Research On Fault Diagnosis Of Wind Gearbox Based On Feature Fusion And XGBoost Algorithm

Wind energy has the advantages of no pollution and low cost,which makes wind energy become one of the most potential and most dynamic clean energy sources.The installed capacity of wind farm has a large growth space and the cost decreases rapidly.Gearbox is one of the important transmission components of wind turbine.Due to the bad operating environment,complex structure and with the increase of running time the components are corroded,gradually aging and other reasons,the gearbox would fail and bring huge economic losses.Therefore,the fault diagnosis of gearbox has always been one of the important topics of scientists.In this paper,an experimental platform for fault diagnosis of gearbox is set up,and the bearings and gears with more fault frequency are taken as the research object,the vibration data under different operating conditions are collected,and the fault diagnosis model is established to ensure the operation reliability of wind turbine.The main work of this paper is as follows:(1)This paper briefly introduces the development situation of wind power industry,the fault diagnosis technology research status of wind turbine at home and abroad and the basic structure of gearbox.The fault type of the rolling bearing,the fault type of the gear in the gearbox and the frequency domain form of the fault of bearing and gear are simply explained;(2)This paper introduces the feature extraction method of wind gearbox vibration data in time domain,frequency domain and time frequency domain,and proposes the Improved Particle Swarm Optimization(IPSO).The superparameters of VMD(Variational ModeDecomposition)are optimized by IPSO.The vibration signal is analyzed into several modes according to the central frequency,and then the average energy of each modal power spectrum is calculated as the eigenvalue,finally the model of IPSO-VMD-FE(Improved Particle Swarm Optimization and Variational Mode Decomposition and Frequency Energy)is proposed;(3)In order to solve the problem that there may be fixed faults in rolling bearings,a feature fusion algorithm based on CART(Classification And Regression Tree)decision tree is proposed.Firstly,the importance of the features obtained from each sensor is calculated by using the CART decision tree algorithm,and the weights are calculated according to the feature importance,and then the feature fusion of the multi-sensor feature matrix is carried out.Finally,the fusion feature matrix is input into XGBoost(eXtreme Gradient Boosting)algorithm for fault diagnosis.The fault diagnosis model of the CART-XGBost is developed,and it is applied to the fault diagnosis of the bearing of the fan gear box.(4)Aiming at the compound fault problem of gearbox gear,the DCNN(Deep Convolutional Neural Network)is used to extract the characteristic matrix of vibration signal,and then input into XGBoost algorithm for fault diagnosis.The DCNN-XGBoost fault diagnosis model is put forward and applied to the fault diagnosis of fan gearbox gear.