The Study On Detection Of Weak Fault Signal Of Ball Bearing

Ball bearing, which is much widely used in rotating machinery, always decides the whole machine performance. Therefore, it is necessary to detect the running state of ball bearing in order to prevent accidents in time. Some methods to detect the point corrosion of ball bearing is mainly discussed in this paper in terms of signal processing and verified by processing the fault data obtained from Bearing Data Center Website of Case Western Reserve University in American.Based on the analysis of the features of ball bearing faults, the four methods of time-domain analysis, frequency-domain analysis, time-frequency analysis and wavelet envelope demodulation are discussed in this paper.(1) Time-domain analysis is realized by comparing the appropriate parameters of the fault vibration signals and the normal vibration signals of ball bearing. In this paper, the four statistical parameters of kurtosis, peak indicators, margin targets and impulse indicators are chosen.(2) Cepstrum analysis is mainly considered in frequency-domain analysis. First of all, the fault signals are denoised, and then the real cepstrums are obtained; finally the inverse frequency is converted to reach the result. The method can detect the periodicity of signals which is difficult to identify with no impact from transmission path characteristics.(3) The method of time-frequency analysis, which is realized by low-pass filtering first and getting pseudo wigner-ville time-frequency distribution, is discussed. This method can reflect the time-frequency co-feature of the fault signals, so that its information in time-domain and frequency-domain is wholly grasped and the variation of frequency becomes much clear.(4)The vibration signals of ball bearing faults are multi-decomposed by wavelet transform, and the parts of the high-frequency are extracted. Afterward, the Hilbert transform is used to demodulate the parts, and the frequency analysis of the signals demodulated has been done to obtain the wavelet spectra from which the fault characteristic information of ball bearings is obtained.