An integration of heuristic and model-based reasoning in fault diagnosis

Abstract

A human expert utilizes various pieces of diagnostic knowledge very dynamically in problem solving. Recently a few integrated reasoning systems have been proposed to use various types of knowledge in an integrated way, but they have shortcomings in reasoning control and reasoning interactions. To overcome these shortcomings, a diagnostic reasoning system with dynamic interactions that uses diverse types of knowledge with cooperation is proposed. The integrated diagnostic reasoning system contains heuristic and model-based reasoning modes. A type of knowledge in one reasoning mode is shared with another via transformation knowledge. In the integrated diagnostic reasoning system, the heuristic reasoning mode plays a major role in reasoning, and the model-based one fills the gaps in the heuristic reasoning mode. The model-based reasoning mode helps the heuristic one by supporting primitive-level testing, and by generating symptoms and hypotheses. In addition, it undertakes the remaining tasks if the heuristic reasoning mode fails. The heuristic reasoning mode, also, aids the model-based one by pruning the candidate set. By applying the integrated reasoning system to diagnosis of a computer printer circuit the system is shown to be appropriate for the domain that calls for the cooperative use of heuristic and detailed system description knowledge. With the criteria of efficiency and coverage, it has been shown that the proposed system is appropriate when insufficient symptoms and heuristic rules are provided.