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

With the rapid development of the wind power industry,it is very necessary to monitor the working conditions of gearboxes and obtain gearbox fault information in time to ensure the operation safety and maintenance of wind turbines.The gearbox working environment is complex and the amount of data is large.It is difficult to implement traditional fault diagnosis methods.This paper studies the deep migration learning wind turbine gearbox fault diagnosis and classification method,and uses the fault diagnosis knowledge obtained after the gearbox fault public data training to transfer to the wind turbine gearbox fault diagnosis application.By fine-tuning the parameters,the original sample data is obtained.The similar migration model data can accurately diagnose and classify gearbox faults.Deep learning uses feature learning algorithms and hierarchical feature extraction data.A large amount of data is needed to understand the potential data methods.Transfer learning is used in applications where the original sample data is big data and similar to the data of the migration model.Analysis of the fault mechanism and vibration characteristics of the gearbox system of wind turbines.For gearbox pitting,wear and broken teeth faults,the migration learning method is determined to realize the classification of wind turbine gearbox faults;because the vibration signal interference factors of the gearbox are many and present Non-linear and non-stationary,so the SVD matrix decomposition method is used to reduce the noise of the original vibration signal,and the continuous wavelet change method is used to perform time-frequency transformation of the one-dimensional vibration signal as a model database.Through the pre-training data set,create a deep convolutional neural network model,and use the migration learning method to transfer its parameters to the wind turbine gearbox fault diagnosis model.Through regularization,network sparsity,and layer data distribution feature optimization,wind turbines are realized The gearbox fault diagnosis model is optimized and combined with the multi-label classification method,and the feasibility of the deep transfer learning wind turbine gearbox fault diagnosis method is verified.Through method verification of the gearbox fault data,the results show that the fault diagnosis method based on deep migration learning can realize the fault migration of the original vibration signal of different monitoring equipment compared with the traditional intelligent diagnosis method,identify the gearbox fault and obtain a higher value.The accuracy of the diagnosis.