文档介绍：Wavelet Packet Feature Extraction for Vibration Monitoring
Gary Y. Yen and Kuo-Chung Lin
School of Electrical puter Engineering
Oklahoma State University, Stillwater, OK 74078
******@
Abstract structured and could be grouped into two categories:
sustained defects and intermittent defect [9]. For
Condition monitoring of dynamic systems based on sustained defects, Fourier based analysis, which uses sine
vibration signatures has generally relied upon Fourier wave functions as basis functions, provides an ideal
based analysis as a means of translating vibration candidate for extraction of these narrowband signals. For
signals in time domain into the frequency domain. intermittent defects, the signals often demonstrate a
However, Fourier analysis provided a poor nonstationary and transient nature, and carry small yet
representation of signals well localized in time. In this ponents embedded in larger repetitive
case, it is diflcult to detect and identifi the signal signals. In this case, the Short Time Fourier Transorm
pattern *om their expansion coeficients because the (STFT) can be employed to detect the local transient.
information is diluted across the whole basis. The Unfortunately, fixed windowing implies fixed time-
wavelet packet transform is introduced as an alternative fiequency resolution in the time-freq