Method for enhancing the amplitude of fault features in wind power equipment via complex cepstrum-assisted optimization

Abstract

Precise diagnosis of rolling bearing health conditions is essential for reducing the maintenance costs of rotating machinery. To overcome the limitations of conventional spectrum segmentation methods, which rely on fixed filter banks and struggle to adapt to complex modulation characteristics, this paper proposes a cepstrum amplitude modulation (CSAM) method based on the complex cepstrum. The proposed approach first collects and preprocesses the raw vibration signal, and then obtains a complex spectrum via fast Fourier transform. A pointwise logarithmic sampling of this spectrum yields a complex logarithmic spectrum that preserves both magnitude and phase information. Power-exponent modulation is subsequently applied to the magnitude spectrum, and the modulated magnitude spectrum, combined with the original phase spectrum, is transformed back via inverse Fourier transform to generate the complex cepstrum. In the complex cepstrum domain, the harmonic spectral kurtosis is employed as a diagnostic metric to identify quefrency components associated with fault-induced modulation. Both numerical simulations and experimental validations demonstrate that the CSAM method can reliably extract periodic impulse signatures from inner-race and outer-race faults in rolling bearings. Compared with traditional methods, CSAM offers superior modulation-demodulation accuracy and noise immunity, providing an effective and adaptive demodulation tool for the health monitoring of rotating machinery.