The Dipmeter Advisor System - A Case Study in Commercial Expert System Development

Abstract

We discuss commercial expert system development, using our experience with the Dipmeter Advisor system as a case study. While the data is too sparse for definitive results, several ideas have emerged as important and suggestive as guidelines for subsequent commercial expert system undertakings . During the past four years, the Dipmeter Advisor system has migrated from an initial experiment in application of expert system techniques in well-log interpretation to a candidate commercial interpretation system. The system has undergone substantial change: It has been implemented in different configurations, in different languages, and on different computer systems. Our ability to experiment with the system has been greatly enhanced by the tools and ideas of rapid prototyping. We have also observed an oscillation in thrust over time between (i) expansion and change in the domain knowledge, and (ii) selection and design of appropriate expert system tools. Finally, we have found several of the maxims of expert system development to be valid, but question a number of others.