Research On Remaining Life Of Fan Bearing Of Electrical Multiple Unit Traction Motor

The requirements for the reliability,availability,maintainability,and safety of high-speed EMUs were constantly raised with the rapid development of railway transportation and exacerbating market competition.Bearings,whose performance gradually deteriorated under severe environmental conditions for long time,are important parts in mechanical systems.The operating state directly affected the operating state of the entire vehicle and the vehicle's operating safety.The analysis of bearing operation status and prediction of remaining life can not only reduce the malfunction rate of EMUs,improve the safety of EMUs,and avoid catastrophic accidents,but also reduce the maintenance costs of EMUs and increase resource utilization,which is in line with the concept of sustainable development.The research object is the traction motor fan bearing in the article that has been used for four years,whose remaining life was predicted when it will be repaired in the fourth level(It will be used for six years).The bearing life calculation formula proposed by the International Organization for Standardization can only calculate the average life of the bearings without considering the individual differences of the bearings.Although the finite element analysis method can analyze the remaining bearing life under partial fault conditions,the simulation modeling is complicated and the analysis efficiency is low.Therefore,this paper proposed the method based on deep learning to predict the remaining bearing life.The main research contents are as follows:(1)The maintenance process of the EMU was introduced in the article.The difficulties were put forward in the prediction of the remaining life of the traction motor fan bearings.The current research status of bearing remaining life prediction methods,the research status of bearing accelerated life test,and the advantages and disadvantages of various signal processing methods were discussed in the article.(2)The average remaining life of the bearings were calculated according to the working conditions of traction motor fan bearings.The mechanical model of the traction motor fan was established to calculated the radial force on the bearing.Based on the L-P bearing life theory and I-H bearing life theory,considering the influence of factors such as lubrication and temperature on bearing life,the International Organization for Standardization had proposed standardized bearing life calculation formulas to calculate the remaining bearing life of bearings under different reliability conditions,but it represented the average remaining life of the bearing under operating conditions.(3)The bearing accelerated life test was designed to acquire the bearing vibration signal,to analyze and predict the running status of the bearing.The accelerated life test of the bearing was finished by increasing load test,and the vibration signal of the bearing was collected at the same time.Variational modal decomposition method was used to process the bearing vibration signal,and the AR energy spectrum of the entire life cycle of the sample bearing in the accelerated life test was calculated.The K-means algorithm was used to analyze the running status of each sampling point in the whole life cycle of one of the bearings.Based on the analysis result of the running status,the SVM algorithm was used to predict the running status of the remaining bearings.(4)The acceleration bearing life factor was calculated.The finite element analysis model of thermal-transient dynamics was established to simulate the real working conditions of the bearing,and the fatigue life analysis module was used to obtain the bearing life acceleration coefficient.The acceleration coefficient that based on the finite element analysis was compared with the acceleration coefficient of bearing life theoretical construction.(5)The bearing remaining life was predicted.Aiming at the prediction starting sample points of different operating states,the paper proposed different methods for predicting the RUL of bearings.When the bearing operation state was in the decline period,the Moving Window Long and Short-Term Memory(MWLSTM)network was used to predict the AR energy change after the prediction starting point,and the RUL was determined by the energy limitation.The error between the prediction result and the real result was very small,so the prediction method had high reliability.When the bearing operation state was in the stable period,the number of the sample points whose operation state was similar with the operation state of the prediction starting points was large and the degree of similarity was close.Based on the three-parameter Weibull model,the RUL of the bearing was analyzed.The error between the prediction result and the real result was very small,so the prediction method was effective.According to the accelerated bearing life coefficient,the prediction result was converted into the RUL under actual working conditions.