Applying a Qualitative Modeling Shell to Process Diagnosis: The Caster System

摘要

The purpose of knowledge engineering is to develop partial-qualitative models for solving practical problems. These models--called knowledge bases in expert systems--must have appropriate diagnostic knowledge to deal with the real-world problems. In general, solutions to diagnostic problems can be either selected from a set of preenumerated alternatives (for known conditions) or constructed (for novel problems or those that combine multiple, interacting disorders in an unforseen way). While engineering design is often thought of as a constructive problem-solving process, diagnosis is typically thought of as a selection or classification problem. But the solution method is not inherent in the task itself. Instead, it depends on the problem solver's previous knowledge, requirements for customization, and the like. Nevertheless, useful programs can be developed that solve diagnostic problems by selection alone. We believe that starting with a well-defined classification procedure and a relational language for stating the classification model eases the development of a program that diagnosis by selection. To test this thesis. we built an expert system, called Caster, that addresses a particular diagnostic problem: malfunctions in industrial sandcasting. Our goal was to demonstrate that these control structures, developed for a medical diagnosis problem, are general and applicable to engineering applications.