| Vibration analysis has been widely used in the condition monitoring and fault diagnosis of the transmission system. Usually, there is much noise in the vibration signals of transmission system, which makes the spectra of signal and noise overlap in the frequency domain. So, this paper mainly aims at the research of vibration signals processing techniques that suitable for the transmission system. Now Blind Source Separation (BSS) is one of the hottest and most exciting topics in the fields of signals processing, but its real application on the mechanical vibration is few reported. This paper will do such research on the basis of transmission vibration experiment.This paper deals with the research on wavelet analysis and BSS, and then obtains some related algorithms of wavelet de-noising, FastICA and its improved Newton iterations. As the real conditions of vibration signals are often corrupted by noise, a new method is proposed to select the delayed autocorrelation and wavelet de-noising as a signal preprocessing, and then using the improved FastICA to separate the preprocessed signals, which called"wavelet de-noising-BSS-wavelet de-noising"in this paper. After that, a faults experimental platform for the transmission system is established, and a lot of gear vibration signals are acquired with it. At the same time, a research on the bearing fault data is completed, and some useful analysis results are also obtained. By the experiments, this paper checks and proves the practicability and applicability of the new method. These experiments not only build a foundation for the real application of BSS but also provide a new method for fault diagnosis on the transmission system. |
