Research And Application Of Fault Diagnosis And Prediction Algorithm Based On Deep Learning

With the development of a new generation of information technology,industrial intelligence upgrade is an effective way for manufacturing enterprises to maintain competitiveness in the market.The mechanical and electrical structure of industrial production equipment is becoming more and more massive,the production environment is becoming more and more complex,and the occurrence of faults is becoming more and more frequent.How to quickly diagnose the cause of faults when they occur or predict the point in time when they occur before they occur is the mainstream research direction of industrial intelligence,while improving information technology to be applicable to the industrial production environment and deploying it in practice is also a major challenge for industrial This paper addresses the above-mentioned issues.To address the above issues,this paper investigates the fault diagnosis of bearings and the Remaining Useful Life(RUL)method of equipment using deep learning algorithms in conjunction with the actual operation projects of assembly lines,and gives a design scheme for the application of fault diagnosis and prediction system,which is of practical value for the intelligent upgrading of assembly lines.The main work of the thesis is as follows.(1)A reliability analysis based on the fault tree analysis method is implemented for the assembly line.The fault diagnosis method based on continuous wavelet transform and Transformer model is designed for this part by analyzing the faulty parts of the assembly line that are prone to problems.Using the strong image feature learning capability of Transformer model,the vibration signal is converted into a time-frequency image representation by continuous wavelet transform,which is beneficial for Transformer model to learn signal features and output classification results.The final experimental demonstration of the proposed method verifies the performance of the proposed method.(2)A multimodal feature fusion RUL prediction method is proposed.For the problem of different RUL prediction features of devices in different modes,the joint feature learning of high-dimensional data inputs in different modes is performed by Stacked Sparse Autoencoder(SSAE)to solve the problem of mismatch between data features and RUL recession process in different modes;then,in order to solve the problem of long-life temporal features of devices Then,in order to solve the dependency problem of equipment long-life time series features,the joint feature output learned by SSAE is used as the input of the Long Short Term Memory Network(LSTM)model,which mines the long-range dependency features in the time series and finally outputs the prediction results of equipment RUL trend.Experiments on the turbofan aero-engine dataset demonstrate the effectiveness of the proposed model.(3)A data analysis system for intelligent fault diagnosis and prediction is designed and implemented.The system design includes data acquisition,transmission,storage and recall,and realizes the functions of fault diagnosis,fault prediction,historical data query and data analysis,etc.It completely realizes a fault diagnosis and prediction system for assembly lines,which can play a role in monitoring and maintenance of assembly lines and has high practical value.