Research On The Fault Diagnosis Of Rotating Machinery Based On The Optimal Resonance-based Sparse Signal Decomposition

Rotating machinery is widely applied in modern industry. It generally operates under tough working environment which results in the failure inevitably. Unp redicted failures in such machinery will lead to breakdown, economic losses and, even worse, human casualties. Therefore, it is very important and necessary to detect the early fault in rotating machinery. The analysis method based on the vibrational signa l is the main fault diagnosis technique. This technique firstly collects the vibrational signal of rotating machinery, and then extracts the fault characteristic signal from the collected signal and finally detects the fault in rotating machinery.The resonance-based sparse signal decomposition(RSSD) method is proposed by Selesnick and it is an enable-sparse decomposition method. Different from tradition al time-frequency methods, the RSSD method uses Tunable- Q Wavelet Transform(TQWT) to obtain the sparse representation of the complex signal according to different values of the quality factor Q. Next, the high resonance component and the low resonance component can be obtained in a non-line way by using the morphological component analysis method. However, this method is sensitive to noise. The decomposition effect of the RSSD method is not very well when the strong background noise exists. Also, when there are compound faults in rotating machinery, the week fault characteristic signal is always submerged by the strong one and the noise which cause misdiagnosis and miss-diagnosis.Supported by the project of Natural Science Foundation of China, which named as “Research on the resonance-based sparse signal decomposition method and its application in mechanical fault diagnosis(Project Approval Number: 51275161)” and the independent research project of the state key laboratory of advanced design and manufacturing for vehicle body of Hunan University, which name d as “Research on the early fault diagnosis and residual life prediction techniques of key automotive components and parts(Project Serial Number: 71375004)”, aiming at these problems of the RSSD method above, three novel fault diagnosis methods are proposed by combing the RSSD method with the genetic algorithm, the energy operator demodulation method, the stepwise optimization method and the comb filtering method respectively. The proposed methods are successfully applied to the fault diagnosis of rotating machinery.The main researches and the acquired innovative achievements in the thesis are as follows(1) The effectiveness of envelope demodulation analysis is always not very well when the bearing signal is directly demodulated. To address this issue, a novel method based on the resonance-based sparse signal decomposition and tunable-Q wavelet reconstruction was proposed and applied to the bearing fault diagnosis in this paper. In this method, the vibration signal of a rolling bearing is decomposed into the high-resonance component and low-resonance component by the resonance-based sparse signal decomposition method. Then, the low-resonance component is further decomposed into a set of sub-signals by the tunable-Q wavelet method and the proper signal is reconstructed from some selected sub-signals. Finally, the proper signal is analyzed by Hilbert envelope method, the cycle of the periodic impulse component can be acquired and the faults of the rolling bearing can be diagnosed. Simulation and application example show that the proposed method is effective in extracting impulse signal from rolling bearings.(2) The separating effect of RSSD is closely related to the decomposition parameters which can not be selected adaptively. Aiming at that problem, a new method for the fault diagnosis of rotating machinery is proposed based on the energy operator demodulating of optimal resonance components. In the proposed method, the smoothness index and kurtosis are introduced and composed into the evaluation index. And then, the genetic algorithm is used to optimize decomposition parameters in the RSSD method. Next, the optimal high resonance component and the optimal low resonance component can be obtained by using the RSSD method with decomposition parameters. Finally, both optimal resonance components are subject to the energy operator demodulation. Though observing the spectra of instantaneous amplitudes of both optimal resonance components, the faults of the gear and the bearing can be detected effectively.(3) The comb filter can be used to extract the periodic impulse signal and thus it is a useful tool to detect the fault in rotating machine. However, it is a challenge to find the proper fundamental frequency of the comb filter because of the lack of priori knowledge. That is, the measured vibrational signal can not be directly filtered by the comb filter whose fundamental frequency is the fault characteristic frequency. Filtering the measured signal blindly is a time-consuming process. What’s more, the filtering results will be interfered by noises easily. In this thesis, a new approach, namely the Optimal Resonance-based Signal Sparse Decomposition with Comb Filter(ORSSD-CF) method, is proposed and applied to the compound faults diagnosis in rotating machine. Aiming at the problem that the optimization process of decomposition parameters by using the genetic algorithm is time-consuming, the stepwise optimization approach is introduced to obtain the optimal decomposition parameters of the RSSD method. The effectiveness and efficiency of the propose method are verified by application examples.By combining the RSSD method with the tunable-Q wavelet reconstruction method, the genetic algorithm, the energy operator demodulation, the stepwise optimization strategy and comb filtering respectively, three novel fault diagnosis methods are proposed in the thesis and successfully applied to the fault diagnosis of rotating machinery. The simulation and application examples show that the proposed methods have a good application prospect in the fault diagnosis of rotating machinery.