Cyclic Impulse Spectrum Analysis of Time-Frequency Slices for Axle-Box Bearing Fault Identification

For the incipient fault weakness, irregular pulse disturbance, and pseudo-cyclic periodicity of localized damage impact in the fault diagnosis of axle-box bearing (ABB) under complex service environments, the cyclic impulse spectrum (CIS) based on time-frequency spectrogram slicing is proposed to discriminate repetitive transients under noise contamination. In this approach, a new indicator called the cyclic impulse degree (CID) is designed based on the coefficient of variation of the impulse peak moment within a specified time-shifted periodic window, which allows for a generalization of impulsiveness and pseudo-cyclic periodicity to quantify the characteristics of bearing fault impacts. First, the normalized S transform (NST) highlights the energy distribution of fault impacts in the time-frequency plane, from which the CID is constructed to distinguish the cyclic impulses from random transients. Secondly, the NST slice at the frequency associated with the maximum integrated CID is used as a spare representation of fault impact response. Finally, the CIS derived from the NST slice presents the fault characteristic frequencies clearly and accurately. The numerical simulation and data processing results of the ABB faults demonstrated that the CIS can identify the multiple faults of ABB, but also exhibits good performance in resisting random transient interference and discriminating noise-contaminated periodic impulses caused by the bearing localized defects.