| Wind turbine is easy to be affected by wind sand,wind speed and uncertain load,which leads to frequent failure of wind turbine components.In the wind turbine,the planetary gear train is an important part of the speed-increasing gearbox.The operating efficiency and reliability of the wind turbine are directly related to the working state of the planetary gear train.Therefore,it is of great significance to study the fault diagnosis method of the wind turbine planetary gear train for ensuring the continuous and safe operation of the equipment.But in the real industrial scene,the fault data is limited and highly unbalanced,so the accuracy and stability of diagnosis are limited.In order to solve these problems effectively,this paper proposes a Deep learning algorithm based on the combination of Deep Convolutional Generative Adversarial Net(DCGAN)and data enhanced Stacked Autoencoders(DESAE).Convolutional generative warfare network is a new branch of the research field of deep learning,which is effective in solving the problem of small and unbalanced samples.It can learn the characteristics of the original samples,and then simulate the generation of samples similar to the distribution of the original sample characteristics,to achieve data expansion,thus improving the accuracy and stability of diagnosis.The main research contents and achievements are as follows:1.A fault diagnosis model based on DESAE is proposed to solve the problem that fault samples are very few and the accuracy and stability of fault diagnosis are limited.First,the input signal is passed through the encoder,and the encoding area maps it into the hidden layer expression,and extracts the high-dimensional features of the data,then the decoder decodes the extracted high-dimensional features from the hidden layer to the original input data,thus completing the data reconstruction process.The results show that the recognition rate of DESAW model is 95.01% on the simulation data set and 93.14% on the gear data set.2.To solve the problem that the quality and recognition rate of DESAE model generated data are not very high,we combine DCGAN network with DESAE network model,and replace the generation module of DESAE network with DCGAN network,a fault diagnosis model based on DCGAN-SAE is proposed.First,the generator in the GCGAN network generates samples from the random noise that can make the discriminator unable to distinguish the true from the false,so as to expand the data,then the expanded data are input into the classification module of DCGAN-SAE network model for classification recognition.The results show that the recognition rate of DCGAN-SAE model is 99.57% on the simulation data set and 98.28% on the gear data set.3.In order to further verify the validity of the fault diagnosis method based on convolution generation antagonism stack self-compiler in the data set of wind power planetary gear train,the experimental data of wind power planetary gear train collected in the laboratory are taken as the research object,DCGAN-SAE model is applied to fault diagnosis of fault data of wind power planetary gear trains.The results show that the highest recognition rate is 98.51% on the fault data set of wind turbine planetary gear train,while the accuracy rates of DESAE and SAE are 93.24% and 76.43%,respectively,compared with DESAE and SAE,the accuracy of DCGAN-SAE model is improved by 5.27% and 22.08%. |
