Abstract: In addressing the challenge of extracting fault characteristic frequencies from non-stationary vibration signals in rolling bearings, a fault diagnosis method of rolling bearing was proposed based on Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) and Hilbert envelope spectrum method. First of all, the ICEEMDAN was used to decompose the fault signals into a series of Intrinsic Mode Function (IMF). Subsequently, the valid IMFs with high correlation to the original signal were extracted, and their corresponding Hilbert envelope spectrum was generated to extract the fault characteristics frequency of the rolling bearing. Through signal analysis of the outer and inner ring faults of rolling bearings and comparison with Empirical Mode Decomposition (EMD), it is demonstrated that the amplitude of the envelope spectrum after the decomposition based on the ICEEMDAN method can clearly reflect the characteristic frequency of bearing faults. The results affirm the effectiveness of this method in diagnosing the outer and inner ring faults of rolling bearings.

Key words: rolling bearing, fault diagnosis, ICEEMDAN, Hilbert envelope spectrum