Dynamic mechanical system design using qualitative and symbolic reasoning

Technical Report · Tue Dec 06 04:00:00 EST 1988

OSTI ID:6525353

Kim, Young-Jin; Agogino, A M

This report describes the results of a three-month project to perform preliminary research for use in preparation of UCB Campus- LLNL collaborative research proposal. The goal of this research is to develop and implement artificial intelligence methods for the design, modeling and analysis of dynamic physical systems. Bond graphs and a library of dynamic system primitives form the basis of the knowledge representation scheme. Bond graphs abstract physical systems into basic elements that represent localized dynamic properties of small parts of the system. This research deals with mechanical systems, in which there are natural hierarchies of subsystems which are in turn comprised of primitive components. The subsystems developed in this research are divided into three classes: resistance fields, inertial fields and capacitance fields. Basic categories from the mechanical systems domain are examined to derive the characteristics of each subsystem. We propose that future research begin with the development of a library of natural modeling primitives for physical subsystems along with rules for creating a network from such primitives. We modify Karnopp and Rosenberg's mechanical translation construction method to generate the automatic bond graph representation of mechanical system in a rule/frame-based format. We outline a knowledge-based methodology to define a bond graph description of complex dynamic physical systems from an integration of simpler primitives and component classes. We conclude with a framework for integrating earlier work in predicting and directing the behavior of dynamic systems using qualitative and symbolic reasoning for providing a mathematical description of the behavior of the system along with recommendations concerning the design parameters. 7 refs., 2 figs.

• Research Organization: Lawrence Berkeley Lab., CA (USA)
• DOE Contract Number: W-7405-ENG-48
• OSTI ID: 6525353
• Report Number(s): UCRL-21179; ON: DE89006771
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990210* -- Supercomputers-- (1987-1989)
ARTIFICIAL INTELLIGENCE
COMPUTER-AIDED DESIGN
CONTROL SYSTEMS
EXPERT SYSTEMS
FLUID MECHANICS
HYDRAULICS
MECHANICS