The Improved Hilbert-Huang Transform And Its Application In Time-frequency Analysis

The core content of Hilbert Huang transform(HHT) is empirical mode decomposition(EMD) and Hilbert spectral analysis(HAS). This method has good time-frequency resolution and time-frequency concentration and is suitable for nonlinear and non-stationary signal analysis and processing. However, the spline interpolation is in overshoot and undershoot and other issues, causing intrinsic mode function(IMF) decomposed by EMD has two problems:first, the high frequency IMF contains multiple band, causes mode mixing. In second, low IMF has illusive components. This paper puts forward a improved HHT algorithm in order to solve HHT transform problem, and make the application of it in ultrasonic signal processing. The specific research contents are as follows:(1) Study the basic principle of Hilbert-Huang transform and its applications in the signal analysis.(2) In view of the mode mixing problem, this paper put forward a kind of improved HHT algorithm based on wavelet packet which doing pretreatment for signal by wavelet-packet transform firstly, then applying the EMD decomposition.(3) Aiming at the problem of illusive components, puts forward The log-likelihood test(LLT) to identify and remove the illusive components. This method is with the log-likelihood value of each IMF component and the original signal as the judgment basis. A number of simulation signal analysis showed that, this method is more effective and more accurately than the relational coefficient method which is used by the present mainstream in eliminating the illusive components.(4) Based on the simulation of signal analysis, study using the improved HHT method in the engineering test signal and the rotating equipment fault diagnosis signal processing. First of all, for engineering test signal, based on the EMD decomposition, use logarithm likelihood test method to determine the useful components, then is by cross-comparing to extract feature information of material damage within this range. Secondly, for the rotating equipment fault diagnosis signal, using HHT marginal spectrum to reflect bearing fault characteristics, and put HHT marginal spectrum and envelope spectrum analysis results for comparison, found the HHT marginal spectrum of bearing fault analysis effect than does the envelope spectrum difference.