Research On The Generalized Demodulation Time-frequency Analysis Method And Application To Gear Fault Diagnosis

As a universal part of transmitting movement and momentum, gear has been used in mechanical equipment widely. In many cases, gear fault is the main reason leading to mechanical equipment failure. Therefore, gear fault diagnosis is significant. In the research content of the gear's fault diagnosis technique, the characteristics analysis of the fault is the core. The fault characteristics can be obtained by analyzing the original signal for the pattern recognition and the fault diagnosis. Therefore, it is necessary to choose a suitable signal processing method.The generalized demodulation time-frequency analysis approach is a new signal processing method, for which, by using generalized demodulation, the signal with time-frequency distributions that are curves can be transformed into the signal that are lines paralleling the time coordinate axis. Then through a maximal overlap discrete wavelet packet transform (MODWPT), the generalized demodulated signal is decomposed and some single-component signals whose instantaneous amplitudes and frequencies are of physical significance can be obtained. It was proved that this method is superior to the other time-frequency analysis method in many ways, such as Window Fourier Transformation, Wavelet Transformation, EMD method and so on. The generalized demodulation time-frequency analysis approach is very suitable to process the multi-component amplitude-modulated and frequency-modulated (AM-FM) signal. And when the gear has a fault, its vibration signal mostly shows off the multi-component AM-FM signal. Therefore, this method is suitable to process the gear fault vibration signal.The paper mostly completed two sides of the research:the fundamental research to the generalized demodulation time-frequency analysis approach and the research to the gear fault diagnosis method based on the generalized demodulation time-frequency analysis. The main works and innovation points were done as follows:1. There was a research to the choice of the phase function. The research result discovered that even though the each component's time-frequency distribution is not parallel for each other, if only the each component's time-frequency distribution of the signal is in the different rectangle of the wavelet packet time-frequency space and each rectangle of the wavelet packet time-frequency space only obtain the one component at most, at this time the single-component signals'instantaneous amplitudes and frequencies are also of physical significance and the veracious result still can be obtained.2. The generalized demodulation time-frequency analysis approach has very clear advantage in decomposing the multi-component AM-FM signal. The improved generalized demodulation time-frequency analysis approach and the repeated generalized demodulation method were proposed in this paper. And the two methods were applied to decompose the multi-component signal. The analysis result from the simulated signal showed the effectiveness of the two methods.3. For the multi-component signal including the noise, the generalized demodulation time-frequency analysis approach can effectively obtain the main frequency component from the original signal. Therefore, the method can be applied to analyze and process the practical engineering signal.4. The Hilbert spectrum method based on the MODWPT was proposed and applied to the gear fault diagnosis. The analysis result from the simulated signal showed that the Hilbert spectrum method based on the MODWPT is clearly superior to the Hilbert spectrum method based on the EMD. The analysis result from the experimental signals showed that the Hilbert spectrum method based on the MODWPT has very clear advantage to analyze the vibration signal of the gear with fault when it operates in a stable speed.5. The vibration signal of the gear with fault was analyzed when it operates in a mutative speed using the method which assembles the generalized demodulation time-frequency analysis approach and the order analysis method. In order to study the amplitude-modulated signal, the envelope order spectrum method based on the generalized demodulation time-frequency analysis was proposed. In order to study the phase-modulated signal, the order spectrum method based on the generalized demodulation time-frequency analysis and the instantaneous frequency calculation was proposed. The analysis results from the simulated and experimental signals showed the effectiveness of the two methods.