Rolling Element Bearing Fault Diagnosis by Different Data Fusion Techniques

Rolling element bearing is a crucial element of rotating machinery. A sudden failure of the bearing may result in a catastrophic failure. Therefore, the identification of bearing failure in the incipient stage is essential. The vibration signal generated by bearing fault is used for condition monitoring and fault diagnosis. The parameters extracted from vibration data provide the fault indication. However, a single parameter may miss useful information, resulting in less accurate fault diagnosis. The fused parameters after data fusion of individual parameters are more informative and efficient than a single parameter. Mahalanobis–Taguchi–Gram–Schmidt method, principal component analysis, and independent component analysis are the most popular data fusion techniques, and these techniques are applied to three most occurring fault data such as the outer race, the inner race, and rolling elements. The bearing vibration data, available on the NASA Web site, are used for the data analysis and extracting parameters. A comparison is made on the effectiveness of various data fused parameters obtained from different data fusion techniques. The best data fusion technique with respect to each type of defect (inner race, outer race, and ball) is identified.