Abstract
A fault diagnosis method based on signed digraph (SDG method) is suitable for the fault diagnosis of a large-scaled chemical plant because of its diagnosis speed and its easiness in modeling the objective plant. The method, however, has not yet been put to practicable use since it may output multiple candidates for the origin of failure. In this paper, two methods (Delay method and Improved delay method) are proposed to output the candidates in order of the possibility of becoming true origin using the information on delays among the state variables in the objective plant. Furthermore, a quantitative estimation method of the information content using the idea of entropy of the diagnostic result has been developed. The usefulness of Improved delay method has been demonstrated by comparing the information contents of the diagnostic results obtained from SDG method, Delay method and Improved delay method in the experiments with the use of a mini-plant.
