| Summary: | The accurate prediction of reliability for long-time running intelligent satellite power distribution systems is crucial in engineering. In this paper, an adaptive method is proposed to achieve this goal. Based on lifetime and degradation data, an estimator of the reliability for the system is derived by mainly using an additive degradation model of combined Poisson and Gaussian processes. A locally c-optimal approach to choosing effective data from the real-time data flow is given. Associated with the sequence of observed lifetime and degradation data, a robust criterion is proposed to determine an appropriate data subset for reliability prediction. A simulation study shows that the proposed method gives superior performance over the traditional method. Benefiting from adaptive and optimal strategies, the reliability predictions for 16 to 20 years obtained from the proposed method are convincing even if the initial models fitted by the ground test data have deviations from the true models.
|
|---|
