Research On Fault Diagnosis And Remaining Life Prediction Of Rolling Bearing Based On Deep Learning

As the basic part of the rotating machinery equipment,the rolling bearing is very important to the equipment.The quality of the rolling bearing will directly affect the operation of the mechanical equipment.In this context,it is very important to identify and judge the rolling bearing fault and predict the remaining useful life(RUL)of the mechanical equipment,which can reduce or even avoid the accidents caused by the mechanical equipment fault,and provide a safe and stable later maintenance strategy according to the RUL of the equipment.In order to effectively monitor the health status of mechanical equipment,it is necessary to carry out regular fault identification and life prediction of rolling bearing.In order to extract more sufficient feature information of the original vibration signal,and carry out accurate fault feature recognition and RUL prediction,in this paper,a fault identification method based on the combination of time-frequency transformation analysis method and convolutional neural network(CNN)is proposed in this paper,and then proposes a rolling bearing RUL prediction method combined with ensemble empirical mode decomposition(EEMD)and long-term and short-term memory network(LSTM).The research contents are as follows:Traditional fault diagnosis method needs to extract features manually,which leads to some irreversible errors due to human being.This paper proposes a method of fault diagnosis of rolling bearing by inputting vibration signal in time-frequency domain into CNN model to solve this problem.In order to improve the learning accuracy and convergence speed of CNN,Adam algorithm is introduced to optimize,reduce the learning rate of the algorithm,and learn CNN model in an adaptive way.Based on the data of CWRU experimental center,the time-frequency characteristics based on CNN are obtained for analysis and diagnosis.The diagnostic rate is higher than that of the time domain and frequency domain under the same conditions.In order to predict the remaining useful life of rolling bearing,it is necessary to fully extract the characteristic information of bearing vibration signal and construct the bearing health index.This paper proposes a RUL prediction method based on the combination of LSTM and EEMD.Firstly,the vibration signal is analyzed in time and frequency domain,and the basic feature set is selected for RUL prediction in this paper.In view of the fact that the extracted features may contain redundant features that can not be used,the correlation coefficient principle is used to reduce the features,and finally a more sensitive trend feature set is obtained.Finally,the LSTM network model is established to predict the remaining service life of the bearing.The experimental results show that this method is more accurate than the traditional feature extraction method for the life prediction of rolling bearing,and the prediction result is more ideal.Aiming at the problem that the degradation state of rolling bearing is changeable,and the degradation characteristics are not obvious,this paper uses LSTM network to build the health index.When LSTM-HI is 0,the bearing fails,and the effect is remarkable when it is used in the prediction of bearing remaining life.Finally,supported by the data of PRONOSTIA experimental platform,the RUL prediction method of bearing based on EEMD-LSTM is compared with EEMD-BP network and EMD-LSTM network.It can be found that the prediction effect of the proposed method is more significant.