Application Research Of PHM For Rolling Bearing Based On Deep Learning

With the deepening of Internet+ and artificial intelligence+ research,the manufacturing industry is also actively transforming into intelligent production and intelligent maintenance.Aiming at the problems of "under-maintenance" or "overmaintenance" in traditional maintenance strategies,prognostics and health management(PHM)have become the focus of current research.As the basic components of various industrial equipment,rolling bearing failure will directly affect the operation of the equipment,so the PHM research of the bearing has important significance.As a machine learning method that has been developed in recent years,deep learning has good performance and excellent models in image recognition and time series prediction.This paper focuses on the pain points such as difficult fault identification and inaccurate prediction under complex working conditions in the current PHM research of rolling bearings,combining energy spectrum analysis and deep learning methods,extracting fault indicators and establishing relevant models,and deeply studying the deep learning methods in bearing PHM Application,the main work is as follows:1.Aiming at the problem of inaccurate bearing fault diagnosis under variable load,a fault diagnosis method based on Teager energy spectrum and convolutional neural network is studied.On the one hand,the Teager energy spectrum is introduced into the fault diagnosis of the bearing,which can effectively distinguish the bearing state under variable load;on the other hand,the convolutional neural network is used to establish a diagnostic model,which directly extracts and analyzes the Teager energy spectrum.It solves the problem of imperfect artificial feature extraction and can better analyze the energy spectrum features to achieve deep feature extraction.Through the experimental analysis and verification of the bearing fault data set of the Western Reserve University in the United States,a comparative analysis of the network structure,etc.was performed to verify the effectiveness and superiority of the method.2.The frame and method of fault prediction based on the hybrid model of FDI and Multi-LSTM-BP are studied.According to the Teager energy spectrum and its envelope signal,the fault development indicator(FDI)that identifies the fault change trend is extracted;a Multi-LSTM-BP hybrid model fault prediction method is proposed.Finally,the experimental data was verified by the bearing data set of Xi’an Jiaotong University and Shengyang Technology Company.The experimental results show that the framework and method proposed in this paper have higher prediction accuracy.3.Preliminary study of bearing remaining life analysis to explore the impact of bearing failure on bearing life.The main consideration is to combine the development of the fault with the time-frequency domain characteristics of the vibration signal to form a new bearing health index,and use the LSTM method to predict the remaining life.In addition,AGA is used to optimize the selection of LSTM structure and parameters so that it can adaptively adjust the network structure for different bearings.The feasibility and effectiveness of the method were verified by an experimental analysis through the full life cycle data set of Xi’an Jiaotong University.The research results show that,combined with the deep learning method,the fault diagnosis method proposed in this paper can better grasp the data features,solve the problems of fault diagnosis and feature extraction under variable load,etc.Under the experimental environment and data set in this paper,a diagnostic accuracy of 95.4% was obtained.In addition,the fault prediction and life prediction methods based on LSTM have solved the problems of traditional methods relying too much on expert experience,inaccurate prediction effects and poor versatility.Through verification and analysis on public data sets,the accuracy of fault prediction was 92% through 160 test points.For the prediction of the health degradation index of the outer ring fault bearing,the MSE value was 0.0367,the RMSE was 0.19,which can more accurately predict the remaining life.