Fault Diagnosis Of Wind Turbine Gearbox Based On Improved Empirical Wavelet Transform

Wind energy is a kind of clean energy with great utilization potential.Wind turbine is an essential equipment for wind power generation.In recent years,the total installed capacity of wind turbine has gradually increased.Because most wind turbines are installed in the wild,they are prone to failure under the influence of the external environment.Gearbox is a key component in wind turbine with high failure rate and high maintenance cost,Therefore,it is of great significance to study the fault diagnosis of gearbox for the normal operation of wind turbine.The research contents of this paper are as follows:Firstly,the common fault types of gear box gears and the common diagnosis methods of gears are studied,the advantages and disadvantages are analyzed,and vibration signals are selected to study gear faults.The main sources and mechanism of gear vibration are analyzed,and the common parameters and the main components of frequency spectrum in gear vibration are introduced.According to the mathematical model of gear vibration,the influence of different faults on the mathematical model is analyzed,thus the main characteristics of the frequency spectrum when different faults occur on gears are defined.Then,according to the main structure and operation principle of doubly-fed wind turbine,the mechanical part and control part of the wind turbine experimental platform are built by selecting the hardware according to the corresponding principle.The experimental platform can realize variable speed and constant frequency operation.The acceleration vibration sensor and signal acquisition instrument are selected.The vibration signals of normal gear and broken gear under different rotating speeds are obtained through experiments.Fourier spectrum analysis is carried out on the vibration signals of the experiments according to the spectrum characteristics studied in chapter 2.The difference between normal gear and fault gear is compared,and the correctness of spectrum characteristics is verified.The difference and principle between empirical mode decomposition(EMD)and empirical wavelet transform(EWT)in time-frequency domain signal processing methods are studied.Simulation signals are constructed according to gearbox signal types.The decomposition results of the two methods on simulation signals are compared,and the advantages of empirical wavelet transform are verified.The scale space theory is introduced to optimize the frequency band division in the EWT decomposition process.The results of frequency band division of vibration signals by Means method,probability method,Ostu method,K-means clustering method and Empirical law method are compared.The threshold value of frequency band division is determined by using Empirical law method,and the optimized empirical wavelet transform algorithm(SSEWT)is determined,which provides theoretical basis for vibration signal analysis of the test bench.Finally,the vibration signal extracted from the test bench is decomposed by the improved empirical wavelet transform(SSEWT).According to kurtosis index,the impact components of normal gear and broken gear are screened out.Hilbert envelope spectrum analysis is carried out.Compared with direct Fourier spectrum analysis,Hilbert envelope spectrum better distinguishes normal gear and broken gear,and fault diagnosis of broken gear is realized.