Research On Vibration Monitoring And Fault Diagnosis System Of Wind Turbines

With the establishment and implementation of large-scale wind power development plan in many countries of the world, the wind industry grows quite rapidly in several years. But after that, it also has some growing issues. Although the design and manufacturing of the wind turbines has been gradually perfect, its maintenance becomes a great challenge which is now a key problem waited for a solution.The wind turbines are often located in wind gaps such as the mountains, beaches and islands. Due to the location, the impact of the strong wind and extreme variation of temperature, always cause its malfunction. Statistics shows that the failure of gearbox, transmission chain and generator causes the longest downtime and the biggest economic losses. Currently most large-scale wind turbines have their own Supervisor Control And Data Acquisition (SCADA) system, which is used for data acquisition, parameter control, equipment control and alarm, etc. It supplies strong technological flat for the operational reliability of wind power farms, but it lacks effective monitoring and accurate diagnosis function of transmission chain. Foreign condition monitoring and fault diagnosis products are very expensive, and lack of unified industry standards about alarm threshold. And most units’s condition needs expert’s analysis with practical experience. However there is no shaped product in our country. Based on the above, this paper selects the research on the vibration monitoring and fault diagnosis of wind transmission system as the keystone. The main research work and conclusion are as follows:(1) It introduced the composition structure and operational principle of wind turbines, and processed a statistical analysis on the common faults of the wind turbines. Statistics show that the electrical system and control system cause the higher failure rate, but the gearbox and generator fault cause the longest downtime and the large economic losses. This paper analyzes emphatically the fault mechanism and its expression form, typical fault and signal characteristics of the key parts, such as gears, bearings, shaft system.(2) In the feature extraction of the wind turbine vibration signal, the time domain analytical method and frequency domain analytical method of vibration signal were introduced respectively. The analysis results of some living examples indicate that frequency domain analysis can effectively identify the fault frequency in vibration signals from the gearbox. In this paper, the author provided detailed information on the wavelet packet energy spectrum analysis technology, extracted the energy feature of the gearbox vibration signal by adopting the wavelet packet, and took the energy feature vector as the input of the SVM model for fault diagnosis. All of these will lay a solid foundation for subsequent SVM to gearbox fault pattern recognition.(3) As the gearbox vibration signal is nonlinear, has low signal-to-noise and small samples, the author proposed a gearbox fault diagnosis model according to the characteristics of it, and it was based on wavelet packet, genetic algorithm and support vector machine(SVM). Firstly, the gearbox vibration signal was de-noised and preprocessed by the wavelet packet. Then the feature of band-energy was extracted by the wavelet packet decomposition and reconstructed as the input vector of the SVM classifier. Genetic algorithm was used to achieve automatic and optimal choose of the parameter "g" of radial basis kernel function and penalty parameter "C"Finally, the effectiveness and superiority of the proposed model was verified through a series of instance analysis of the gearbox fault diagnosis.(4) This paper introduced the feature extraction and fault diagnosis method of vibration signals. On account of this method, the author applied the Object-orientated technology, adopted C#and SQL Server2008database, preliminary developed a wind turbine vibration monitoring and fault diagnosis system. The system module includes user management, device management, condition monitoring, spectrum analysis fault diagnosis, SVM fault diagnosis and database management. Each module is subdivided into several relatively independent subsystems. And the function of preprocess (such as filtering, average value of function to nought), feature extraction and fault diagnosis for the vibration signals can be achieved, this will make great help to the wind turbine condition monitoring and diagnosis and decision.