Research On Bearing Fault Diagnosis Based On Convolutional And Recurrent Neural Networks Under Big Data

Industrial plants are under tremendous pressure to continue reducing unplanned downtime,but it is quite difficult to effectively classify the types of bearing failures using traditional machine learning related algorithms.Based on the two characteristics of time series obtained from image data with both local features and global characteristics fully considered,this paper can accurately and efficiently complete the acquisition of global angular characteristics by using the unique characteristics of information transfer along the time channel in recurrent neural networks,by using long short-term memory networks(LSTM)and gated recurrent networks(GRU),and at the same time using the combination of LSTM and The combination of LSTM and GRU network can also effectively overcome the problems of long time memory loss and gradient disappearance and explosion,but the combination of LSTM and GRU will lead to long training time and insufficient local characteristics afterwards;at the same time,the combination of LSTM and GRU will lead to long training time,insufficient local features,and inability to perform distributed computation afterwards,However,the combination of LSTM and GRU will lead to long training time,insufficient local characteristics,and inability to perform distributed computation and other characteristics afterwards,A one-dimensional convolutional neural network is added to build a deep network diagnostic model of bearing faults,which is used to extract local features to establish a better CRNN(CNN–LSTM–GRU–LSTM)model.Finally,the CRNN diagnostic scheme and different models provided in this paper were tested against each other and migration validation experiments were conducted,through which the model has a high accuracy of bearing fault identification.In the actual production life process,due to the development of society,the explosion of bearing failure data has made it more and more difficult to store bearing data and diagnose bearing failure.To cope with the shortcomings of CRNN such as slow training speed and poor stability under big data.In the paper,we migrate the CRNN model to a big data platform and use Hadoop,an excellent big data platform,to achieve distributed storage of data.At the same time,the distributed framework of Tensor Flow,which has better scalability and ease of use,is used to rewrite the CRNN model,so that the distributed bearing fault diagnosis model of CRNN is constructed,and optimize and fine-tune the network parameters of the distributed model,and finally the better DCRNN(Distributed CRNN)model was finally determined.