AN ARTIFICIAL-INTELLIGENCE SYSTEM OF TROUBLE DIAGNOSIS FOR AIRCRAFT ENGINES

Reported here is about the trouble diagnosis system for AN-24 aircraft engine which has been realized by inputting the experiences of the re pair mechanics or experts of the engine as a computer software. The sy stem is composed of following four sections which are called ''model'' ; a phenomena model, an inference model, a learning model, and an inte rpretation model. Therefore, the system is called as ''model diagnosis system''. These four models are relatively independent which makes pa rallel operation, easy debugging, and addition of new knowledge possib le. The experience of the engine experts has been stored initially to outer knowledge base in the computer. Intermidiate knowledge which ari ses on the process of the inference is treated at inner knowledge base . The inner knowledge base adopts a blackboard structure. This makes t he system not only able to diagnose the vague preconditioned reason, b ut also to diagnose the unpreconditioned one by Learning. The validity of the system was proved from some experiments.