Research On Rolling Bearing Fault Identification Based On Deep Residual Shrinkage Network

With the development of machinery manufacturing industry,condition monitoring and fault diagnosis of operating equipment are of great significance to reduce maintenance costs,reduce losses and improve system reliability.Bearing is one of the most frequently used parts in mechanical equipment,so the accurate and effective use of bearing failure identification methods can help ensure the sa fety and stability of online equipment.Traditional methods often need to rely on manual and expert,which increases the complexity and lag of fault identification and is not conducive to the realization of intelligent fault identification and detection of online equipment.Therefore,in order to achieve intelligent fault identification detection for rolling bearings,this paper investigates the method of bearing fault identification based on deep residual shrinkage network.In this paper,we analyze typica l and several faulty bearings using rolling bearing data published by Western Reserve University as an experimental study.In view of the shortcomings of the existing rolling bearing fault identification method,the rolling bearing fault identification method is improved.In this thesis,the main work includes the following areas:(1)This paper describes the connection between solid media and vibration signals,illustrates how vibration signals of different faults are generated in rolling bearings,and specifically introduces the current signal processing methods commonly used.(2)Based on the extracted vibration characteristic parameters of rolling bearing faults as samples,the rolling bearings under different working conditions were characterized,and the normal rolling bearings and faulty rolling bearings were correctly determined and identified according to a series of characteristic parameter change conditions.(3)Improve a fault identification method-deep residual shrinkage network,solve the problem of low sensitivity and accuracy of traditional fault identification.The method targets the performance change process of rolling bearings from normal to failure,and constructs full-cycle failure sample points for rolling bearing failure identification by collecting multiple random points.Based on the CNN model,the residual network is added and the soft threshold method is used to reduce the sample noise.After the simulation of the fault signal,the analysis of the actual measured signal,the fault identification with CNN as the model and the deep residual network as the model and the comparison of parameters,the actual results prove that the method of this paper can have high fault identification accuracy.(4)For the current problem of fault identification accuracy,a method of fault identification based on CNN model of DRSN network combined with DS evidence theory is proposed.DS evidence theory is used to fuse the results of multiple fault identification and output the final results.The test results show that the method proposed in this paper can identify the rolling bearing failure in various cases,and compared with other identification methods,there is a high accuracy rate,which proves the correctness of the proposed method.