Research On Fault Diagnosis For Wind Turbine Gearbox Based On Fuzzy Neural Network And Evidence Theory

In current stage,wind power is the most promising renewable energy.In order to improve the reliability of wind turbines and the generation of wind power,the advanced operation and maintenance technologies are required.Gearbox is one of the key equipment of doubly-fed and semi-direct-drive wind turbines.Gearbox bears the various loads from the extreme environment and is prone to failure.The failure of gearbox will affect the performance of wind turbines,reduce the power generation and bring great challenges to the operation and maintenance of wind turbines.Therefore,it is significant to study the methods of vibration monitoring and fault diagnosis for wind turbine gearbox.Taking the wind turbine gearbox as the object,we study the gearbox fault diagnosis method.The main work includes:(1)The fault characteristics of wind turbine gearbox are analyzed.The fault types related to bearings,gears and shafts are studied.The spectrograms of faults are analyzed and the fault characteristics are summarized.These work provides a scientific basis for intelligent fault diagnosis.(2)To decrease the noises in the gearbox vibration signals under various loads?an adaptive lifting wavelet denoising method is proposed.This method can effectively reduce the noise effect and improve the signal-to-noise ratio of vibration signals.The characteristics of vibration are extracted according to time domain,frequency domain and wavelet packet energy.(3)In view of the strong impact loads of wind turbine,three fault diagnosis models based on PSO-FNN are established according to time domain,frequency domain and wavelet packet energy.The models are verified with the test samples to realize the accurate diagnosis of gearbox high-speed shaft and intermediate shaft faults.The experience results illustrate that the rightness of the three models.(4)The diagnosis model of fused gearbox fault is proposed based on the D-S evidence theory.The model makes full use of the advantages of each diagnosis model.It improves the discrimination for gearbox fault diagnosis.Moreover,it has better fault tolerance and error correction than a single diagnosis model.