) identified. Despite extensive developments in DFB identification approaches, their performance often degrades under severe interference conditions due to the inherent methodological limitations. To address these deficiencies, this research introduces a novel DFB identification approach called Mehpsogram. Initially, a new pyramid-structure-based multi-scale adaptive spectral segmentation method is proposed to segment the spectrum of the bearing vibration signal into discrete subbands. Subsequently, a modified envelope harmonic product spectrum (MEHPS) is presented to suppress the interference frequency components while highlighting the spectral peaks generated by the fault characteristic frequency and its harmonics in the discretized squared envelope spectrum of the filtered subband signals. Finally, the optimal DFB is determined through frequency domain signal-to-noise ratio analysis of the MEHPSs derived from the filtered subband signals. Comprehensive evaluations on simulated and real-world datasets consistently demonstrate that the proposed Mehpsogram method outperforms existing representative DFB identification methods.