Research On Time-frequency Analysis Methods And Its Applications To Rotating Machinery Fault Diagnosis

Study of rotating machinery fault diagnosis technology is of important significance,for theprotection of the safe operation of equipment, reducing major economic losses and avoid cata-strophic accidents.Most vibration signal of rotating machinery is non-stationary signal,time-frequency analysis method can also extract local information of the vibration signal in time do-main and frequency domain,so it is suitable for rotating machinery fault diagnosis. However, thetime-frequency analysis methods such as Short Time Fourier Transform (STFT), Winger-VilleDistribution,wavelet transform and Hilbert-Huang Transform have their own shortcomings,so itis urgent to study new fault diagnosis methods for rotating machinery. In this paper, the theoriesof the frequency slice wavelet transform, local mean decomposition, the intrinsic time scaledecomposition and its applications in rotating machinery fault diagnosis is studied deeply. Itsmain contents are as follows:1. Aiming at the problem that the noise in the signal will reduce the frequency resolution offrequency slice wavelet transform analysis,a time-frequency slicing fault diagnosis method forbearing based on morphological filtering,autocorrelation analysis and frequency slice wavelettransform is proposed.A multi-structural element difference morphological filters is proposed,and its noise reduction effect is better than a single structure element morphological filter,thesimulated signal and bearing fault diagnosis analysis of examples demonstrate the effectivenessof the method.Afault diagnosis method for gear based on morphological filtering,autocorrelationanalysis and frequency slice wavelet transform is proposed. Autocorrelation denoising is carriedout before the gear fault signal is decomposed by frequency slice wavelet transform analysis, andthe autocorrelation denoising can highlight the failure feature to improve the frequency resolu-tion.2. The principles of local mean decomposition and1.5dimension spectrum is discussed, thesimulation signal analysis shows its characteristics. Aiming at the problem that the noise in thesignal impact the results of local mean decomposition, a fault diagnosis method based on localmean decomposition and1.5dimension spectrum is proposed.Aiming at the problem of lowefficiency of local mean decomposition, a B-spline local mean decomposition (BLMD) method based on B-spline interpolation is proposed. A BLMD time-frequency analysis is proposed andapplied to the bearing and gear fault diagnosis.A fault diagnosis method based on BLMD withbi-cepstrum is proposed and applied to the bearing and gear fault diagnosis.The analysis ofsimulated signal and engineering examples of bearing and gear fault diagnosis demonstrate theeffectiveness of the method.3. Aiming at the limitations of non-stationary signal processing method and the distortionproblem of the intrinsic time scale decomposition, a B-spline improved intrinsic time-scaledecomposition (BITD) method is proposed, on the basis of BITD, a local energy spectrum basedon BITD is proposed.Aiming at the non-stationary characteristics of a gear fault vibration signal,a gear fault diagnosis method combining BITD and homomorphic filtering demodulation isproposed.In this approach, firstly, BITD is applied to decompose the vibration signal into a finitenumber of proper rotation components, then by using the correlation coefficients, the properrotation (PR) components best representing the fault information are selected to extract the faultcharacteristics through the homomorphic filtering demodulation.The method is effectively veri-fied by simulation signals and engineering examples of gear fault diagnosis.4. The principle of stochastic resonance denoising is discussed,and a feature extractionmethod combined with BITD and stochastic resonance is proposed,and analysis of the simulatedsignal and experimental signal verify the effectiveness of the method,EMD-based denoising me-thods is studied,on the basis of analysis the limitations of EMD-based denoising methods,twothreshold de-noising methods based on BITD is proposed and applied to the bearings faultdiagnosis.And the simulated signal and experimental signal verify the effectiveness of themethod.5. On the basis of analysis the principle of permutation entropy and basic scale entropy,amethod combined with BITD and permutation entropy is proposed,BITD method is applied todecompose the vibration signal into a finite number of proper rotation(PR) components,then thefirst four proper rotation components are selected to calculate permutation entropy,then the en-tropy values are input to a SVM-based classifier to distinguish the rolling bearing fault types. Byanalyzing the experimental data,the results show that the proposed method can diagnose the faultcategories effectively. Aiming at the nonlinear,non-stationary characteristics of a gear vibrationsignal,and difficulty in obtaining a large number of fault samples,a gear fault diagnosis method combining BITD and base-scale entropy is proposed.Firstly,BITD method is applied to decom-pose the vibration signal into a finite number of proper rotation(PR) components,then the firstproper rotation components is selected to calculate base-scale entropy,then the base-scale en-tropy values are inputted to a SVM-based classifier to distinguish the gear fault types.By analy-zing the experimental data,the results show that the proposed method can diagnose the faultcategories effectively.