| Most previous diagnosis expert systems use either statistical association between phenomena and components, or the causal relation that the failure of the component causes the phenomena. This thesis presents a new diagnosis expert system that uses both the statistical association and causal relations. A new statistical model of diagnosis has been proposed. The model distinguishes itself from previous models by including the relation between the failing situations and the failure of components. The actual diagnosis process has been analyzed to understand how human diagnosis experts use heuristic and causal knowledge systematically. It attempts to use three kinds of causal relations: first, a situation causes the failure of a component, second, the failure of the component causes phenomena, and third, the failure of one component causes the failures of other components (or the failure of a function). The use of the knowledge of the system structure in diagnosis is discussed. Mapping to a function of the system, an indirect mapping method, has be proposed. The advantages of indirect mapping methods are discussed. A diagnosis expert system has been developed to utilize the results of the theoretical studies and has shown promising potential over current diagnosis expert systems. |
