Sparse Representation Of Signal Transients And Its Application In Machinery Fault Feature Extraction

The operation status of the key equipments in machinery industry, transportation, energy, metallurgy, petrochemical and national defense industry has a close relationship with the normal operation and healthy development of the national economy. The key components of the equipment, running in such complex and severe conditions as heavy load, elevated temperature, would inevitably experience performance degradation and failure. To diagnose the failures of the components, guarantee the safety operation of the equipments and avoid catastrophic accidents, it is of great significance to extract the fault features of the key components.This thesis is funded by "Signal Transients Extraction under the Frame of Sparsity and Its Application in Rotating Machine Fault Prognosis"(No. 51375322) supported by the Natural Science Foundation of China. With the aim of fault feature extraction of rolling bearing and gearbox, a novel feature extraction method based on sparse representation theory is proposed in this thesis. The theoretical research and application research are studied in depth, respectively.The failure types and the characteristics of the fault vibration signals of the bearing and the gear, which has localized fault, were analyzed respectively. The transient nature and sparsity nature of the vibration signal is revealed. The key issues of the signal sparse representation are also revealed in this thesis. All the works provides the theoretical foundation of the research in this thesis.The construction of over-complete representation dictionary is one of the key issues of the signal sparse representation. According to the characteristics of the fault vibration signal of rolling bearing and gearbox, a method of constructing the dictionary adapted to the signal is proposed in this thesis. In this method, the correlation filtering is firstly applied for the optimal wavelet atom. Then the dictionary can be constructed by translating the time parameter of the atom. By this method, the similarity of each atom in the dictionary and the signal being analyzed is guaranteed.Solving the objective function in the representation model is another key issue in signal sparse representation. A novel transient sparse representation method for optimizing the data fidelity in the objective function is proposed in this thesis. The effectiveness of the proposed method is then verified by numerical simulation. Moreover, the effects of the parameters-the Lagrange multiplier and the penalty parameter-on the representation result, and the existence and uniqueness of the solutions are also investigated. Finally, the effectiveness of the proposed method for fault feature extraction is verified through the applications in fault rolling bearing and fault gearbox, respectively.A novel transient sparse representation method to optimize the penalty term in the objective function is proposed in this thesis. A quadratic function of the general form is designed in this thesis to ensure the realization of the algorithm. The effectiveness of the proposed method is verified by the numerical simulation. In order to test the noise tolerance of the proposed method, the simulation tests with different noise amplitudes are also investigated. Finally, the effectiveness of the proposed method for fault feature extraction is verified through the applications in fault rolling bearing and fault gearbox, respectively.In this thesis, through the research on the transient feature extraction based on sparse representation theory, it is confirmed that both the proposed methods are effective in detecting the fault feature of machinery components, which has theoretical and practical value for fault extraction and diagnosis of rotating machinery.