Research On Residual Life Prediction Of Satellite Platform Based On Bayesian Theory

Residual life(RL)estimation is significant in reliability engineering.Traditional methods based on failure data show their limitations for satellite platform,which is characterized as high reliability,long lifetime and small sample size.Meanwhile,with the development of data gathering technology,multi-source reliability information can often be available for satellite platform.Based on the Bayesian theory,an approach for RL estimation by fusing multi-source information is proposed.By utilizing various kinds of reliability data,the precision of estimation for components,subsystems as well as systems is improved obviously.The main contents are as follows:(1)An approach of estimating the RL of exponentially distributed components by fusing multi-source information is proposed.Priors are determined by historical lifetime data and expert information,and the corresponding posterior distributions can be obtained after fusing field lifetime data.Additionally,consistency test is conducted and the joint posterior distribution is the weighted sum of these posterior distributions with weights calculated by the ML-Ⅱ method.Finally,Bayesian estimation can be calculated based on the posterior distributions.Accuracy of the proposed methods have been validated by extensive simulation.And the digital computer is taken as an example to validate the effectiveness of the methods.(2)As for Weibull distributed components,a method of estimating the RL by utilizing multi-source information has been proposed.Using historical lifetime data,lifetime data of similar products,degradation data and expert information,corresponding priors can be determined.After fusing lifetime data,posterior distributions can be obtained and the joint posterior distribution is the weighted sum of these posterior distributions.Considering the complexity of posterior distributions,a sample-based method is proposed to estimate the RL.Sufficient experiments and the case study of momentum wheel prove that the Bayesian method performs well.(3)Method for RL estimation of satellite platform system based on the fusion of multi-source information fusion has been discussed.Firstly,estimation results of the lower-level structure is necessary to be converted to the upper-level system.In this paper,on the assumption that the components are exponentially distributed or Weibull distributed,the closed-forms for lifetime and RL of typical systems are derived,and numerical methods are given as well.Then,by fusing the converted result and multisource information of system,the RL for system can be estimated.Simulation experiments and an example of a subsystem on the satellite platform validate the method.