Issues involved in developing a failure analysis expert system for metallic materials using rule and case-based reasoning

Several issues are discussed to aid the development of rule and case-based reasoning expert systems (ES.) for the failure analysis of metallic materials. This approach applies to generic failure (fatigue, corrosion, deformation, erosion, etc.) diagnosis and focuses on the structure, planning and elicitation of failure analysis knowledge related to metallic components. An effective failure analysis ES., based solely on rules or past case histories, entails either a complex set of rules or many cases. The best methodology is a hybrid approach of rules and case histories as this mimics the approach used by consultants [1, 2]. A structured plan is very important; it should include the scope, semantics, a questioning strategy and deductive rules. This necessitates understanding the main engineering failure modes, and how these are associated with different materials, processing methods, and components. Developing the cases and remedial actions is a matter of collecting together information, which is considerably easier than developing the semantics, the questioning strategy and the deductive rules. An effective questioning strategy involves: selecting the description, order, weighting, types and number of questions by which to ascertain the most important facts. Finally, it is difficult to balance between just enough information needed for a non-expert and not an excessive amount of unnecessary information for an expert in failure analysis. This can be overcome somewhat by using hypertext.