Research And Application Of Rolling Bearing Life Prediction Based On VMD And LSTM

Prognostics Health Management is an important part of equipment life cycle management.One of the key technologies is to realize the prediction of remaining useful life.Accurate prediction of RUL can reduce production losses and save maintenance costs.As a key component of mechanical equipment,the reliability and life of rolling bearings will affect the performance and life of the entire rotor system and even the entire mechanical equipment.Therefore,accurate prediction of rolling bearing RUL is of great significance.Based on the research and development of the gearbox system fault simulation software,this paper studies the vibration signal noise reduction,feature extraction and RUL prediction algorithm for the rolling bearing RUL prediction problem in this software,and gives specific details in combination with the gearbox system fault simulation software The design and implementation of the main work are as follows:(1)Through literature research,the domestic and foreign research status of vibration signal feature extraction,life prediction and time series prediction are analyzed.(2)Based on the fact that noise in the vibration signal of rolling bearings will affect the accuracy of RUL prediction,a noise reduction method based on variational mode decomposition is studied.The original vibration signal is decomposed by variational modal decomposition,and components with correlation coefficients greater than or equal to 0.3 are selected to recombine to obtain a reconstructed signal,thereby achieving noise reduction of the vibration signal.Compared with EMD noise reduction and wavelet threshold noise reduction,simulation results show that the noise reduction method based on variational mode decomposition is superior to EMD noise reduction and wavelet threshold noise reduction.(3)A total of twenty initial features in time domain and frequency domain are extracted from the noise-reduced signal.Through the analysis of feature life trend,the degradation features that can describe the degradation trend of rolling bearings are selected from the initial features,including root mean square,peak-to-peak value,and average frequency eight degradation features are used for RUL prediction of rolling bearing.(4)The LSTM network is used to predict the rolling bearing RUL,and then the experimental analysis is carried out based on the rolling bearing vibration signal collected by the PRONOSTIA test bed,and the comparison experiment with the support vector regression model and the BP neural network.The experimental results show that the LSTM prediction result is better than the support vector regression and BP neural network.(5)Based on the above work,this paper designs and implements the gearbox system fault simulation software..The overall design of the software and the key technologies involved are described.Finally,the application of rolling bearing life prediction in gearbox system typical fault simulation software is explained.