Pattern Recognition of Bearing Faults using Smoother Statistical Features

A new diagnostic scheme is presented for ball bearing localized faults, which utilizes preprocessed time domain features based pattern recognition (PR). Vibration data is acquired from faulty bearings using a test rig, and the features are extracted from the data segments that are preprocessed prior to use in the fault classification process. The preprocessing involves smoothing of the features, which reduces the undesired impact of noise and vibration randomness on the PR process, and thus enhances the diagnostic accuracy. The results are compared with a similar scheme in terms of minimum features requirement to achieve an optimum classification accuracy, and the feature processing based proposed scheme provides better results. keywords: Fault Diagnosis, Vibration Analysis, Feature Processing, Pattern Recognition.