The Fault Dectection Method Research Of The Key Components Of Wind Turbine Drive System

Under the background of "energy crisis and ecological crisis" which is caused by resources shortage of the oil, coal and others, wind power as a renewable environmentally friendly energy is widespread concern in today’s society. With installed capacity of fan increases year by year, highlighted the importance of the maintenance services of fault detection and repair year by year. Due to the fan installed on the high tower, once the fan drive system fails, the maintenance cost will be high. In this paper, fault detection methods of bearing and gear are studied, the main work includes the following aspects:(1)In view of the traditional method of denoising method has neglected the influence of impulse, a new method for rolling bearing fault based on mathematical morphological filter and improved wavelet is presented. Firstly, the morphological filter has a certain inhibition effect on the interference pulse, so it can be used to remove the impulse disturbance of rolling fault vibration signal. Due to the defects of the traditional threshold function, the improved wavelet denoising method is proposed to remove the white noise. Then Empirical mode decomposition method is used to decompose the signal after de-noising. Finally, the characteristic frequency of the intrinsic mode function is compared with the actual fault characteristic frequency, and the conclusion is drawn.(2)When the rolling bearing partial damage or defects, its vibration signal are showed nonstationary and nonlinear properties, modulation and other characteristics and the extracted characteristic parameters can’t very well for fault detection. Probabilistic neural network model based on characteristic parameters were proposed. Firstly, empirical mode decomposition method is used to extract the intrinsic mode function, which is formed by the signal adaptive decomposition. Then, the multi feature parameters are extracted. Finally, the probabilistic neural network model is applied to the rolling bearing fault detection of wind power generator in order to improve the accuracy of rolling bearing fault detection.(3)To improve the classification accuracy of the wind turbine gear fault detection, we put forward a method of fault detection based on SVM trained by improved shuffled frog leaping algorithm(ISFLA-SVM). Because the parameter selection for penalty factor and kernel function in SVM have a great effect on the classification accuracy, we should use the improved shuffled frog leaping algorithm for SVM parameters optimization, use the optimized parameters to train machine, then trained model is applied to fault detection, then three groups of data in UCI are used for performance evaluation, finally ISFLA-SVM model will be applied to the gear fault detection of wind turbine gearbox.