This paper describes an Expert System developed to predict whether a pay zone will produce hydrocarbon or water. This Expert System was applied to the Caballos and Villeta Formations in the Putumayo basin and the Mirador Formation of the Cusiana Field, both in Colombia. Log analysis "rules" relating the combination of logs (SP, GR, caliper, resistivity, density, neutron, sonic, etc.) and basic information are included in the knowledge base.
The Expert System uses the Kappa-PC Shell and is composed of three sub-systems: 1) an inference engine, 2) a knowledge base, and 3) a user interface. We also consulted human experts to formulate and expand the knowledge base. Additionally, each formation was characterized using a Log Interpretation System (QLA2). The results from each characterization were used to develop cross plots. The QLA2 cross plots provided practical correlations from which to formulate "rules" for the Expert System.
Logs from nine wells in the Cusiana Field were used to evaluate the Expert System. The recommendations of the Expert System agreed with conventional log analysis theory as well as with the known field results. Furthermore, it minimized the time and cost required to evaluate the pay zones. In conclusion, this Expert System can be used to predict whether or not a zone will produce oil or water with a high degree of accuracy.