| The objective of this research is to develop and implement a generic diagnosis model which will be the first step towards the development of intelligent diagnostic systems for complex systems. The model developed in this research is based on the theory of hierarchical systems and the hybrid reasoning approach of artificial intelligence.;From the theory of hierarchical systems, a system can be described in terms of three types of hierarchies: strata, layers, and echelons. The bases of these hierarchies are the level of abstraction, the level of decision complexity, and the organizational level, respectively. Strata and echelons are used in this model.;The two types of knowledge used in artificial intelligence are deep and shallow knowledge. Deep knowledge, based on the structural, functional, or behavioral information of a system, is concerned with how and why the system works and what it does. Shallow knowledge, based on production rules, provides specific details for possible circumstances in operating the system. Diagnostic reasoning is the process of finding or locating faults responsible for a symptom observed. In this model, deep reasoning uses a deep KB (Knowledge Base), defined as functional blocks in terms of strata, and shallow reasoning uses a set of shallow KBs, consisting of production rules represented as echelons. The approach for hybrid reasoning used in this model is characterized by "deep reasoning first, then shallow reasoning.";Given a failure probability of each functional block in the deep KB, the diagnostic strategy begins with selecting the most unreliable functional block. The shallow KB, corresponding to the finally-selected functional block, is then explored using the information on degree of belief, test cost, and entropy.;The model is implemented on KEE (Knowledge Engineering Environment), a shell for building an expert system, installed on the Symbolics, 3640 machine, utilizing an example of a CIMS (Computer-Integrated Manufacturing System). |
