Research On Prediction Of Remaining Useful Life For Rolling Bearings Based On RNN And TL

Prognostics and Health Management(PHM)is an effective way to ensure the safe operation of equipment,and Remaining Useful Life(RUL)prediction is the most core technology among them.Rolling bearings,as key components of rotating machinery,are widely used in various equipment,and their health largely determines the smooth operation of the equipment.In order to solve the problem that the failure modes of rolling bearings are diversified due to the changing working conditions,which makes it difficult to predict their remaining useful life,this paper combines the Long Short-Term Memory(LSTM)neural network and transfer learning(TL)to predict RUL of rolling bearings under multiple failure modes.The specific research contents are as follows.(1)Established health indicators for rolling bearings.Firstly,feature extraction is performed on the vibration signals of rolling bearings in the time,frequency,and time-frequency domains to establish a 30 dimensional feature matrix;Then,the first feature dimensionality reduction is carried out according to the temporal correlation and monotonicity of features;Then,Principal Component Analysis(PCA)is used for the second feature dimensionality reduction;Finally,Adaptively Spatial Feature Fusion(ASFF)is used to fuse the multidimensional feature matrix into one-dimensional feature vectors,and the final health indicators are constructed.(2)Divided the health status of rolling bearings.Based on the degradation trend of rolling bearings,an improved Fuzzy C-Means Clustering(FCM)was used to divide the degradation process of rolling bearings into four stages: health,slight wear,severe wear,and complete failure.The inflection points of two stages were identified,and the degradation and failure points were clarified.(3)Predicted the remaining useful life of rolling bearings under different failure modes.A double-layer LSTM rolling bearing RUL prediction model was constructed to address the problem of rolling bearing RUL prediction under a single failure mode;Considering the gradual diversification of the working environment of rolling bearings,which leads to the diversification of failure modes,LSTM and TL are combined to predict the RUL of rolling bearings under multiple failure modes.(4)Developed a prototype system for predicting the remaining useful life of rolling bearings.Based on the requirement analysis of the system,determine the specific functions that the system needs to have,including data import,model training,lifespan prediction,and data storage;Next,design and develop the process and framework of the system;Finally,implement each functional module.Research has shown that the RUL prediction model for rolling bearings constructed in this article has improved average prediction accuracy in both single and multiple failure modes,and can serve as a basis for Prognostics and Health Management of rolling bearings.