Knowledge acquisition in a machine fault diagnosis shell

Abstract

The knowledge acquisition tools and techniques discussed in the literature deal primarily with the acquisition of human expertise applied in a particular problem domain. The acquired expertise generally includes both the decision making strategies of the human expert and the descriptions of the application problem. This acquisition process can become quite repetitive and time consuming when developing a number of application expert systems which use similar problem solving expertise but differ only in their application details. In this paper, we address this issue by discussing the details of the knowledge acquisition system we have designed and developed for use in a customized machine fault diagnosis shell. The knowledge acquisition system functions as a module of the diagnosis shell and acquires details of application machinery for which diagnosis expert systems are to be developed. The acquired application specific knowledge is combined in the shell with predefined generalized diagnosis strategies and application diagnosis expert systems are rapidly generated. The designed knowledge acquisition system has been implemented using Lisp on the Symbolic Lisp machine and has been validated by acquiring and verifying the design descriptions of a Cincinnati Milacron 786 robot. In this paper, the issues related to the development of a knowledge acquisition system for a customized problem solving shell, such as the machine fault diagnosis shell, and the details of the implemented knowledge acquisition system are discussed.