Research On Remaining Useful Life Prediction Method Of High Speed Rolling Bearing Of Power Machinery

As one of the important basic components in power machinery and equipment,rolling bearings are prone to failure in the harsh environment of high speed and heavy load.Therefore,rolling bearings’ operating conditions are essential to the safe and reliable operation of mechanical equipment.It has important theoretical significance and engineering practical value for reasonable repair and maintenance programs,if it can accurately predict the remaining useful life of rolling bearings.At present,the data-driven method for predicting the remaining useful life of rolling bearings is the mainstream among many research methods.On this basis,this article focuses on the prediction of the remaining useful life of rolling bearings.The main research as followings:(1)This thesis introduces the basic structure of rolling bearings and establishes a simplified mechanical model of rolling bearings.The vibration mechanism is analyzed from the internal and external aspects of the rolling bearing.And the failure mode of the rolling bearing is analyzed.(2)A method for constructing health indicator(HI)of rolling bearings based on clustering by fast search and stack denoising autoencoder(CFS-SDAE)is proposed,and the model structure of CFS-SDAE and the construction process of health indicator are explained in detail.The comparison between the CFS-SDAE method and the traditional root mean square(RMS)and principal components analysis(PCA)methods on the PHM2012 rolling bearing data set shows that the HI curve constructed by the proposed method is smoother and has a better trend.(3)The basic theory and methods of the long short term memory neural network(LSTM)and the attention mechanism are introduced,and the attention mechanism is embedded in the LSTM network to predict the remaining useful life of rolling bearings.Validation on the joint data set shows that the robustness of the Attention-LSTM network is better than that of the existing LSTM network,the prediction result is close to the true life value,and the average prediction accuracy of the model can reach more than 92%.(4)Aiming at the LSTM network that only considers the positive dependence of the time series,the Bidirectional LSTM(BiLSTM)network is used to capture the patterns that may be ignored by the LSTM network.Therefore a prediction model based on the convolutional neural network(CNN)and BiLSTM hybrid neural network is proposed.This model uses the CNN network to mine deep features,and inputs the deep features into the BiLSTM network to predict the remaining useful life.The experiment is verified on the PHM2012 rolling bearing data set,and the results show that the proposed method is better than other CNN,LSTM,and BiLSTM models,and the average prediction accuracy can reach 90%.