Weak Fault Feature Extraction of Rotating Machinery Based on Double-Window Spectrum Fusion Enhancement

Early weak fault feature extraction is difficult due to the interference by wideband random noise and other vibration sources in the same frequency band. In this paper, a weak fault feature extraction method of rotating machinery based on the double-window spectrum fusion enhancement (DWSFE) is proposed to address this issue. This method is established through combining fast Kurtogram algorithm and spectrum fusion enhancement. In this scheme, the fast Kurtogram algorithm is utilized to suppress the interference of weak fault feature extraction by noise signals outside the optimal resonant band and to achieve the signal envelope demodulation in the passband of the optimal resonant region. Subsequently, discrete short-time Fourier transforms (DSTFTs) with long- and short-time windows are constructed to obtain the time & x2013;frequency spectra with different time & x2013;frequency resolutions, respectively. By integrating the advantages of the time & x2013;frequency resolutions of these spectra, the spectrum correlation analysis denoising and the spectrum amplitude association enhancement are proposed to suppress the random interference in the passband and to highlight the signal characteristic frequency. Furthermore, the weak fault features are separated from other interference of vibration sources in the same frequency band, and finally, the weak fault feature extraction of rotating machinery can be achieved. The simulation and application results show that this method can enhance weak fault features while denoising and achieve the weak fault feature extraction of rotating machinery under strong noise.