| Autor: |
Shukla, Prashant Kumar, Roy, Vandana, Chandanan, Amit Kumar, Sarathe, Vivek Kumar, Mishra, Pankaj Kumar |
| Zdroj: |
SN Computer Science; November 2023, Vol. 4 Issue: 6 |
| Abstrakt: |
The sound and vibration signals coming from the spinning machinery are non-stationary signals, and to analyse them, we require information on both their time and frequency components. The signal in the time domain is converted by the Fourier transform into one in the frequency domain, which does not include any information about the passage of time Wavelet transforms, can represent non-stationary signals in time and frequency. It can breakdown source signals of different time and frequency resolutions, which are characteristic of rotating machine fault mechanisms. Rotating the machine achieves this. Wavelet can filter raw waveforms and compress massive amounts of data without losing information. Thus, eliminating background noise improves sound-based defect diagnosis. Sound and vibration data recovered wavelet energy features and wavelet energy to entropy characteristics. Eight decomposition stages extracted wavelet features. Wavelet characteristics have nine coefficients: one approximate and eight detailed. Decision trees were used to select the best wavelet features. Wavelet energy features outperform wavelet energy-to-entropy ratio features. The classification algorithm's mean classification efficiency percentage for vibration signals of class 24 is 94.23 and for sound signals, 96.73. |
| Databáze: |
Supplemental Index |
| Externí odkaz: |
|
Full text from SpringerLink