The Combination of the Hilbert Transform and the Fourier Transform for Mechanical Vibration Signal Analysis
Xiao Wang,Yan Zhu,Bo Yu
DOI: https://doi.org/10.2991/icmmcce-15.2015.317
2015-01-01
Abstract:In this paper we propose one method for mechanical vibration signal analysis. The given mechanical vibration signal is firstly decomposed into several intrinsic mode functions (IMFs) by the empirical mode decomposition (EMD) method. The Hilbert transform of each IMF, calculated by a recently developed fast algorithm, is then implemented and the analytical signal corresponding to each IMF is obtained. After writing the analytical signal in polar coordinates, we do the refinement Fourier spectrum analysis for the phase function of every analytical signal. By comparing the behavior of these analytical signals, which correspond to different IMFs, in the frequency domain with the characteristic frequency of the mechanical system, we distinguish the IMFs which contain the effective information from the IMFs which consist of noise in the original signal. Numerical results show that the combination of the Hilbert transform and the Fourier analysis works well for practical signals. Introduction Since the mechanical systems such as the industry manufacture and the transportation tools play more and more important role in the world, the study of vibration signal generated by mechanical system is very important. The analysis of mechanical vibration signal has a lot of applications in different fields, for example, the fault diagnosis for mechanical systems including machines or some components of machines, the vibration monitoring and control, the production quality control, and the noise source location and noise cancellation. Generally speaking, the mechanical vibration signal analysis consists of the extraction of information from measured signal patterns. Traditional approach to investigation of mechanical vibration signal can include a spectral analysis based on the Fast Fourier Transform (FFT) [1] and a time-frequency distribution analysis based on the wavelet transform [2]. The classical spectrum analysis method based on FFT has the advantage of high speed but suffers from unsatisfactory understanding of the underlying mechanical system, especially in processing the time-varying vibration signals, such as weak signal submerged by the side lobe of strong signal, low-frequency resolution, and signals corrupted by strong noise. The time-frequency distribution analysis method based on wavelet transform can decompose any signal into a linear combination of a set of basic functions which are dilations and translations of a given mother wavelet. If the mother wavelet has some good properties such as compact support, high vanishing moments and so on, then this decomposition of the signal will present a reasonable time-frequency distribution, from which we can find out the signalu0027s behavior in different frequency bands and at different time durations. However, every time when we try to obtain the time-frequency distribution based on the wavelet transform, we need to choose one suitable mother wavelet, which is, in some sense, empirical. Currently, the application of Hilbert transform to mechanical vibration signal analysis attracted considerable attention, especially after the advent of the Hilbert-Huang transform, which consists of 4th International Conference on Mechatronics, Materials, Chemistry and Computer Engineering (ICMMCCE 2015) © 2015. The authors Published by Atlantis Press 1668 the EMD method and the Hilbert spectrum analysis [3]. For example, some authors combined the EMD method and the wavelet transform to present a fault diagnosis for wind turbine gear box [4]. For more information of the use of Hilbert transform in mechanical vibration signal analysis, please refer to [5]. The success of the Hilbert-Huang transform lies in two facts. The first fact is that, by using the EMD method, it can decompose any signal into a sum of IMFs adaptively and quickly. The second fact is that it can present a global Hilbert spectrum analysis, from which we can easily find out the instantaneous frequency of the analyzed signal. However, there is no need to implement the Hilbert spectrum analysis in mechanical vibration analysis because the main advantage of the processing of mechanical vibration signal is that we actually know some information of the analyzed signal before we analyze it, especially in the frequency domain. For example, when we do fault diagnosis for gear box, we do have some priori knowledge about the behavior in the frequency domain of the signal generated by the defect gear box. Keeping this in mind, we can only use the EMD part of the Hilbert-Huang transform to decompose the signal into several IMFs. These IMFs may contain the noise, the characteristic information of the mechanical system and some other information. The next step is to find out the IMFs which contain the effective information and the IMFs which contain the noise information. For this purpose, we calculate the Hilbert transform of each IMF, from which we obtain the analytic signal corresponding to each IMF. Thus, we can get the instantaneous amplitude function and instantaneous phase function with respect to each IMF. After applying the refinement Fourier spectrum analysis for all these phase functions and comparing them with the characteristic frequency of the underlying mechanical system, we can extract the informative IMFs from the given signal. This paper is organized as follows. Section 2 describes the algorithms, including the EMD method, the calculation of the Hilbert transform, and the refinement Fourier spectrum analysis. Section 3 presents several numerical results. The conclusion will be drawn in Section 4. Algorithm Description In this section we will describe our algorithm which consists of three steps. The first step is to decompose the given mechanical vibration signal into several IMFs by using the EMD method. The second step is to implement the Hilbert transform on these IMFs so that we get the corresponding analytic signal, from which the phase functions are obtained. Using the refinement Fourier spectrum analysis, the third step is to find out the effective IMFs according to the characteristic frequency of the mechanical system. The EMD method was proposed by Huang et al. [3] in 1998. It is a sifting process based on the local oscillation of the signal. The main advantage of this method is that it does not need to pick up any basis functions, which is necessary by the conventional method such as Fourier method and wavelet transform method. This advantage makes the EMD method beyond the choice of different basis functions, which depends on different signals to be dealt with. By this method, the signal will be decomposed into a finite sum of IMFs and a residue, where the IMFs should satisfy: (i). The number of the extrema and the number of the zero crossings are equal or differ at most by one; (ii). At any point, the mean value of the envelopes defined by the local extrema is zero. The detailed algorithm of EMD method can be found in [3]. From the view of mathematics, we can say that the IMFs are adaptive basis functions for some function space. Thus, the EMD method actually presents an adaptive function decomposition method algorithmically. After the sifting process, the original signal will be decomposed into several IMFs. The next step is to find out the IMFs which contain the effective information and the IMFs which contain the noise information. Generally this is not an easy task. However, the main advantage of the mechanical vibration signal analysis is that we have some priori knowledge about the exact characteristic information, in the frequency domain, of the given mechanical vibration signal. Thus, the Fourier spectrum analysis for these IMFs will be applied. However, before this procedure, we will try to “separate the amplitude part and phase part of these IMFs to get a more meaningful Fourier spectrum analysis. For this purpose, we calculate the Hilbert transform of these IMFs in next step.