Life Prediction Of Satellite Key Components Based On RVM-PF

Life prediction of satellite key components can help ensure the satellite stable, efficient operation, is one of the core space technology. Now, the methods of satellite components life prediction based on data are usually by the processes of doing ground tests, collecting sample data, building degradation models and predicting the useful life. But due to the complex space environment, the in-orbit data of satellites are often uncertain and incomplete, which make the traditional prediction models can not effectively eliminate the influence of noise, and so, can not be applied in orbit prediction. In this case, the paper proposes a data-driven prediction algorithm using relevance vector machine and particle filter. The algorithm weakens the impacts of the physical characteristics, finds rules and builds models from the in-orbit performance data, solves the in orbit life prediction of satellite key components. The main contents of the paper are as follows:First, propose a degradation modeling method based on state-space model. Introduce the principle of relevance vector machine, analyze the impact of the parameters and the limitation of application. As the particle filter algorithm need the given state-space model, deduct the observation equation and the prediction equation from the data through RVM regression fitting and re-fitting.Second, propose a life prediction method based on particle filter. Introduce the principle of particle filter. In the basis of the established state-space model, give the complete processes of the RVM-PF prediction algorithm, and analyzed the algorithm features combined with the instances of sample data.Third, take the satellite momentum wheel as example, analyze the failure mechanism, select bearing temperature as the life characteristic, and verify the algorithm performance combined with the measured data. The results showed that prediction accuracy is higher, and the prediction algorithm can solve the problem of satellite key components life prediction based on in-orbit data.Finally, concluded the main works and key innovations of the paper, looked ahead the further study contents.