Design And Implementation Of Sliding Bearing Operating Temperature Prediction System Based On Deep Learning

Bearing is a part where generator failures occur frequently and plays a very important role in generators.Various power generation companies spend a lot of energy to regularly maintain and replace bearings every year.To determine whether the bearing is good or whether it needs to be replaced,the most commonly used and most feasible method is to measure the temperature of the bearing.Therefore predicting future temperature changes is very important for detecting bearing failure and preventing motor accidents.The author of this article is involved in a threering bearing working parameter prediction project for a hydropower plant.The main purpose is to develop a system that can predict the bearing temperature so that effective measures can be taken to enable the bearing to run healthily and effectively for a long time.This article is divided into five parts to elaborate the selection and development design process of various technical theories of the system in order to develop such a system.First,it introduces the purpose and significance of system development,the domestic and international research status of time series prediction,and the current research status and time-phase results of time convolutional networks,which can help readers understand the main areas and theoretical knowledge of the application.Secondly,it introduces the relevant theories of bearing failure,including bearing structure,bearing failure and common treatment measures,time series prediction algorithms,including RNN,LSTM and TCN,which are involved in this article.Thirdly,the article introduce the design of the time series prediction module,including data preprocessing,model performance measurement,model data processing flow and optimization of deep learning.Then,the results and analysis of the experiment are introduced including the experimental development environment and data sources,and the time series prediction experiment which is design to compare the advantages and disadvantages of the three algorithms in time series prediction,and the optimal parameter settings of the three algorithms.Finally,the design and implementation of the bearing working parameter prediction system are introduced including the functional analysis,and the overall design of the system which contains the overall architecture design,functional modules design and database design.The system can be applied to the actual industry after the system's functional testing.The system realizes user management,temperature prediction,abnormality and fault warning,fault management and statistics of power generation related data in general.It has been tested to meet the needs of actual industrial applications and it plays a good auxiliary role for the power plant workers' work.It's a good attempt to apply deep learning theory in the industrial field.