Reference: Mohammed, J. L. Combining Experiential and Theoretical Knowledge at Multiple Levels of Granularity in the Domain of Semiconductor Fabrication Processes. KSL, November, 1989.
Abstract: Engineering knowledge incorporates both empirical and theoretical knowledge, and exists at multiple levels of abstraction, approximation and granularity. Thus, representing and reasoning with engineering knowledge requires the ability to represent and reason with experiential knowledge and theoretical knowledge at multiple levels of granularity, and also meta-level knowledge concerning when it is appropriate to apply each type of knowledge, and how to translate between them. Three existing systems, PIES, DAS and CPS, encode knowledge about a single engineering domain: semiconductor fabrication. These systems encode, respectively, experiential knowledge concerning how to diagnose failures in manufacturing, and theoretical models at two different levels of granularity concerning how the manufacturing process works (which also support the same diagnostic task). It is proposed that a single system be constructed that will integrate the representations and reasoning mechanisms of the PIES, DAS and CPS systems. The system's primary high-level task will be diagnosis of problems in the manufacture of integrated circuits. The primary purpose will be study the issues surrounding the integration and utilization of knowledge at different levels of granularity. A secondary purpose will be to investigate an approach to the 'knowledge re-use' problem based on a taxonomy of reasoning tasks.
Full paper available as hqx.