1989 Volume 15 Issue 2 Pages 258-268
Although utilizing heuristic knowledge of various types effectively is indispensable to operating a chemical plant safely, very few methodologies for utilizing it on a computer-based system have been reported. The importance of this matter is increasing especially in the case of fault diagnosis. The work presented here addresses the following two issues : classification and encapsulation of heuristic knowledge of human operators which must be utilized in diagnosing failures of chemical plants, and establishment of a framework for developing a fault diagnostic system which mimics the manner of human operators.
When the plant control system has given an alarm, the operators first try to narrow down their focus to a bounded section where the fault may exist by testing if none of various relations inhering in the plant is violated. Then, they try to find the possible cause of the fault by simulating qualitatively the phenomenon that is occurring in the section. A computer-based system has been developed together with a knowledge base by following this procedure. The system has also been developed in open-ended fashion by using multi-language paradigm. The system is easy for human operators to accept, because the diagnostic algorithm is similar in manner to their own.
