Data Driven Prediction Method For Degradation Trend Of Rolling

The rolling bearing is a very important component in the mechanical structure.The main function is to transform the sliding friction into the rolling friction,so that the friction between the mechanical components is greatly reduced.According to the statistics,the damage of the 70% large mechanical structures is related to the bearing damage.It can be seen that the prediction of the degradation trend of bearings is very important.For the prediction of the degradation trend of bearing,we should focus on solving two problems.First,build appropriate health factors to evaluate the performance degradation of rolling bearings.In this paper,we choose the health factor based on the principal component analysis,that is,to integrate the information of multiple characteristic parameters and construct the health factor.Second,choose the appropriate degradation state prediction model and predict the degradation trend of bearings accurately under limited h istorical data.On the basis of the existing historical data,the prediction is to make a scientific prediction of the data after the current time node by using the corresponding analysis method or data processing model to analyze the data.The commonly us ed prediction models include curve fitting model,neural network model,support vector machine model,Elman model and so on.In this paper,the prediction of degradation trend of rolling bearings is discussed under the condition of less historical data and adequacy.In the short term,the commonly used BP neural network and support vector machine are used in the short term prediction,and the third chapter also compares and analyzes the advantages and disadvantages of the two methods.The method of time SFA M network and Elman network is applied in the long term prediction.At the end of the fourth chapter,this paper compares two long-term forecasting methods,and gives relevant conclusions.The practicability of Elman network is better than that of SFAM network.However,when SFAM runs data for some bearings,the prediction of degradation trend is more accurate.In the end,a hybrid prediction model is built.Based on the hybrid prediction model of BP neural network and support vector machine,the error of the hybrid prediction model is smaller than that of any model.It provides a new idea for the prediction of the trend of the rolling bearing degradation.