Professor, Royal Military College, Kingston, Ontario
Pages: 15 Published: Jan 1995
An ideal expert system should closely reproduce the functions normally performed by the original source expert(s). But the complex transformation of a lifetime of industrial expertise into conveniently accessible software requires the willing and active participation of domain experts whose knowledge is rarely organized for such a transformation. The matching of logical and meticulous knowledge elicitation with memory-based judgments and expertise has often proved to be a difficult task. This paper describes an approach to construct a knowledge elicitation shell specifically adapted to the field of corrosion-related problems and expertise. The knowledge elicitation shell was structured along the lines of what has recently become the framework for the transfer of corrosion information to management and design engineers. The recent work published by Professor R. W. Staehle on the general topic of lifetime prediction and the impact of stress corrosion cracking on materials was used for this task. One obvious advantage of using an established formalism is the availability of supporting documentation and background material. The advantage of using Professor Staehle's work in particular is the clarity of his points of view that reflect a solid career in materials science and engineering. In this paper, the environment-induced cracking of high-strength aluminum alloys is used as an example to illustrate how the elicitation shell could be activated. It is believed that, by using the elicitation shell, human experts will have to rationalize some aspects of their expertise and subsequently become able to establish new links between facts and corrosion data. The process of building tables of certainty factors during the shell operation and interaction with users is also discussed.
lifetime prediction, corrosion, stress corrosion cracking, expert systems, knowledge-based systems, knowledge elicitation, computerized material property databases, computers, data systems, materials databases
Paper ID: STP15410S