Neural Network Based Ball Bearing Fault Detection Using Vibration Features for Aerospace Applications 942168
Traditionally, ball bearing condition monitoring is done by a human expert whose judgement is based on bearing vibration and temperature. In this paper, a method is described for classifying normal ball bearings and damaged ball bearings using scalar features, derived from their vibration signals, and a feedforward multi-layer neural network, trained using the back propagation algorithm. Two experimental test rigs, used for acquiring the vibration signals for the two types of ball bearings studied here, are described. Several scalar features, derived from the raw vibration signals, are discussed. Next, training of a feedforward multi-layer neural network with these scalar features, using back propagation algorithm, is presented. It is shown that with these scalar features, the neural network is successful in classifying normal and damaged ball bearings.
Citation: Haddad, S. and Chatterji, G., "Neural Network Based Ball Bearing Fault Detection Using Vibration Features for Aerospace Applications," SAE Technical Paper 942168, 1994, https://doi.org/10.4271/942168. Download Citation
Author(s):
Sam David Haddad, Gano B. Chatterji
Pages: 10
Event:
Aerospace Technology Conference and Exposition
ISSN:
0148-7191
e-ISSN:
2688-3627
Related Topics:
Neural networks
Fault detection
Bearings
Test facilities
Vibration
Education and training
Mathematical models
SAE MOBILUS
Subscribers can view annotate, and download all of SAE's content.
Learn More »