Bearing fault diagnosis using multi-layer neural networks

Wang, Ling and Hope, A.D. (2004) Bearing fault diagnosis using multi-layer neural networks Insight, 46, (8), pp. 451-455. (doi:10.1784/insi.46.8.451.39377).


[img] PDF 23201.pdf - Version of Record
Restricted to Repository staff only

Download (994kB)


This paper presents an investigation into bearing fault diagnosis on centrifugal pumps which can be applied to the waste water industry. After the establishment of a vibration monitoring system in a pumping station, signals were recorded regularly to build up a vibration database for future monitoring. Conventional methods were used to analyse the vibration signals and some bearing defects were successfully detected.
However, the major challenge for this project was to use a vibration monitoring system to predict pump bearing faults automatically instead of by analysing the data off-line. This paper proposes a solution based on artificial neural networks (ANNs), which is a powerful technique for pattern recognition and can be applied to the classification of pump faults. Due to the long period of time required to obtain essential information from the pumping station for neural network training, a test-rig pump was established in the laboratory to simulate the common pump faults, including typical bearing defects. Bearing faults were simulated by generating pit marks on the bearings using electrical discharge machining (EDM.). Vibration signals in the time-domain were collected and pre-processed to act as the inputs to neural networks. The neural networks were then trained and their classification accuracy rates evaluated. In this study neural network models were designed using the Matlab Neural Network Toolbox and the models which successfully classified the vibration signals were chosen for pump fault diagnosis.

Item Type: Article
Digital Object Identifier (DOI): doi:10.1784/insi.46.8.451.39377
ISSNs: 1354-2575 (print)

ePrint ID: 23201
Date :
Date Event
Date Deposited: 24 Mar 2006
Last Modified: 16 Apr 2017 22:46
Further Information:Google Scholar

Actions (login required)

View Item View Item