Research On Bearing Fault Diagnosis Based On Deep Auto-Encoder Network

Rolling bearings are an important part of rotating machinery.Failures can directly lead to major economic losses and even casualties.Due to the harsh and complicated environment of the rolling bearing,the fault diagnosis effect is poor.Therefore,effective bearing fault diagnosis methods are of great significance to the normal operation of the machine.In the complex environment,the traditional bearing fault diagnosis method relies on professional diagnostic knowledge to manually extract fault features,and it is difficult to ensure the accuracy of fault features.The deep auto-encoder network(DAEN)uses the greedy layer-by-layer training method to automatically capture useful information directly from the original vibration signal.Through DAEN used its strong feature extraction capability and powerful computing power,the accuracy of bearing fault diagnosis can be guaranteed and the cost can be reduced.Therefore,the study of bearing fault diagnosis methods based on DAEN is essential to ensure efficient and high-speed safe operation of industrial equipment.Firstly,this paper studies the bearing fault diagnosis method.The common methods of bearing faults are studied.The advantages and disadvantages of classical fault diagnosis methods and feature extraction methods are analyzed,and their feature methods are compared.The techniques of deep learning is researched and the principles of DAEN is analyzed.The cloud adaptive particle swarm optimization(CAPSO)algorithm is studied and its advantages and disadvantages are analyzed in depth.Secondly,an improved DAEN algorithm is presented.The method uses the maximum correlation entropy as the loss function to reduce the influence of noise on the accuracy of fault diagnosis.The method uses the maximum correlation entropy as the loss function to reduce the impact of noise on the accuracy of fault diagnosis,and the model uses the randomness and stability of CAPSO algorithm to optimize the connection weight of DAEN,to reduce the constraints on the weights and extract fault features adaptively.Efficient and accurate fault diagnosis can be implemented with the Softmax classifier.Based on this,a bearing fault diagnosis model is established.Finally,a bearing fault diagnosis model based on DAEN is established,which realizes intelligent fault diagnosis.It has trained and learned the bearing fault vibration data set,which is the vibration signal of the rolling bearing,The data set is divided into training samples and test samples.The result is the number of layers and the number of hidden layer nodes of the best auto-encoder(AE).Principal component analysis is performed to verify the performance of the proposed algorithm in classification.The results of test show that the proposed method has higher diagnostic accuracy and more stable diagnosis results than those based on the Support Vector Machine(SVM)and the Back Propagation algorithm(BP)under appropriate parameters.