Geostatistics and Artificial Intelligence Applied to Stratigraphic Correlation
Michael Edward Hohn, Maxine V. Fontana
Programs using artificial intelligence techniques and geostatistics have been written to match lithologic units on driller's logs with formation tops determined from geophysical logs. These programs form part of a larger system for storage, retrieval, and manipulation of oil and gas data.
Records on most older wells include only a driller's log with lithologic data. The few formation tops reported may have resulted from incorrect correlations, yet the geologist may want to use these wells for drawing maps or compiling statistics. Given a well location, a program computes kriged estimates of formation tops from nearby wells represented in the data base by geophysical logs. A second program matches these estimates with lithologies given on the well's driller's log. This program uses a set of rules on allowable matches, and a simple rule processor similar to an expert system. Rules can be used to give formation-to-lithology matches and, in the few cases where actual formations are named on the driller's log, attach confidence levels to the reported correlation and alterna ive matches.
These programs allow a geologist to use a data base of lithologic information allied with reliable geophysical data from a small subset of wells. Geostatistical estimation and artificial intelligence lead to an incremental refinement of the data base through addition of geophysical logs from new wells and rules for matching formation tops with lithologic units.
In research supported by the Gas Research Institute, this procedure has been successfully applied to a previously undivided section of several thousand feet of Devonian shales.
AAPG Search and Discovery Article #91043©1986 AAPG Annual Convention, Atlanta, Georgia, June 15-18, 1986.