| The mechanical fault diagnosis is very important to ensure the normal operation of the rotating machinery. Therefore, it has extremely significant reality meaning and practical merit. As we all know, extracting the fault feature from the vibration signals is the key of fault diagnosis. Most of the fault vibration signals of the rotating machinery are non-stationary. Consequently, selecting appropriate non-stationary signal processing methods is the precondition for the success of fault diagnosis.In general, the gear vibration signals which measured at equal-time-interval are the multi-component non-stationary signals and have the low signal-to-noise ratio, which makes it is very difficult to extract the fault features from the gear vibration signals with the existing signal processing techniques. On the other hand, under the early stage of rotor’s rub-impact failure, the fault feature is very weak and is always hidden in the strong power frequency signal component. Hence, it is very difficult to be detected. The multi-scale chirplet path pursuit (MCPP) method is a non-stationary signal processing method which is proposed in recent years. With the best path algorithm of the MCPP method, the signal component with the largest energy in the original vibration signal and its instantaneous frequency can be estimated by connecting the selected chirplets and their corresponding piecewise linear frequencies, respectively. For this reason, the MCPP method is suitable to analyze the non-stationary signals whose instantaneous frequencies change in wide range. In the early stage of rotor’s rub-impact failure, the power frequency component whose energy is the maximal among all the components in the rotor vibration signal can be picked up from the original rotor vibration signals by MCPP method. Thus, the MCPP method can be used to fault diagnosis of the gears with time-varying rotational speed, and can also be used to detect the weak rubbing fault features from the rotor vibration signal. Supported by National Natural Science Foundation (No.50875078) and Specialized Research Foundation for the Doctoral Program of Higher Education (No.20090161110006), this dissertation takes deep researches on the fault diagnosis of gears with time-varying rotating speed and the early detection of the rubbing fault of rotors based on MCPP method.The main researches and the acquired innovative achievements are as follows: (1) The gear vibration signals which obtained under speed-up and speed-down process are non-stationary. Therefore, the FFT-based spectral analysis of these gear vibration signals will be blurring, and have no physical meaning. Aimed at this problem, whether or not the FrFT method can be used to process the gear vibration signals which measured under speed-up and speed-down process is discussed, and the difficulties existing in the application of FrFT is further studied. Then, based on MCPP and FrFT, a novel approach for gear fault detection under speed-up and speed-down process is proposed. For a given gear vibration signal which collected under speed-up and speed-down process, the gear vibration signal segment for which FrFT is useful can be extracted from it (with the MCPP method), and the best order of FrFT for the extracted gear vibration signal segment can be calculated out, simultaneously. This is then followed by the FrFT spectrum of the extracted gear vibration signal segment. If the modulation sidebands are visible in the FrFT spectrum, we can judge the existing fault on the gear.(2) Under the time-varying rotational speed condition, the modulation sidebands of the gear vibration signals are always varying with the rotational speed, which makes the demodulation analysis of these gear vibration signals are very difficult. Because the MCPP method can estimate the instantaneous frequency of the signal component whose energy is the largest among all the components of the original gear vibration signal, it can be used to estimate the rotational speed of the gear. Then, the order-tracking of the gear vibration signal is feasible. Furthermore, the gear vibration signal which obtained from the engineering practice always has the low signal-to-noise ratio. Hence, the morphology analysis which is an effective tool to extract impulsive signal components of the vibration signal and has the robust anti-noise ability is introduced. Then the order multi-scale morphology modulation method based on multi-scale chirplet path is proposed.(3) When the mechanical systems broke down, the vibration signals are always companied with noise and modulations, and that they are always showed determined nonlinearity and non-Gaussian, which makes it very difficult to discern the condition of mechanical systems. Aimed at this problem, by combining the bi-spectrum analysis and the MCPP method, a novel approach namely the order bi-spectrum analysis based on the MCPP method is proposed to detect the fault of the gears under the time-varying rotational speed condition. The experimental study indicates that the proposed method can distinguish different conditions of gears.(4) Setting up the suitable searching range of the frequency offset and the frequency slope rate is the key for a successful application of the MCPP method. In order to acquire the suitable searching range of the frequency offset and the frequency slope rate, the singular value decomposition (SVD) method is introduced, and then the MCPP method based on SVD is presented. The proposed method uses the SVD method for preprocessing the vibration signal, and then from the Wigner-ville spectrogram of the preprocessed signal, the scopes of the frequency offset and the frequency slope rate of the vibration signal can be roughly estimated.(5) In the early stage of rotor’s rub-impact failure, the fault feature is very weak, which is always hidden in the strong power frequency signal component. A new signal decomposition method, i.e. the multi-scale chirplet sparse signal decomposition (MCSSD) method which is proposed by combing MCPP and the sparse signal decomposition method is used for early detecting the rub-impact fault of the rotor systems. Then, an early detection method of the rubbing fault for rotors which based on the MCSSD method is proposed. The simulations and the experiments confirmed the validity and the superiority of the proposed method.The MCPP method can use the chirplet atoms whose instantaneous frequency curves are linear straight lines to adaptively match the signal component with the largest energy in a multi-component vibration signal. Meanwhile, the instantaneous frequency of this signal component with the largest energy can be estimated by jointing the piecewise linear frequencies of the chirplets which is used by the MCPP method. Owing to the above ability, the MCPP method is suitable for processing non-stationary signals. Under the time-varying rotational speed condition, the gear vibration signals which measured at equal-time-interval always have low signal-to-noise ratio and possess some properties, such as non-stationarity, nonlinearity and non-Gaussianity, which makes the fault diagnosis more complex. Therefore, as a novel non-stationary signal processing technique, the MCPP method is introduced in and combined with FrFT, multi-scale morphology analysis, bi-spectrum, and SVD. Then, various combination methods are proposed and applied for fault diagnosis of gears under time-varying rotating speed. The results of the simulations and experiments indicate that the combination methods can be effectively applied to the fault diagnosis of gears under the variable rotational speed condition. Furthermore, considering at the early stage of rotor’s rub-impact failure, the rub-impact fault feature is too weak to be detected directly. The MSCCD method which proposed by joining MCPP and the sparse signal decomposition method is applied to detect the feeble fault feature. The simulations and experiments show that the MSCCD method is feasible for characteristics detection of the rub-impact fault in the rotor system. |
PDF Full Text Request