The system be down for regular maintenance on April 3rd, 2024 from 8:00-10:00am.
Fault detection of rolling element bearings
Abstract
The early detection of bearing failure requires methods that are sensitive to impulsive signals and their changes. The difficulty of early bearing defect detection is that the signal component generated by early defect is very small compared to other vibration sources. It was found that techniques focus at signal changes in the higher frequency is more sensitive to early bearing defect because the periodic impact generated by early bearing defect will resonant at high frequency regions with less noise interference. The most popular technique is the high frequency resonant technique (HFRT) by demodulating the envelope of the bandpass signal using Hilbert transform. There are two major limitations of using the HFRT. First of all, the HFRT demands on knowledge of the resonance frequency range where the defect generated impulses is more pronounced with respect to normal system vibrations. The second limitation is when there are multiple defects developed at the same time, a traditional HFRT may not be able to identify each defects due to the overlap of each transient component.A resonant frequency band selection method based on the wavelet packet transformation is developed to search the optimal resonant packet automatically using kurtosis and changes of signal energy.Two new HFRT algorithms based on wavelet packet transformation and quadratic energy detector are developed. The results from the synthetic bearing defect signal shown that the performance of newly proposed HFRT algorithms are sensitive to early bearing defect and are able to detect multiple defects which cannot be detected by the traditional HFRT method based on Hilbert transform. A series of experiments using recorded signals from both the artificially defected bearings and bearings running from normal till failure are performed. It was found that both methods work well for field applications.When the signal to noise ration is low, we can pre-process the signal with harmonic analysis to extract the fault related impulse component to improve the performance of HFRT algorithms for bearing defect detection.
Collections
- Mechanical engineering [431]