Building Expert Systems for Controlling Complex Programs

Application Systems

Volume

Proceedings of the AAAI Conference on Artificial Intelligence, 2

Knowledge/Databases

Production rule schemes have proven quite effective in concisely representing expert knowledge in several application areas. Yet, there are many problems for which one would like to take advantage of additional knowledge that may not be easily represented in these surface level models. One class of problems of particular practical interest are those in which we would like to have a computer-based system give interactive advice on how to control and interpret results from a set of complex and interrelated applications programs. The advice may refer to interpretations of current results, possible experiments that should be performed with the help of the applications programs, and indications of inconsistencies in specific analytical procedures and in problem solving sequences followed by the user. In the present paper we report on our experiences in designing an expert system (ELAS), of the type described above, for well log analysis in oil exploration and production. We have integrated a production rule advice model (using the EXPERT system) with existing Amoco software for well-log analysis and display. In doing so, the original system for well-log analysis was reorganized so that its use could be monitored and controlled, and its knowledge structured according to the types and sequences of methods used by expert analysts. By varying the assumptions and parameters used in the different individual analyses, our goal is to make available interactive interpretations of the alternative approaches that an expert might take to a complex problem of well-log analysis.