Research On Lifetime Prediction Method For Satellite Equipment Based On Information Fusion Theory

Rapid development of aerospace industry makes lifetime prediction become a core technology for in-orbit spacecraft equipments to assure their reliability and safety. It is difficult, however, to synthesize limited reliability information and acquire precise lifetime due to the complexity, redundancy and small-sample-data property of aerospace equipment, and that becomes an urgent problem to be solved.Information fusion based lifetime prediction methods in multilayers are put forward. Matlab simulations based on satellite momentum wheel are conducted. Specific work and research in this thesis are summarized below.1. The Bayesian method and its relevant knowledge are adopted in data level fusion algorithm to process multisource's life distributions as apriori information. Two data based methods are put forward: Correlation function algorithmic calculates the correlation between multiple apriori distributions and the test samples ditributions, Maximum entropy and Moment matching mehtod measures the entropy of multisource information. Both of they eventually derive the weight of multisourses for lifetime prediction fusion.2. An information method based on D-S evidence theory is proposed to obtain and collect both internal and external information from multisource to build D-S evidence collections. Then, D-S combination rule is used to obtain reasonable weight allocation for prior distributions. After that, fusion posterior distribution is figured out and the value of life parameter is also estimated. Finally, according to parameter's estimated value, the reliability lifetime prediction for momentum wheel is derived.3. A lot of lifetime prediction methods have been reported in existing literatures, but no evaluation method is available to assess their prediction performances and applicability. Thus, an evaluation system is developed for those lifetime prediction methods. Triangular fuzzy number is introduced in the analytic hierarchy process(AHP) to determine the weight of each index, which can lead to more reasonable and precise evaluation results. Finally, reasonable weight allocation is derived according to the score results of multiple lifetime prediction methods.