Research On Fault Diagnosis Method Of Rotating Machinery In Nuclear Power Plant Based On Deep Learning

Rotating machinery plays a key role in the operation of nuclear power plants,and its state directly affects the stable operation of nuclear power systems.Ensuring the safety and reliability of nuclear power systems has always been an important subject in the field of nuclear engineering.Therefore,it is of great engineering significance to study the fault diagnosis methods of rotating machinery in nuclear power plants.In today’s digital age,artificial intelligence is developing rapidly,researchers are committed to applying the achievements of artificial intelligence technology to fault diagnosis,and intelligent fault diagnosis technology has been gradually being applied in various fields.In large and complex nuclear power systems,there is a complex nonlinear relationship between the fault symptoms of rotating machinery and the cause of the fault.However,the nonlinear mapping ability of traditional shallow model is limited,and sometimes it is difficult to establish a good mapping relationship,which affects the diagnosis accuracy.Therefore,this dissertation studies the fault diagnosis method of rotating machinery in nuclear power plant based on deep learning,signal processing technology and multi-sensor data fusion technology.This dissertation takes the rotating machinery such as bearings,shafts,and motors,which are abundant and indispensable in nuclear power plants,as the research object.After fully investigating the research status of rotating machinery fault diagnosis technology at home and abroad,as well as the application of deep learning and data fusion in fault diagnosis and fault feature extraction respectively,the mechanism analysis on common faults of rotating machinery such as bearings,shafts and motors is carried out,and the relationship between the causes,fault forms and fault characteristic frequencies of common faults of bearings,shafts and motors is established.In view of the importance of the frequency domain information of rotating machinery and the unique advantages of data fusion technology,it is proposed to combine multi-sensor data fusion technology with fast Fourier transform method to achieve feature extraction of rotating machinery signals,to provide effective and reliable input information for fault diagnosis modules.The deep learning model has strong learning expression ability and feature extraction ability,which effectively overcomes the inherent defects of the traditional shallow model.In this dissertation,a deep residual neural network is constructed on the basis of research on the development background of deep residual neural network and the composition of its important components.Taking the deep residual neural network as the fault diagnosis module,combined with the proposed feature extraction method based on multi-sensor data fusion technology and fast Fourier transform,the fault diagnosis model of rotating machinery in nuclear power plant is designed and constructed.The corresponding experimental bench is designed and built according to the research object,and the fault diagnosis model is applied to the experimental bench to verify the proposed method of this dissertation.Through experiments,the feasibility and effectiveness of the fault diagnosis method of rotating machinery in nuclear power plant based on deep learning is proved.