A MULTIEXPERT KNOWLEDGE SYSTEM ARCHITECTURE FOR MANUFACTURING DECISION ANALYSIS (LEARNING, PROCESS PLANNING, EXPERT SYSTEM, ARTIFICIAL INTELLIGENCE, KNOWLEDGE FUSION)

The use of AI/expert systems in evaluating AI technology for candidate manufacturing applications is addressed. The problem domain is focused on evaluating the technical feasibility of applying AI technology to process planning tasks. The research asserts that decision problems of this nature are too qualitative and complex for mathematical models or simulation, and are characteristic of problems for which techniques such as heuristic search and pattern matching are well suited. A key issue of importance to this research is the assumption that the above type problems are typical of most problem domains in that they do not rely on just a single expert for advice, but utilize the expertise of many knowledgable sources. But because each source can potentially conflict, the contemporary approach to building multiple expert systems has been to reach consensus before developing the knowledge base. This research presents a new approach to multiple expert systems (herein referred to as multiexpert knowledge systems) which addresses a limitation of contemporary expert system architectures. The developed architecture resolves the problem of accommodating multiple (potentially conflicting experts, and uses multiple lines of reasoning to advantage in learning new knowledge. The results demonstrate that such a system is capable of (1) obtaining useful advice from conflicting expertise and (2) learning new knowledge to be applied to the problem of interest--the evaluation of AI technology for process planning applications. When interfaced to an 'induction logic' program, the developed multiexpert knowledge system has the potential of becoming the first 'intelligent' expert system.