Remaining Life Prediction And Experimental Research Of Metro Traction Motor Bearing

The remaining life prediction of rolling bearing can not only provide basis for making maintenance strategies,but also can prevent accidents.In order to properly evaluate the remaining life of the rolling bearing and make reasonable maintenance plans,this paper selects the metro traction motor bearing as the research object,based on the rolling bearing fatigue life test-bed,carries out the fatigue life test of the metro traction motor bearing,through collecting and analyzing the vibration signal of the metro traction motor bearing,selects the characterization index of the bearing performance deterioration trend,and then,the characteristic data of bearing vibration signal is extracted.Then,the extracted characteristic data is input into neural network for training.Finally,the remaining life of metro traction motor bearing is predicted.Firstly,the fatigue life test platform and signal acquisition system of metro traction motor bearings are built,and the orthogonal test scheme is designed.The fatigue life test is carried out within the load range specified in the national test standard,and the remaining life data of bearing in 120 hours are collected.Secondly,the stability and sensitivity of the time-domain characteristic parameters of metro traction motor bearings are compared and analyzed,and the RMS characteristic parameters are initially selected as the index to characterize the declining trend of bearing performance.And then,the RMS characteristic data of six kinds of fault bearings are extracted.In view of the fact that the amplitude of RMS is not uniform when characterizing the law of bearing degradation,the relative root mean square characteristic index is designed.Six relatively smooth trend curves of bearing performance degradation are obtained through data denoising,and the bearing failure time is determined.Then,the relative root mean square characteristic data samples of bearing noise reduction in six fault states are analyzed,each of the six groups of data selects one as the test set,the rest as a training set,input BP neural network to predict the remaining life of bearing.Finally,the whole remaining life cycle of metro traction motor bearing is divided into three different degradation stages.With the help of the advantages of SVM in small sample classification,the degradation state of bearing is evaluated.The results show that the predicted remaining life of metro traction motor bearing is basically consistent with the real value.The remaining life of metro traction motor bearing can be predicted by the method of combining the relative root mean square feature with BP neural network,and good results can be achieved.