Research On Residual Life Prediction For Typical Satellite Platform Subsystem Based On Multi-source Information Fusion

Satellite platform subsystem is a typical kind of complex system with high reliability,long life and small sample test,residual life(RL)prediction of which is an important problem in satellite engineering.Various reliability information,such as expert information,life data and degradation data,can be collected for satellite platform subsystem.However traditional methods are no longer valid for the purpose of information fusion,so this paper proposed methods on RL prediction based on multi-source information infusion,which can improve the accuracy of RL prediction.In this paper,related issues are researched form satellite platform unit to satellite platform subsystem and main achievements are listed as follows:(1)Method of RL prediction for satellite platform unit fusing no-failure time censored life data,few-failure life data and expert information.Firstly,expert information about reliability are expressed as parameters' prior distributions.Then life data are fused with expert information with Bayes methods to determine the residual life probability density function,as well as the RL distribution.Lastly,some kind of GPS receiver is taken as example,and a simulated method and comparison with other method are provided to prove the effectiveness of the presented method(2)Method of RL prediction for satellite platform unit fusing no-failure time censored life data,few-failure life data and degradation data.Based on Wiener process,assumed that drift parameter ? and diffusion parameter ? are random variables,life data are fused with degradation data with Maximum Likelihood methods and Bayes methods.Besides,real-time RL probability density function is obtained and parameters estimation are achieved by MCMC methods.Lastly,Infrared earth sensor is taken as example to prove the effectiveness of the presented method.(3)Method of RL prediction for satellite platform subsystem by analytic derivation and simulation.Firstly,reliability model for satellite platform subsystem is built.Then based on the RL model of unit,RL distributions of different reliability structures are obtained by analytic derivation and simulation.Result of RL prediction for satellite platform subsystem is provided fusing RL distributions of different reliability structures.Lastly,the present method is valid for units for different distributions and is convenient for engineering application.