Multi fault detection of the roller bearing using the wavelet transform and principal component analysis

Vibration monitoring and analysis techniques are the key features of successful predictive and proactive maintenance programs. In this work, advanced vibration analysis techniques like Wavelet transform, principle component Analysis (pcA) and squared prediction Error (spE) have been used to detect the faults in bearing. Discrete Wavelet Transforms (DWT) decomposes signal to high and low frequencies. pcA is employed to extract important feature and reduce dimension. spE is used to detect the bearing faults. The experimental data is collected from Spectra Quest's Machine Fault Simulator (MFS-4) apparatus. In this study, four rollers were bearing defects (ball defect, outer race defect, inner race defect and combined defect) for 1" and 3/4" bearing. From the results, the suggestion techniques can be used to detect multi-faults in the bearings. The results show that the best wavelet function is coiflets4 in this method.