Research On Prediction Method Of Rolling Bearing Performance Degradation Trend

Predicting the state of important components in the mechanical equipment system can detect the abnormal conditions of important components in time,and carry out prevention and elimination work to avoid the expansion of faults.Rolling bearings are the most widely used parts in industrial production and daily life.The normal operation of rolling bearings will endanger the performance of rotating machinery and equipment.Therefore,it makes a lot of sense.Taking rolling bearings as the research object,the methods of the improved particle swarm optimization(PSO)algorithm parameter optimization Long Short-Term Memory neural network(LSTM)model and the time-based sequence model Holt-Winters improved by Grey Wolf Optimizer(GWO)are proposed to predict performance degradation of rolling bearings.First,the method of combining the Kernel Joint Approximate Diagonalization of Eigen-matrices(KJADE)and the two types of models is used to extract the performance degradation index of the rolling bearing.Then,in order to improve the prediction accuracy and adjust complex network structure parameters automatically,a method of the improved PSO algorithm parameter optimization the LSTM model is proposed.The PSO algorithm is improved to better the number of hidden layer nodes and batch size of LSTM,so as to speed up the speed of finding the optimal parameter combination and improve the prediction accuracy.Finally,the optimized model is used to predict the performance degradation trend of rolling bearings.Experimental results show that proposed method has better effects than LSTM,SVR and CNN in predicting the performance degradation of rolling bearings.Secondly,a method of rolling bearing performance degradation prediction based on GWO optimization HW model is researched.The degradation process of rolling bearings is a complex process,and the relationship between the degree of failure of the bearing and the change of time has the correlation characteristics of time series.The HW model has a strong advantage in dealing with non-stationary time series.Aiming at the problem of HW model parameter selection,the GWO algorithm is used for optimization.Under the same model parameters,single-step prediction and multi-step prediction are performed respectively using HW model;under the same prediction step size,GWO is used to improve HW parameters,and the results are compared with the ELM method.The experimental results show that the proposed method is well applied.Based on the full life data provided by the the Cincinnati University and the data collected from the bearing fatigue test bench of our laboratory,this paper makes an analysis.The research in this thesis indicates that the two proposed methods can predict the degradation trend of rolling bearings well,and the prediction accuracy is also improved compared with other methods.The work in this article has a positive effect on the research of the prediction of the performance degradation of rolling bearings,and helps to make the corresponding maintenance plan in time to prevent machine failure.