Research On Fault Diagnosis Of Wind Turbine Based On Multi-scale Chirplet Sparse Signal Decomposition

Wind power utilization and wind turbine capacity is increasing rapidly with the steady declining of fossil fuels and increasing environmental pressures. Harsh natural environment and complex load lead to a higher failure rate of large wind turbine. The research on condition monitoring and fault diagnosis technology of wind turbines is of essential importance to the safety and stability of wind turbines.Condition monitoring and fault diagnosis of large wind turbines has two features: First, it is required to establish the corresponding system model of condition monitoring and fault diagnosis in accordance with the structure and failure features of wind turbines,Second, the monitoring signals measured in a wind tribune under variable speed and load mode are usually non-stationary signals. This thesis, funded by project“Sparse Signal Decomposition Based on Multi-scale Chirplet and Its Application to Mechanical Fault Diagnosis”(Project’s Serial Number: 50875078) supported by National Natural Science Foundation of China, and project“Research on condition monitoring and fault diagnosis of large wind turbine”(Project’s Serial Number: 20090161110006) supported by The National High Technology Research and Development of China 863 Program, researched the fault diagnosis problems of large wind turbines.The main research works include:(1) According to the structure and failure features of wind turbine, the system model of condition monitoring and fault diagnosis of wind turbines is established and a brief introduction of the fault diagnosis technology of large wind turbines is presented.(2) In order to extract the fault features from non-stationary signals of a large wind turbine under time-varying rotational speed condition effectively, the sparse signal decomposition method based on multi-scale chirplet is introduced in the fault diagnosis of wind turbines. The method, with good time frequency gathering property, decomposition adaptability, expression sparsity and strong noise immunity, is especially suitable for the analysis of non-stationary signals.(3) Aiming at the problem that the modulation sidebands of vibration signals of the gearbox of large wind turbines with variable rotating speed are difficult to identify, an order analysis method based on multi-scale chirplet and sparse signal decomposition is proposed. In the proposed method, a vibration signal is decomposed to obtain the meshing component and the modulation sidebands by using the sparse signal decomposition based on multi-scale chirplet, and simultaneously, their instantaneous frequencies are obtained. Then the rotating speed signal can be got by software method without the tachometer. Based on the obtained rotating speed, the even angle resampling on the sum of the meshing component and the modulation sideband components are carried out, so the order spectrum of the resampled signals can be obtained, which can be used in the fault diagnosis of the gearbox. Simulation and practical application examples show that the proposed method has a strong noise immunity and is better than the traditional order method in identifying the modulation sidebands.(4) Aiming at the problem that the fault characteristic frequency of the roller bearing of large wind turbines with variable rotating speed are difficult to identify, a new generalized demodulation method based on multi-scale chirplet and sparse signal decomposition is proposed. Firstly,the multi-component signal is decomposed by using of the sparse signal decomposition based on multi-scale chirplet and the mono-component signals and its phase functions are obtained. Then,based on the obtained phase functions of mono-component signals, the generalized demodulation method is used to transform the original non-stationary signals into stationary signals. When the rotating speed of a bearing is varying with time, the bearing fault vibration signals are non-stationary signals. In the proposed method, the non-stationary enveloping signals of bearing fault vibration signals are transformed into stationary signals by using of the generalized demodulation method based on multi-scale chirplet and sparse signal decomposition. According to the relationships between the frequencies of enveloping signals after generalized demodulation and the rotational frequency, faults of the bearing can be identified. Simulation and practical application examples verify that the proposed method is more effective than the conventional enveloping spectral analysis technique in extracting the features of roller bearing fault vibration signals.In this thesis, the system model of condition monitoring and fault diagnosis of a large wind turbines is established, a suitable method for dealing with the multi-component non-stationary signals by sparse signal decomposition based on multi-scale chirplet is introduced. Based the method, the thesis proposes an order analysis method and a generalized demodulation method based on multi-scale chirplet and sparse signal decomposition. These methods can be effectively applied to the fault diagnosis of gears and roller bearings of large wind turbines under time-varying rotational speed condition.