Research On Fault Diagnosis Methods For Rotating Machinery Based On Hilbert-Huang Transform

At present, the rotating machinery has been widely used in almost all of the industry sections. Therefore it is extremely significant to detect the fault of rotating machinery. As we known, extracting fault feature is the key of fault diagnosis. To extract fault feature effectively, signal processing-based methods are widely used today. Due to the fact that most of fault vibration signals of the rotating machinery present nonstationary properties, it is essential to choose appropriate signal processing methods that are suitable for nonstationary signals to extract fault feature.The time-frequency analysis methods are widely used in rotating machinery fault diagnosis because they can provide both time and frequency domain information of a signal simultaneously. However, the time-frequency analysis methods such as windowed Fourier transform (WFT), Wigner-Ville distribution (WVD) and wavelet transform have their own limitations. Therefore, in the quest of accurate results of fault diagnosis, it is necessary to adopt the novelty theories and signal processing methods to improve the techniques of fault diagnosis. Recently, a novelty time-frequency analysis method, Hilbert-Huang transform (HHT), which is suitable for nonstationary signals, has been put forward and confirmed to be superior to the other signal processing methods in many applications. Supported by National Natural Science Foundation, this dissertation prior introduces HHT into rotating machinery fault diagnosis, whose aim is to extract fault feature of roller bearing, gear and rotor system by using HHT.The main research work of this dissertation includes two aspects. One is the theory research of Hilbert-Huang transform; the other is the research of the fault diagnosis approaches for the rotating machinery based on Hilbert-Huang transform. The main innovative work is as follows:1. The theory of Hilbert-Huang transform is studied. Thus, both the IMF criterion problem and the end effects are solved.(1) It is for the first time that an analysis method for signal instantaneous attribute using wavelet transform based on EMD is proposed, by which the complete time-frequency distribution of nonstationary signals can be obtained. Thus, the problem that appears when using wavelet analysis to calculate the instantaneous physical parameters of the broadband nonstationary signals could be solved, and the end effects that must be solved when HHT is used to calculate the instantaneous physical parameters of the nonstationary signals are avoided.(2) According to the integrity and orthogonality of EMD method, it is for the first time that the energy difference tracking method is proposed to determine the "sifting" times. The IMFs obtained by the proposed method would not only meet the orthogonality but also reflect the intrinsic information of the signal.(3) The time series prediction approaches based on support vector regressive machines and time-varying AR model are proposed, by which the end effects can be restrained effectively and the accurate IMFs and instantaneous frequency and instantaneous amplitude can be obtained.2. Hilbert-Huang transform is applied to fault feature extraction of the rotating machinery and corresponding fault diagnosis methods are proposed.(1) The time-frequency entropy method based on Hilbert-Huang transform is proposed. The study results show that the time-frequency entropy of the normal gear vibration signal is relatively big, while the time-frequency entropy would decrease when faults occur in gear. Therefore the time-frequency entropy based on Hilbert-Huang transform could be served as the feature parameter for the classification of gear work state.(2) To target the modulation characteristic of the gear fault vibration signals, it is for the first time that the frequency family separation method based on EMD is proposed. The research results show that proposed method can separate the frequency families effectively. Hence, the disadvantage of band-pass filter in the conventional envelope method could be overcome.(3) When the rotor local-rub faults occur, the fault vibration signal would present the amplitude modulation characteristic due to the periodic rub between the rotor and the stator in the rotor rotation. According to the amplitude modulation characteristic of the rotor local-rub fault signals and the characteristics of EMD method, a method of rotor local-rub fault feature extraction based on EMD is proposed. The research results show that proposed method can effectively separate the local-rub fault information and the background information of vibration signal.3. Hilbert-Huang transform and some relevant mathematical methods are combined to extract fault feature of the rotating machinery and corresponding fault diagnosis methods are proposed.(1) The concept of local Hilbert marginal spectrum is introduced. Targeting the modulation characteristics of the roller bearing fault vibration signals and the disadvantage of the conventional envelope method, a method of fault feature extraction based on local Hilbert marginal spectrum is put forward, in which wavelet packet...