Representing and Reasoning about Physical Systems from a Functional Viewpoint
Mahmoud Pegah, Jon Sticklen, James K. McDowell · 1993
A pivotal challenge of the 90s is to develop engineering methodologies which robustly address issues such as design for manufacturability, design to requirements, and conceptual design in engineering domains . Over the last several years, we have explored ways in which a functional reasoning viewpoint can be utilized to address such issues . In this report, we describe our extension and application of the Functional Modeling (FM) approach to represent and qualitatively simulate a significant portion of the fuel system of the McDonnell Douglas F/A-18 aircraft . Our general goals have been two fold: to test the scalability of our approach against a formidable real world problem, and to extend our approach to include a library facility from which standard parts may be instantiated into an evolving engineering design. Results support attainment of these goals . Knowledge acquisition for the research reported here centered on reverse engineering from a technical manual for the F/A-18 fuel system . Our KA experience on this project indicates that a functional viewpoint in general provides a strong backbone for reverse engineering . Reasoning explicitly about physical systems offers a way to circumvent the brittleness of reasoning systems built solely on associational knowledge . In addition, MBR is attractive because it captures an intuition that is especially cogent in engineering areas : in order to troubleshoot a device, or redesign a device to new specifications, or . . ., it is useful to know how the device works i.e., to represent and reason with a model of the device . There are two variations on the above theme. Each is involved in the larger picture of representing devices in the world and reasoning about them . One branch ofresearch has focused on how models of behavior are derived . The naive physics work of deKleer (deKleer & Brown, 1984), Forbus (Forbus, 1984), Kuipers (Kuipers, 1984), and Bylander (Bylander, 1986) exemplify this research. The second variation focuses on how models of behavior are used, as typified by the circuit diagnosis work of deKleer and Williams (deKleer & Williams, 1987) and Davis (Davis & Hamscher, 1988), and the function-based thrust of Chandrasekaran (Sembugamoorthy & Chandrasekaran, 1986), Franke (Franke, 1989), Sticklen (Sticklen,Chandrasekaran, & Bond,