Research On Fault Diagnosis Method Of Motor Bearing Based On Deep Learning

The development of fault diagnosis technology is closely related to the development of new technologies.As the production environment of industrial machines becomes more and more complex,the requirements for fault diagnosis technology are gradually increasing.The normal operation of motor bearings is of great significance to social production and life and personal safety.In this thesis the vibration signal of the motor bearing is taken as the research object,and the deep learning method is applied to the fault diagnosis methods.The IGSA-SDAE and VMD-IGSA-CNN methods are verified by experiments.This thesis first expounds the purpose and significance of motor bearing fault diagnosis.Based on the domestic and foreign literatures related to fault diagnosis in recent years,this thesis summarizes the domestic and international research status of traditional integrated fault diagnosis technology and deep learning fault diagnosis technology.To research the fault diagnosis of motor bearings,the cause and vibration mechanism of motor bearing faults should be analyzed.This thesis studies on the basis of deep learning algorithms.Two kinds of deep learning models stacked denoising autoencoder(SDAE)and convolutional neural network(CNN)are improved.SDAE has local optimization problem of parameters in the training process.The optimization degree of network model parameters directly affects the performance of diagnosis.Based on SDAE,the improved gravitation search algorithm(IGSA)is used to optimize the network.The parameters are proposed based on the IGSA-SDAE fault diagnosis model.IGSA solves the problem of GSA boundary optimization and improves the slow convergence of GSA in the optimization process.In order to improve the noise immunity of the whole system and improve the accuracy of fault diagnosis,the variable mode decomposition(VMD)is introduced to denoise the whole vibration signal.Then the noise-rejected signal is input into the CNN.The IGSA is used to optimize the network parameters,and a new method based on VMD-IGSA-CNN for motor bearing fault diagnosis is proposed.In this thesis,the IGSA-SDAE and VMD-IGSA-CNN methods are validated in the experiments using the data from the Case Western Reserve University bearing test center.Two network structures are compared and selected during the experiment.The training accuracy and loss function curves are recorded during the experiment and the final test accuracy was obtained.These results are the criteria used to evaluate the pros and cons of the fault diagnosis model.Finally,the simulation results are compared with BP,SVM,traditional aggregate fault diagnosis and unimproved deep learning model.The combination of theoretical experiments proves that the proposed algorithm has higher diagnostic accuracy.A motor bearing fault diagnosis platform was designed for the experiments.