Research On Fault Diagnosis And Model Migration Method For Motor Bearings Based On Residual Networks

With the development of intelligent technology,data-driven end-to-end fault diagnosis has become a hot topic.Traditional data-driven fault diagnosis methods rely on manual feature extraction,and the model’s noise resistance and generalization are poor,which will affect the final diagnosis results.This article proposes an improved Residual Network(Resnet)for fault diagnosis of motor bearings.This method can directly extract the fault features of the original signal,avoid manual feature extraction,and improve the accuracy of motor bearing fault diagnosis.By introducing the bearing data set of Case Western Reserve University,the training link and testing link are placed in the data set to deduce the algorithm of the improved residual network,and by selecting the hyperparameter of the network through multiple sets of comparative experiments,an improved residual network is derived.This network utilizes the introduction of an improved residual module to enhance the network’s feature extraction ability,improve the accuracy of motor bearing fault diagnosis,and enhance the model’s noise resistance;By comparing with other intelligent fault diagnosis algorithms,the effectiveness and practicality of the proposed improved residual network are demonstrated;And provide visual analysis of the algorithm.In order to solve the problem of poor generalization ability of the model in engineering practice,this paper optimizes the improved Resnet as the main network of model transfer learning tasks.The optimization of migration tasks is improved from two aspects: data level and algorithm level.At the data level,generative adversarial networks are used for data augmentation;At the algorithm level,the Dropout layer and domain adaptation module are introduced to improve the accuracy of model diagnosis.Design migration experiments for multiple variable operating conditions and switching prototypes based on measured datasets,and validate the improvement and enhancement of model migration method generalization performance through experiments.The experimental results indicate that the proposed model provides a feasible method for the engineering application of motor bearing faults through improvement.