The learning aspects of the FAULT FINDER expert system

The FAULT FINDER expert system implements fault isolation decisions for any target system or equipment that is modeled by lowest replaceable units (LRUs). This expert system utilizes a database to represent each LRU, a status interface to obtain LRU status, and a knowledge base to store the rules of fault isolation for the target system. The expert system has learning capabilities in the database, the knowledge base, and the inference procedure. Another aspect of learning which influences the structure of the knowledge base is that each rule has parameters associated with it to store the information learned as a result of user feedback and the inference process. The certainty or possibility associated with the conclusion of each rule is adjusted as the system runs and gains experience. The inference procedure uses fuzzy logic for premise matching certainty and the combining of premise certainties for the rule firing certainty. Also presented are issues of knowledge structure and possible types of fault isolation knowledge.<
>