Research On Fault Feature Extraction And Intelligent Classification Of Wind Turbine Gearbox Based On Depth Learning

The normal operation of wind turbine gearbox directly affects the wind field personnel and equipment safety and economic benefits,so condition monitoring and fault diagnosis of the wind turbine gearbox is of great significance.In this paper,the shortcomings of the traditional diagnostic methods of gearbox are researched,and then two new algorithms are proposed based on the empirical mode decomposition(EMD)method,energy entropy theory,the feature selection method based on the feature distance evaluation,the deep learning and the particle swarm optimization support vector machine.It is proved by diagnosing the gear and rolling bearing that the new algorithms are more suitable for the fault feature extraction and intelligent diagnosis of wind turbine gearbox.Firstly,the basic structure and working principle of wind turbine are researched.The basic structure of the wind turbine gearbox is researched,the characteristic frequency of each part of the gearbox is calculated,and the typical faults and signal characteristics of each part of the gearbox are analyzed.Secondly,the traditional methods of fault feature extraction for wind turbine gearbox are researched,including time domain analysis,frequency domain analysis and time-frequency domain analysis;Several commonly used time-frequency analysis methods are compared,and the EMD method and energy entropy theory are studied;the principle of denoising auto-encoder is researched,extracted feature of the wind turbine gearbox vibration signal using deep learning.Again,according to the traditional gear fault diagnosis of wind turbine gearbox in the artificial frequency domain feature extraction low efficiency and low accuracy,this paper proposes a method of fault diagnosis which combines the deep learning extraction in frequency domain features with the artificial extraction in time domain features.Firstly,mathematical statistics and deep learning are used to extract fault features.Then,the feature vector extracted from the deep learning is combined with the time domain feature extracted from the artificial method to form the augmented feature vector.Then put the extracted feature input the particle swarm support vector machine for feature classification.Finally,the diagnosis results of the multi stage gear transmission system show that this method can effectively improve the accuracy of diagnosis.Then,as to the vibration signal of the rolling bearing of the wind turbine gearbox is non-stationary and nonlinear;when the extracted signal contains a large number of redundant feature vectors,the classification accuracy will be reduced;and it is difficult to obtain a large number of fault data,an intelligent fault diagnosis method for rolling bearing based on EMD,feature selection and deep learning is proposed.Firstly,EMD algorithm is used to decompose the non-stationary rolling bearing vibration signal into a number of stationary intrinsic mode functions;then,the EMD energy entropy of the vibration signal and the statistical characteristics of the time domain and frequency domain of each IMF component are extracted to form the feature vector;then,the feature selection algorithm based on the feature distance evaluation is used to select the features;finally,the sensitive features are selected into the deep learning network for fine tuning,and the feature learning and state recognition classification model of rolling bearing is obtained.The experimental results show that the proposed method can effectively extract the fault features and improve the accuracy of diagnosis.Finally,by using the advantage of LabVIEW simple and efficient graphical programming and the powerful signal processing ability of MATLAB,a wind turbine gearbox condition monitoring and fault diagnosis system is designed,which can realize vibration signal acquisition,time domain analysis,frequency spectrum analysis and EMD time-frequency analysis and other functions.Through the analysis of the experimental fault signal,it is proved that the fault diagnosis system can realize the fault diagnosis.