Decentralized Diagnosis with Isolation on Request for Spacecraft*

This paper presents a decentralized diagnoser architecture for fault detection and isolation of continuous systems with a focus on space applications. The diagnosis algorithm utilized in the architecture is based on analytical redundancy relations. Local diagnosers work on functional subsystems with a supervisory diagnoser at the higher level responsible for resolving ambiguities arising from the interaction between subsystems. We demonstrate how the decentralized architecture can be used to address some of the key problems associated with spacecraft diagnoser design. These issues are the opacity of diagnoser structure, integration of diagnoser design and development into the system engineering framework, and the reduction of computation and communication overheads. Varying diagnosability levels can be realized depending on the mission phase. We develop diagnostic models of a satellite attitude determination and control system, and then design a decentralized diagnoser based on these models. Simulation results for a case study are presented.