Research On Fault Characteristic Analysis And Diagnosis Method Of Wind Turbine Gearbox

Along with the wind power technology mature, the market gradually expanded, wind power has become one of the key development energy in the world. According to statistics, the gearbox transmission failure rate is very high in the wind turbine. So it is becoming more and more significant to research the importance and urgency of gearbox fault diagnosis method.The paper mainly researches the vibration signal analysis and fault diagnosis methods about the gearbox of wind turbine, in order to improve the reliability of wind turbine. The major objects are important components of gearbox such as gears and bearings. This work mainly researches the three aspects:(1) The paper studies the gearbox fault mechanism of large-scale wind turbine. Gearbox vibration data from sensors often contains a large number of noise information. This will bring great influence to the fault diagnosis and state evaluation. The paper improves the conventional wavelet threshold de-noising method. Through MATLAB simulation analysis, the fault diagnosis example of gear and bearing are analyzed, and the good results for the improved method are achieved.(2) Fault diagnosis method of gearbox of large wind turbines is studied. An energy analysis solution of vibration signal based on the wavelet packet analysis(WPA) and the vibration signal analysis is given.By calculating the energy of nodes, the corresponding relationship under the various work condition is obtained. The node coefficient of energy changes obviously is reconstructed. And envelope spectrum analysis is carried on these reconstructed node coefficient. The experimental resultsshow envelope spectrum more effective on fault position and fault type.The paper introduces an energy moment parameters, which considering the energy distribution along with the time parameters change bases on the energy analysis solution of WPA. The characteristic parameters related to work condition of time domain and frequency domain signal are extracted,combines with energy moment parameters to build characteristic vector prepared for the subsequent diagnostic analysis. Through the gearbox fault experiment platform to test and analysis data bases on the wind turbine, the effectiveness of the proposed method is verified.(3) The support vector machine(SVM) nuclear parameter optimized by particle swarm optimization algorithm(PSO). This work establishes WPA-PSO-SVM diagnosis model and achieves intelligent pattern recognition under various operating conditions. The SVM normal kernel function adopts a single kernel function, its’ generalization and learn ability is too single; the classification accuracy is not high. This work builds a linear combination of hybrid kernel function which based on the own advantage of RBF kernel function and polynomial kernel function.Through adjusts weight to balance the generalization and learn ability of this algorithm. Finally, through the gearbox fault experiment platform to test and analysis data based on the wind turbine, verifies the effectiveness of the proposed method under the condition of different rotation speed.