Wind Turbine Bearing Fault Diagnosis Based On Vibration Signal Analysis

The emergence of the world energy crisis, the clean energy has been rapid development, wind power generation is relatively mature development of clean energy, the proportion of the more and more important in the power system. The rapid development of wind power, accelerate the development of wind turbine, wind turbine development trend is large, increasing the capacity of unit. However, because of the wind turbine is relatively harsh operating environment, the failure rate is relatively high, but once the big failure, not only cause damage to the wind turbine itself, the impact of wind power units, the stability of the grid causing some damage. This resulted in the development of wind power a certain extent, so the fault diagnosis of wind turbines on the development of the normal operation of wind turbines and wind are of great significance.(1)This article first to the wind turbine bearing vibration signal detection method are studied; then wavelet packet energy and HHT theory of wind turbines in a wind turbine bearings changed conditions and bearing failure occurs when loose extracted feature vectors; Finally, support vector machines for wind turbine bearing fault types are classified.(2) When the wind wind turbine spindle bearing conditions change, the frequency and energy are the main features of bearing vibration signal. This paper choose the HHT analysis and wavelet packet transform as the bearing vibration signal feature extraction method. HHT analysis is developed in recent years to signal adaptive time-frequency analysis method, time-frequency characteristics of both also have adaptive; The thesis using the HHT analysis of wind power bearing the actual fault signal in the time-frequency analysis, and to extract the signal energy of the IMF. Wavelet packet analysis of the instantaneous time-varying signal has certain adaptability, and has high resolution in high frequency and low frequency at the same time. Thesis on the basis of wavelet packet decomposition, to extract the frequency band energy of rolling bearing, and envelope spectrum analysis was carried out.(3)Using the characteristics of the extraction, expect using different vector feature vector to study the effect on the support vector machine(SVM) classification effect. Due to relatively obvious characteristics of the feature vector construct support vector machine classifier can 100% correct classification, which also shows the support vector machine in a small sample classification does have obvious advantages. Above all, prove that combined with wavelet packet, HHT and support vector machine(SVM) on the diagnosis of wind turbine bearings is meet the requirements of diagnosis, to achieve the desired research purposes.