Research On The Application Of Hilbert-Huang Transform In Vibration Signal Analysis Of Rrotating Machinery

Monitoring and diagnosis in terms of vibration signal is the main means of management and maintenance for large rotating machinery. Due to the fact that most of fault vibration signals of rotating machinery present non-stationary properties, it is important to research advanced non-stationary signal process method to ensure the safe and reliability operation of large rotating machinery. Compared to the Fourier-based linear and stationary spectral analysis, the Hilbert-Huang transform (HHT) which based on the empirical mode decomposition (EMD) was considered a great breakthrough and widely used in engineering filed. However, some shortcomings of the HHT have been found, aiming at the existing problems in practical application and considering the characteristics of vibration signal in rotating machinery, an improved HHT is proposed in this paper. The Hilbert spectrum features of several typical faults are also analyzed according to faults simulation experiments and practical faults, which can provide some useful guidance for fault diagnosis of rotating machinery. The main works of this dissertation are as in the following:1. Improved algorithm of HHT achieved in several aspects. The boundary process technique in HHT discussed and the reason of end effect is analyzed. Summarized the existing boundary process methods, a wave matching extending method is put forward to solve the end effects both in EMD and Hilbert transform by only once extension. In order to solve the overshoots and undershoots in cubic spline interpolation, an envelope fitting algorithm based on constrained spline interpolation is proposed, which can decrease error in EMD and enhance signal resolving power in HHT. Because of over decomposition, EMD generating undesirable intrinsic mode functions (IMFs) at the low-frequency region, called pseudo components. The mutual information between each IMF and original signal adopted to identify these pseudo components. Moreover, the mode mixing caused by abnormal events in HHT is studied, which shows that the result of the first IMF has the most influence on the whole precision of EMD. The mode mixing is also solved by adding a high frequency harmonic signal.2. The application of mathematical morphology in vibration signal pre-process is studied and a new de-noising technique of vibration signal is proposed by using open-closing and closing-open combined morphology filter. Furthermore, the ability of morphological filter dealing with the different type and intensity noises is discussed. It shows that the application of mathematical morphology to the signal preprocess in HHT can greatly reduce the influence of complex abnormal events and decreased mode mixing.3. The HHT is used for vibration-based fault diagnosis by fault simulation experiments on rotor test bed. The Hilbert spectrums of rotor running across critical rotating, run-up process under local rub-impact and oil film instability fault are given respectively. Besides, the Hilbert spectrum characteristics of several typical faults, such as rotor startup, unbalance, local and full annular rubbing ,oil whirl and oil whip , are analyzed respectively. The method also applied in practical faults diagnosis, such as rubbing and oil whip.