Research on the Diagnosability of a Satellite Attitude Determination System on a Fault Information Manifold

Abstract

In this paper, a new method for fault diagnosability research based on information geometry is proposed. The problem of the diagnosability evaluation of dynamic system faults is transformed into a distance calculation problem on a manifold. The Fisher information distance is used to realize a quantitative judgment of diagnosability, and a quantitative evaluation index of the fault diagnosability of a satellite attitude determination system is designed. This includes a fault detectability index and a fault isolability index. The validity and superiority of the new indexes are verified through a mathematical simulation. In addition, the fault information is visually presented by the geodesics of the fault manifold, and the properties and behavior of the fault are mined and analyzed on the fault information manifold, which lays a foundation for further exploration of fault information through geometric methods.