A new shell for the development of alarm pattern recognition expert systems

Abstract

This paper presents the first version of GENESIS, an expert system shell suitable for the development of alarm pattern recognition expert systems (APRES). GENESIS Includes a series of algorithms and procedures especially designed for a rapid and systematic construction of APRES. The Inputs required by GENESIS are the fault trees of the system under analysis, the probability of occurrence of each fault in the trees, and the set of symptoms (alarms and measurements) associated to the occurrence of each individual fault. With this Information, GENESIS generates a set of production rules which relates faults and symptoms. The shell uses these rules and the probability of occurrence of each of the faults In order to generate optimal alarm pattern recognition strategies (algorithm of the Inference engine). A strategy helps the operator to recognize which alarm pattern is occurring without having to search the entire set of patterns. All the alarm pattern recognition strategies are generated off-line, as a consequence, the response of the system will be very fast. This feature makes GENESIS a powerful tool for the development of real-time APRES.