Research On Fault Analysis And Diagnosis Of Key Components Of Wind Turbines In Complex Working Condition

In recent years, the global wind power industry has developed rapidly, however,the costs of wind turbine maintenance has been high all the time, which seriously restricted the development of the wind power industry. Wind turbines work environment was extremely complex, affected by fluctuations in wind speed and load changes, the vibration signals had non-stationary, nonlinear, time-varying characteristics. The traditional fault diagnosis method, ignored the impact on wind turbines’ dynamic characteristics which made by the wind turbines’ complicated working conditions. To solve this problem, based on wind generator fault status change process and multi-sensor fusion, we proposed a fault diagnosis and quantitative analysis method of wind turbines in complex working condition.Aim at the problem of fault diagnosis and quantitative analysis of wind turbines, based on Hilbert-Huang transform and information entropy, a method named Hilbert space feature entropy was proposed. First, divided the signal into several time-frequency areas by Hilbert-Huang transform, then did singular value decomposition to the energy distribution matrix of signals, last the Hilbert space feature entropy could be calculated according to the definition of entropy. In addition, a improved method about the Hilbert-Huang Transform was propsed to sovle the problem of end effects.To validate the algorithm, unbalance-rubbing coupling faults experiment and looseness-rubbing coupling faults experiment was designed by rotor test rig. Collected the rotor fault signal at different speeds, then analysed the test data by Hilbert space feature entropy, the type and the extent of fault could be depicted by the curve of entropy changed as rotor speed.Aim at the problen of fault diagnosis of wind turbines in complex working conditions, first analysed the vibration of drive systems of wind turbines in complex working conditions. Analysed the impact of wind turbines on its vibration control strategies, and then analysed the vibration of wind turbines’ bearings in different wind speeds conditions and different load conditions. The signals of bearings were studied from different viewpoints such as the time domain, frequency domain, time-frequency domain, Hilbert space feature entropy. The rule of wind turbines’ vibration changed with wind speed and load was summed up.On this basis, the bearings fault and gearbox fault were studied. The vibration model of wind turbine bearings in different wind speed was given here, the vibration signal of normal bearing and fault bearing was analysed and compared. Analysed the signal acquired in different wind speed by Hibert space feature entropy, the bearing fault could be diagnosed intuitively by the curve of entropy changed as rotor speed. Then analysed the bearings monitoring data of a month by Hilbert space feature entropy, the result showed that the method could effectively and quantitatively depict the bearing fault processes, and the bearing fault could be diagnosed early according to the mutation point of the entropy value. The method for calculating characteristic frequency of gear was given here, analysed the mesh frequency of signal of normal gearbox and fault gearbox, the results showed that this method can effectively analyze gearbox fault, but could not reflect the degree of fault, and the result of the diagnosis was not intuitive, the process was more complicated. To reflect the operational status of the gearbox more comprehensive, fusion analysed the signal of gearbox acquired in multi measuring point, multi-speed, multi fault state by Hilbert space feature entropy. Then the entropy plane of gearbox vibration signal could be calculated to depict the Hilbert space feature entropy of gearbox vibration signal changed with the measuring point and rotation speed. The gearbox fault could be diagnosed intuitively by contrasting the entropy of normal gearbox and fault gearbox. Compared the entropy plane of fault gearbox,which showed that the method could quantitatively depict the fault state of gearbox.