Relational Approach To Fault Diagnosis BasedOn A Functional Model

Matra Marconi Space has been developing operational expert systems for spacecraft fault diagnosis using hybrid approaches, including techniques related to model based approach and fault trees. Keeping in mind the operational constraints of exhaustivity and efficiency that guided the initial choices, the reasoning paradigms developed are now refined, to improve the solutions reached. The paper describes some of the techniques implemented so far, and proposes extensions to the relational approach to automated diagnosis, in order to include more of the available knowledge to refine the representation of influences, by distinguishing certain and possible influences, and through a gradation of the uncertainty introduced.