Owie-Driven spectrum discrepancy impulse reconstruction: An adaptive impulse extraction methodology for rotating machinery

Rotating machinery faults generally result in the generation of impulse components, accompanied by multi-order harmonics. However, the impulse phenomena may inherently be submerged within noisy stationary components, which introduces challenges to identify machinery faults. Existing methods for impulse extraction are mainly divided into two categories: signal decomposition and deconvolution-based methods. The former fundamentally rely on single band-pass filtering, making them incapable of extracting impulse components distributed across multiple frequency bands. The latter functions as frequency-shaping filters, which are prone to cause waveform damage to impulse components. To overcome these limitations, this paper proposes an adaptively impulse extraction methodology for rotating machinery, defined as spectrum discrepancy impulse reconstruction (SDIR). Its innovation lies in the accurate differentiation of each frequency corresponding to impulses, stationary components or noise based on spectrum discrepancies for subsequent signal reconstruction. The spectrum of impulses is determined by comparing the original spectrum with the spectrum after highlighting impulses using optimal weight impulse extraction (OWIE). Additionally, noise is split out through a threshold determined according to the Rayleigh distribution of noise spectrum amplitude. The effectiveness of the proposed SDIR is validated through simulation case studies and experimental fault signals. Results demonstrate that the SDIR is able to decompose the raw signal into three intrinsic sub-signals adaptively for impulses extraction with considerable accuracy.