Research On Fault Diagnosis Of Wind Turbine Gearbox Based On Vibration And Stator Current

In the past twenty years, China’s wind power industry has rapidly developed, but the frequent failure of the wind turbine transmission parts of the wind turbine system has brought great challenge which seriously affected operation reliability, increased maintenance costs and other aspects. Gearbox as the key equipment of the main drive system of turbine, the structure is very complicated, and it is a complex subsystem with nonlinear and strong coupling characteristics. It can effectively avoid the occurrence of accidents and reduce the cost of operation and maintenance, which is of great significance to ensure the economic benefit. And gear wear fault with incremental, compared with the sudden faults, it is easy to be overlooked and harm is great. The current detection method is generally to measure wear, oil detection and iron spectrum analysis method, has high cost and more time-consuming, and the result does not have universality. In view of the above problems, this paper based on numerical analysis and combined with experiments, respectively, from the two aspects of vibration signal and the stator current signal studying the wind turbine gearbox wear fault. The method of the underlying process model(latent process mode, LPM) dynamic parametric modeling was applied, by the description of the dynamic signal timing characteristics and then the high order time series data were decomposition and parameter was estimated that analyzing trend and selecting feature. The main work of this paper is as follows:(1) Summarized the structure characteristics and the fault types of the wind turbine gearbox, and analyzing the failure mechanism, the characteristics of the gearbox vibration fault, the fault detection principle of the vibration and current signal. Aiming at turbine gear early wear fault characteristics are difficult to extract and quantify, the feature extraction of LPM dynamic modeling method and multi-scale entropy quantitative method was used to studying the wear fault of the gearbox.(2) According to the characteristics of non-stationary and multi-component of vibration signal, the LPM feature extraction method is studied, and building classification detection based on support vector machine(SVM) model, meanwhile, the parameters of LPM algorithm was optimized. The feasibility of the method is verified by using the whole life data of the intelligent maintenance center of the United States, and through constructing the experimental platform of simulation experiments verify the validity of the underlying process model.(3) Researching characteristic decomposition and quantitative analysis method based on stator current. Due to the stator current signal detection has the advantage of easily access, low cost and high signal-to-noise ratio, studying the non-impact signal in the stator current mode characteristics based on LPM algorithm, and combined with multi-scale sample entropy quantifies the fault feature vector to verify the effectiveness. By setting up the wear fault diagnosis platform of the wind turbine gearbox, the accuracy and reliability of the model are proved.