Detection and recovery of fault impulses via improved harmonic product spectrum and its application in defect size estimation of train bearings

Abstract

Abstract The detection and recovery of impulsive signature play a vital role in the diagnosis and prognosis of rolling element bearings. Though different approaches have been proposed to deal with this problem so far, challenges still exist when they are applied to the bearings operating under harsh working conditions. The difficulties mainly come from the multi-resonance and multi-modulation characteristics of bearing vibration signals. To overcome this limitation, a new methodology for the detection and recovery of fault impulses is presented in this paper. First, an improved harmonic product spectrum (IHPS) is proposed to detect and identify the multiple modulation sources buried in a vibration signal. With this method, the fault-related impulsive features could be recognized, while the influence caused by non-fault modulation is eliminated. On this basis, a harmonic significance index is further established to quantify the diagnostic information contained in a narrow band signal. By utilizing this index, the optimal resonance band where the fault impulses are most significant could be accurately determined. Finally, IHPS and sideband product spectrum are integrated to reduce the in-band noise and further recover the fault impulses. The performance of this method is evaluated by both simulated data and real vibration data measured from a train wheel bearing with a naturally developed defect. Compared with Kurtogram and Protrugram, the proposed method can detect the resonance band more precisely even in the presence of heavy noise and other impulsive vibration sources. Moreover, with the impulses recovery scheme, the double impact phenomenon caused by a distributed defect is extracted successfully. Benefiting from this, the defect size of a bearing can be estimated from its vibration signal without dismantling, which makes it a promising tool for the bearing diagnosis and prognosis in industrial applications.