Fault Detection and Diagnosis of Rolling Element Bearing based on Neural Network



Published Jul 14, 2017
Chaeyoung Lim Seokgoo Kim Joo-Ho Choi


In this paper, a neural network based diagnosis technique is developed to detect fault and estimate the severity of the spalls at the inner race, outer race, roller and cage. To this end, a bearing test rig is developed, in which the normal and faulted bearings with differing spall size at different locations are operated under accelerated loading conditions. Features are extracted from the bearings, which include the time-based indicators such as RMS, peak, crest and kurtosis, frequency based indicators obtained by envelope analysis, and timefrequency based ones like wavelet decomposition. Neural network model is constructed using the features for the classification. The model is then applied to diagnose the fault in the new bearings, which includes the identification of the fault and severity. Particular attention is made to the study of statistical significance to check the validity of the model.

Abstract 33 | PDF Downloads 35




de Azevedo, et al. (2016). A review of wind turbine bearing condition monitoring: State of the art and challenges. Renewable & Sustainable Energy Reviews, 56, 368-379.
Randall, R. B., & Antoni, J. (2011). Rolling element bearing diagnostics—a tutorial. Mechanical systems and signal processing, 25(2), 485-520.
Siew, et al. Fault severity trending in rolling element bearings.
Lessmeier, et al. (2016) Condition Monitoring of Bearing Damage in Electromechanical Drive Systems: A Benchmark Data Set for Data-Driven Classification.
Samanta & Al-Balushi (2003). Artificial neural network based fault diagnostics of rolling element bearings using time-domain features. Mechanical systems and signal processing, 17(2), 317-328.
Special Session Papers