Abstract
This paper proposes a method for diagnostic problem-solving with fuzzy abductive reasoning. Diagnosis is considered as a process to derive a set of hypotheses that explains the observed symptoms using the causal relationship from causes to symptoms. This process of inference is called abduction. The observed symptoms frequently include vagueness or fuzziness in their degrees of occurrence. Therefore, fuzzy abduction introducing the fuzzy theory might be effective for diagnoses. However, in diagnoses in the real world, it is risky to assume that operators can find all of the symptoms. They may fail to find some symptoms. In such cases, the conventional fuzzy abduction sometimes fails to obtain the solutions, because it tries to find causes that are the complete explanation of the given symptoms. In this paper, we introduce the concept of “cover” to cope with the problem, and expand the fuzzy abduction with it. The new idea is to find a set of causes that “covers” the observed symptoms, not that derives the symptoms exactly. We then apply the expanded fuzzy abduction to a diagnostic problem of a diesel engine for ships, and confirm its usefulness. Furthermore, we compare the fuzzy abduction with another approach, the inverse problem of fuzzy relational equation, which has been used to infer causes from observed symptoms.
