Research On Fault Diagnosis Method Of Wind Turbine Gearbox Based On Deep Learning

Gearbox is the indispensable transmission device of large scale wind turbine,concerning the operation of the whole wind energy system.Wind turbines work in bad environment like desert,wild and isolated island,which leads to high failure rate of gearbox.Research on fault diagnosis of wind turbine gearbox has great significance in ensuring the proper operation and the economic benefits of wind farm.Based on the operation data,this paper uses deep learning in the gearbox fault diagnosis of wind turbine.This main content of this paper includes:(1)Based on the analysis of the fault forms and mechanism of planetary gearbox,the time-domain and frequency-domain characteristic of vibration signals under typical fault conditions are studied.(2)An improved nonlinear threshold function of wavelet de-noising algorithm is proposed,which solves the problems of complex components,low signal-to-noise ratio in vibration signal analysis.The non-linear curve replaces the linear curve of the traditional threshold function is able to avoid the discontinuity of the hard threshold function at the threshold and overcome the constant deviation of the soft threshold function simultaneously.(3)Based on empirical mode decomposition(EEMD)and time series analysis,a feature extraction strategy of gearbox vibration signal is proposed.The proposed feature extraction strategy solves modal overlapping and false components in the traditional EMD method.The time series model of signal is judged by partial correlation function,and the model parameters of high frequency components are extracted as its feature vector.(4)A gearbox fault diagnosis algorithm is proposed by combining the deep belief network with genetic algorithm.A deep belief network diagnosis framework is constructed,and genetic algorithm is used to optimize the network parameters.By comparing the simulation results with traditional BPNN and SVM,the results show that the proposed method is effective and accurate.The methods and strategies proposed in this paper provide a new solution for the technical system of gearbox fault diagnosis of wind turbine.