Monitoring abnormalities within reactor coolant pumps due to vibration, we
introduce a Wigner distribution (WD) to a vibration monitoring system for a
nalyzing vibration signals, and develop an on-line diagnostic method using
the neural network employing back propagation algorithm. A rotor kit is use
d to simulate the abnormal conditions in the pump such as bearing rubbing,
shaft bending, misalignment, etc. The data collected from the kit are analy
zed using both a conventional system based on the Fourier transform (FT) an
d the software developed with the WD in this study, and they are compared.
The study shows that spectrum represented by WD makes it easy to analyze th
e vibration signals of a rotating machine and the developed diagnostic meth
od will help the operators grasp the cause of vibrations. (C) 2000 Publishe
d by Elsevier Science Ltd. All rights reserved.