Research Of The Residual Useful Life Prediction And Maintenance Optimization Based On Service Status For Rolling Bearings

Nowadays,rail transit has become the lifeblood of China's urban transportation system with continuing rapid developing speed.However,the existing maintenance technology can hardly catch up with the need of safety assurance measures for massive amount of rail vehicles under increasing operating pressure.Rolling bearing is a key component of the vehicle,of which the operating status directly relates to vehicles running safety.However,this component is highly prone to crack or even fracture failures due to running under a harsh condition in long-term.Therefore,in order to avoid accidents,to improve safety support capability,to increase the efficiency of vehicle operation and to reduce maintenance costs,it is of great significance to offer scientific maintenance decision support to vehicle bearings.Given this,this thesis carried out a whole procedure-from status feature extraction,data analysis and forecast,to decision optimization-to ultimately realize the proactive maintenance for rail vehicle bearings in full life cycle:1.Aiming at extracting fearures that can effectively characterize the bearing degradation state in full lifetime,this thesis adopted multiple methods.Firstly,the vibration signal is decomposed by three methods:wavelet packet analysis,empirical mode decomposition and local mean decomposition.Then,the direct time domain features,frequency domain statistical features and time-frequency energy features are selected to characterize bearing states.By means of testing the the features above by different fault types and different fault degree data,the features with good performance are choosen to extract the full life bearing state.Finally,feature fusion and dimensionality reduction are carried out by principal component analysis,and the main principals are used as final features to represent bearing state.2.According to the requirement of high reliability in rail operation,this thesis employed Weibull distribution based on reliability analysis to describe the general rule of bearing degradation.Combining with the proportional hazard model,the state features are introduced to the model in the form of covariates.So a residual useful life prediction model based on sevice state is established.Specifically for solving the problem of parameter estimation difficulty,an improved inertia weighting factor adjustment particle swarm optimization algorithm is proposed.Finally,the experimental results show that the proposed method can well track the whole life trend of bearing and give good prediction results,which is in accordance with the practical application requirements.3.Based on the research results of recidual useful life prediction,this thesis furthrer studies the optimization of the maintenance plan.According to the current operation and maintenance problems and the expectations of the operators,a fuzzy multi-objective maintenance model with the objective of maximum availability,minimum cost under reliability constraint is established.In terms of the current monitor applications,the model of online and offline state monitoring mode are discussed in detail respectively.As for the offline mode,an adjusting monitor interval method is proposed.Finally,two kinds of bearing state data is used to verify this model.The model offers the best maintenance time and maintenance strategy based on bearing state,which accomplished a whole process of the scientific proactive maintenance for the rail rolling bearings.