Early Fault Feature Extraction Based on Improved 2.5-Dimensional Envelope Spectrum
Fengqi Zhou, Fengxing Zhou, Yan Baokang · 2023
Owing to the small transient pulses in the early fault signal, particularly in the presence of strong background noise that interferes with the detection of shocks, the efficient extraction of the fault characteristics of rolling bearings is challenging. To facilitate better extraction of the bearing fault information at early stages, this study proposes a modified 2.5-dimensional (2.5D) envelope spectral fault feature extraction method. As higher-order cumulants are still affected by noise in practice, this study improves the 2.5D cumulants to better eliminate Gaussian noise. Subsequently, the improved 2.5D spectrum is combined with a spectrum envelope to extract the characteristics of the bearing fault signals, and the information entropy is introduced to evaluate the parameter values. Experimental data show that the improved 2.5D envelope spectrum better extract the bearing fault information than the 1.5-dimensional and 2.5D spectral methods. Thus, the proposed method offers certain advantages in terms of fault signal extraction