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

Energy is not only the support of a country’s national economy,but also an important pillar for the development of a nation.In order to alleviate the problem of energy shortage,China’s wind power industry has developed rapidly in recent years,with a substantial increase in power generation capacity.The gearbox of the wind turbine is the main component of the wind turbine unit,and its working conditions are very complex,which is very prone to accidents.Ensuring its efficient and stable operation is the technical difficulty of the wind turbine unit.Therefore,fault detection in the gearbox of wind turbine is very meaningful.In this paper,the fault diagnosis of wind turbine gearbox is studied under the condition of rare fault data.In this paper,the theoretical basis of the algorithm model of the generation of countermeasure network and convolutional neural network is analyzed and summarized,and the algorithm model established by myself is introduced.Through the theoretical overview of the relevant algorithm models,and the analysis of the algorithm models derived from them,the algorithm model to be used in this paper is introduced.For the generation of data,the depth convolution generation countermeasure network algorithm is used,and then the fault diagnosis is carried out by building a multichannel weighted convolution neural network algorithm.Then,the improved depth convolution generation countermeasure network model is used to generate the fault data of fan gearbox.In the constructed model,the discriminator network layer adopts the form of convolutional neural network.The network layer of the generator combines the upper sampling layer and convolution layer to replace the transpose convolution operation.The input of the generator,in addition to the random sampling noise,increases the input of real data as an additional condition for data generation research.In the constructed model,the network layer mainly adopts the form of convolutional neural network,and uses the strong extraction ability of convolutional neural network for image features to transform the generation research of one-dimensional vibration data into the generation research of image data,so as to achieve the purpose of expanding the data set.Finally,through the multi-channel weighted convolution neural network model,the fault diagnosis of fan gearbox is studied.The EMD method is used to decompose the data,and then the first three with the largest kurtosis value are selected as multi-channel inputs.By building a multi-channel weighted convolution neural network layer to extract the dynamic features of the input images of each channel,and then weighted fusion,it can comprehensively extract fault features to increase the accuracy of recognition.It is compared with SVM,BPN,DBN,SDAE and other depth learning algorithms in terms of the accuracy of fault diagnosis,which verifies the effectiveness of the methods used in this paper and improves the accuracy of fault diagnosis.