Artificial-Intelligence-Based Quantitative Fault Diagnosability Analysis of Spacecraft: An Information Geometry Perspective
Dayi Wang, Fangzhou Fu, Han Yu, Weimeng Chu, Zhigang Wu, Wenbo Li
Abstract
To provide a theoretical foundation for the design of spacecraft and a reference for their on-orbit adjustment, this study develops methods for the quantitative diagnosability analysis of spacecraft through the introduction of a Riemannian manifold and artificial intelligence into a diagnosability analysis framework. These developed methods have three main advantages. First, they do not rely on any assumption regarding the statistical distribution of the observations. Second, all types of faults can be analyzed without information loss. Third, the computational burden is reduced, making it possible to perform on-orbit fault diagnosability analysis. The effectiveness and feasibility of the proposed methods are then verified via a numerical simulation and a spacecraft attitude control system.