In this research, three main time-frequency/time-scale (TF/TS) analysis algorithms - windowed Fourier transform (WFT), Wigner-Ville distribution (WVD), and wavelet analysis (WA), have been studied and implemented. We compare those different algorithms using an identical set of synthetic signals, from which one can realize their individual features. To justify the effectiveness and to highlight the requisite of using TF/TS analysis, two test-benches have been conducted and investigated. One is to analyze transient vibration signals measured on the end-effector of an industrial robot during plane motion. The other is to analyze nonstationary vibration response of a rotor-dynamic system with both an electro-mechanical clutch and brake. From the second case study, we conclude that the subsequent post-processing would be as essential as the TF/TS analysis itself. Namely, an appropriate post-processing technique is a key step to automated condition recognition rather than only via human intervention.
|Number of pages||4|
|State||Published - 1996|
|Event||Proceedings of the 1996 3rd International Conference on Signal Processing, ICSP'96. Part 1 (of 2) - Beijing, China|
Duration: 14 Oct 1996 → 18 Oct 1996
|Conference||Proceedings of the 1996 3rd International Conference on Signal Processing, ICSP'96. Part 1 (of 2)|
|Period||14/10/96 → 18/10/96|