Fault Diagnosis of Rolling Element Bearing Using Artificial Neural Networks

Abstract

Bearings are one of the most crucial components of rotating machinery. The condition of bearing contributes to overall machine performance. It is thus important to analyze the faults in the bearing in earlier stage in order avoid catastrophic failure. The condition monitoring based on vibration measurement can be used to identify defects in bearing. The present study is carried out to identify the effect on vibration spectrum of a ball bearing having defect at inner race using FFT analyzer. An experimental setup is developed. The fault is created at the inner race of the bearing by using electrical discharge machining (EDM). The frequency spectrum is acquired for faulty as well as healthy bearing using FFT analyzer. It can be concluded that high Peak in amplitude of vibration observed at BPFI in frequency spectrum indicates that fault is present at inner race of ball bearing. Only experimental observation methods depend on human knowledge and experience, which shows error. Therefore, it is necessary to detect fault without human intervention automatically using a computer. This study gives overall review for fault diagnosis using Artificial Intelligence. The experimental data is used for training and testing the Artificial Neural Network (ANN) to detect the fault automatically in MATLAB environment. The accuracy of ANN is found to be 94.27%. The results are in close agreement for the similar condition available in literature.