The Reliability Assessment And Remaining Useful Life Prediction Of Rolling Bearing Based On The LSTM Network

Rolling bearing is an important basic component used in mechanical equipment,its operating state is closely related to the overall performance of mechanical system.Therefore,it is of great significance to evaluate the reliability and predict the remaining useful life of rolling bearings.In the current industrial background of big data and artificial intelligence,the intelligent monitoring and prediction methods for key components of mechanical systems based on deep learning technology have received more and more attention.In this paper,the full-life vibration signal of rolling bearing is taken as the research object,and the data-driven research method is used to construct the deep learning neural network prediction model,which provides technical support for mechanical equipment failure warning and maintenance strategy.The main research contents of the thesis are as follows:(1)Discuss the background and significance of the topic selection,describe the common methods and ideas for the reliability evaluation and residual life prediction of rolling bearings.The application of the trend prediction method is deeply studied,and the research progress of the recurrent neural network and its variant forms in the prediction model is discussed,then the life prediction idea is proposed.Introduce the typical failure modes and fault characteristics of bearings,lead into feature extraction method based on vibration signals,which lays a theoretical foundation for the following.(2)A reliability evaluation and prediction method for rolling bearings based on Weibull proportional hazard rate model and the Long Short-Term Memory is proposed.The kernel principal component analysis is used to extract the covariates to establish the WPHM and evaluate the reliability of rolling bearing.Combine with the neural network prediction model,the influence of the LSTM network parameters on the prediction results is studied and the prediction ability of the neural network is verified.(3)In order to solve the problem of feature parameters selection in life prediction process,three feature parameter evaluation indexes of correlation,monotonicity and robustness are proposed,and the feature parameters characterizing the degradation trend are quantitatively evaluated and screened.The data-driven LSTM network prediction method is used to constructed time-series data by using the selected feature parameters and extract the trend variation characteristics to realize the remaining useful life prediction of rolling bearings.The validity and practicability of the prediction method are verified by rolling bearing full life testdata.(4)Based on the previous theoretical analysis and algorithm,the advantages of LabVIEW platform and Python language are combined to develop a rolling bearing reliability evaluation and life prediction system,the system can realize mechanical system data acquisition,mechanical components online monitoring and the practical application of deep learning method.