Envelope Demodulation Analysis Methods Of Vibration Signals: Research And Application

This academic dissertation has devoted its research subject to demodulation or enveloping based methods of mechanical faults vibration signals. Some existing methods, such as the Hilbert transform, gerneralized demodulating-filtering, energy operator, cyclostationary and empirical mode decomposition based methods, were briefly reviewed for their principles, features and disadvantages. Analytic wavelet transforms, S- transform and the stochastic resonance effect were explored in the dissertation to develop new demodulation methods for cycle-modulating signals, which are expected to have good noiseproof properties.Analytic wavelets are complex wavelets, whose real part and imaginary part can constitute an Hilbert transform pair, and usually used to extract the signal envelope. The Morlet wavelet is the most familiar one amongst them. A sufficient condition for a analytic wavelet to be an Hilbert transform pair is presented, that is the analytic wavelet with real valued frequency representation. For this group of analytic wavelets, the conclusion that the real part and imaginary part of their transform coefficients also constitute an Hilbert transform pair was perfectly deduced. Upon these discussions, a straightforward deduction is that Harmonic wavelets and combined Harmonic wavelets also belong to above mentioned analytic wavelets. They can also be used to the envelope-demodulation of mechanical faults vibration signals. Combined Harmonic wavelets were proposed to design a comb-filter and a envelope-demodulator for a specified cycle-modulating signal. This technique integrates comb-filtering and envelope-demodulating, and can obtain a clear and noiseproof envelope spectrum, compared to the band-pass filtering based demodulation techniques.S-transform is a new time-frequency analysis method, and simultaneously has good features of continuous wavelet transform and short-time Fourier transform. It has a interrelation with a concrete continuous Morlet wavelet transform, is a phase correction of the Morlet wavelet transform. The fact that the slice of S-transform time-frequency spectrum could be the envelope of the amplitude-modulated signal is employed to detect and extract a periodical amplitude-modulating signal and its envelope, combing with a robust method for periodicity detection and characterization of irregular cyclical series in term of embedded periodic components. This periodicity detection method is based on singular value decomposition (SVD), and has the advantage over the spectrum analysis and autocorrelation function analysis in detecting periodic components embedded in stronger noises. Using simulated signals and the faulty bearing vibration signals, the proposed envelope-demodulation method was proven to be noiseproof.The stochastic resonance effect has particular advantages on enhancing and detecting weak signals. Enhancement and extraction of the weak low-frequency amplitude-modulated signals were studied using the combination of the stochastic resonance (SR) and the common envelope demodulation analysis. The SR effect of the signal was realized only for the low-frequency amplitude-modulating signal, not for high-frequency carrier signal, by using the step-changed numerical algorithm and the adjustment of the bistable system parameters. An automatic searching strategy for optimal the algorithm step and the bistable system parameters was adopted in order to achieve the maximum SR effect. The effectiveness of the proposed method was demonstrated on both simulated signals and real vibration signals of a low-speed and heavy-duty gearbox. It was proven to be superior to the spectrum analysis and the common envelope-demodulation analysis.The proposed envelope-demodulation methods in this dissertation have be demonstrated on their effectiveness and superiority to the common envelope-demodulation methods. It's worth reling that they would have a promising application for the fault diagnosis of the gear and rolling element bearings.