Ball bearing early fault detection using wavelet analysis

Detection of ball bearing faults has been for long time a great concern in the field of condition monitoring of machines. One of the standard methods used in industry is the vibration envelope method or High-Frequency Resonance Technique (HFRT). In the present work, wavelet transform is implemented in HFRT for detecting the fault signatures that are buried in mechanical and electro-magnetic induced vibration signals. Induction motor ball bearing vibration signals are filtered using Daubechies orthonormal wavelet family and the filtered signal is amplitude demodulated using a digital envelope. The spectrum of the signal envelope clearly identifies the incipient faults. The results obtained are superior to those using the classical HFRT because the signal to noise ratio is increased and therefore the ability and accuracy in detecting early failures are improved. The strength of this technique lies in its robustness to high background noise levels and its virtual independence on the mechanical carrier frequencies. In this way the need for band pass filters centered on resonant structural frequencies is eliminated. This paper presents the results of an experimental study performed on induction motors ball bearings in the presence of an overwhelming electro-magnetic noise, and with limited access to bearing housing.