Toward Reasoning Methods for Automatic Mechanical Repair

A knowledge representation scheme, QUORUM (Qualitative reasoning Of Repair and Understanding of Mechanisms), has been constructed to apply qualitative techniques to the mechanical domain, which is an area that has been neglected in the qualitative reasoning field. In addition, QUORUM aims at providing foundations for building a repair expert system. The problem in constructing such a representation is the difficulty of recognizing a feasible ontology with which we can express the behavior of mechanical devices and, more importantly, faulty behaviors of a device and their causes. Unlike most other approaches, our ontology employs the notion of force and energy transfer and motion propagation. We discuss how the overall behavior of a device can be derived from knowledge of the structure and the topology of the device, and how faulty behaviors can be predicted based on information about the perturbation of some of the original conditions of the device. Necessary predicates and functions are constructed to express the physical properties of a wide variety of basic and complex mechanisms, and the connection relationships among the parts of mechanisms. Several examples analyzed with QUORUM include a pair of gears, a spring-driven ratchet mechanism, and a pendulum clock. An algorithm for the propagation of force, motion, and causality is proposed and examined. Comments University of Pennsylvania Department of Computer and Information Science Technical Report No. MSCIS-87-64. This technical report is available at ScholarlyCommons: http://repository.upenn.edu/cis_reports/629 TOWARD REASONING METHODS FOR AUTOMATIC MECHANICAL REPAIR Pearl Pu MS-CIS-87-64 GRAPHICS LAB 16 Department of Computer and Information Science School of Engineering and Applied Science University of Pennsylvania Philadelphia, PA 191 04-6389