An integration of object-oriented knowledge representation and rule-oriented programming as a basis for design and diagnosis of technical systems

Abstract

We describe progress made within the FORK project, whose goals are the implementation of a primarily object-oriented knowledge representation system and its application to the design and fault diagnosis of technical systems. Whereas the kernel of the FORK representation system is completely object-oriented, the system as a whole is supposed to integrate a variety of different programming styles. In the following, an extension for rule-oriented programming is described, which raises the descriptive power of the FORK system beyond that of LOOPS. As an application of the rule-oriented component, a constraint language has been implemented which plays an important rule in our approach to the design and fault diagnosis of technical systems. In parallel to the implementation of the FORK system, a first study in the field of diagnosis has been conducted, aiming at a clarification of the basic problems and representational needs. After considering more traditional rule-based approaches to the diagnosis problem, we concentrated on an approach known as “based onstructure andbehavior”. Starting with an algorithm to diagnose multiple failures in electronic circuits, considerable extensions had to be made for the more complicated case of electromechanical systems. Since the object-oriented implementation of this diagnosis system, DIAGTECH, had to obey the restrictions of a PC, only a subset of the features offered by the FORK system was available. DIAGTECH is a hybrid system, because it also supports the rule-based style of diagnosis, for which our logic-based “expert system shell” DUCKITO is used as a subsystem. The next steps in the FORK project will include the development of a general logical framework, comprising a logical reconstruction of object-centered representations, retrieval of complex descriptions by unification, and deductions on structured objects. The problem of nonmonotonicity will be dealt with on the meta level by a module similar to deKleer's ATMS — which was not available for DIAGTECH. Further progress shall be achieved by concentrating on a general treatment of the problem of time in modelling technical systems which is to our opinion one of the most important issues.