Wavelet De-noising Technique for Roller Bearing Fault Detection

Bearings are the most important and frequently replaced component in vast majority of rotating machines. The vibration signature of a damaged bearing consists of exponentially decaying impact response that occurs periodically at the characteristic defect frequencies. At the early stages of bearing failures, these defect frequencies are masked by high amplitude machine noise hindering its detection by conventional diagnostic technique. The use of wavelet analysis to diagnose faults in roller bearing is examined in the present paper. Applying the wavelet transform to time signal obtained from the sensor placed on a bearing housing gives the information about the bearing defect frequencies even at very low signal to noise ratio. In the present paper, use of wavelet transform technique has been illustrated through numerically simulated bearing faults and experimental signal from faulty bearing in order demonstrate the effectiveness of technique in detecting bearing faults. The wavelets transform technique has been compared with the spectrum analysis method. Both the spectrum and wavelet time frequency methods have been applied to simulated and experimentally measured signal for bearing; The paper shows that wavelet transform technique can be effectively used for fault detection in rolling element bearings at a very early stage.