An improved Hilbert-Huang transform and its application in vibration signal analysis

The vibration generated by industrial machines always contains nonlinear and non-stationary signals. Recently., a number of new methods have been proposed to analyse these signals. One of the promising methods is the Hilbert-Huang Transform (HHT). The HHT is derived from the principals of empirical mode decomposition (EMD) and the Hilbert Transform. When applying the HHT, first, the EMD will decompose the acquired signal into a collection of intrinsic mode functions (IMF). The IMF is a kind of complete, adaptive and almost orthogonal representation for the analysed signal. Since the IMF is almost monocomponent, it can determine all the instantaneous frequencies from the nonlinear or non-stationary signal. Second, the local energy of each instantaneous frequency can be derived through the Hilbert Transform. Hence, the result is an energy-frequency-time distribution of the signal. Since applying the process of HHT is not computational intensive, the HHT becomes a promising method to extract the properties of nonlinear and non-stationary signal. However, after the completion of a thorough experiment, the result generated by the HHT has its deficiency. First, the EMD will generate undesirable IMFs at the low-frequency region that may cause misinterpretation to the result. Second, depends on the analysed signal, the first obtained IMF may cover too wide a frequency range such that the property of monocomponent cannot be achieved. Third, the EMD operation cannot separate signals that contain low-energy components. In this study, new techniques have been applied to improve the result of HHT. In the improved version of HHT, the wavelet packet transform (WPT) is used as preprocessing to decompose the signal into a set of narrow band signals prior to the application of EMD. With the help from WPT, each IMF derived from the EMD can truly become monocomponent. Then, a screening process is conducted to remove unrelated IMFs from the result. Both simulated and experimental vibration signals of having a rotary system with the fault of rubbing occurred have proven that the improved HHT does show the rubbing symptoms more clear and accurate than the original HHT. Hence, the improved HHT is a precise method for nonlinear and non-stationary signal analysis. (c) 2004 Elsevier Ltd. All rights reserved.