Computer-Aided Lithostratigraphic Correlation Using E-Logs

FANG, J. H., H. C. CHEN, A. W. SHULTZ, and W. MAHMOUD, University of Alabama, Tuscaloosa, AL

Various attempts have been made in recent years toward computer-assisted well-log correlation in hopes of removing subjectivity which is inherent in manual correlation. However, this application of computer technology has achieved only limited success and popularity among geologists, due partly to the use of nongeologic jargon describing machine correlation and partly due to the publication of these papers in journals seldom read by geologists.

We have coded a computer program which performs machine correlation, allowing human intervention at several stages. First, the interval to be correlated is segmented (by machine or manually) into zones, then the four attributes (depth, thickness, log amplitude, and log shape) of each zone are used to perform matching. In essence, this is a pattern recognition approach, but not by comparing just one feature at a time, but by considering all attributes jointly and simultaneously in a "gestalt" manner.

The program starts by digitizing logs, smoothing log signatures, and carrying out zonation between marker beds. Preliminary marker beds are either predetermined by the geologist, or are the result of iter-ative matching. For each pair of zones in two different wells, a difference is computed by comparing strings of attributes. In this manner, each pair of wells produces a difference matrix with one cell for each combination of zones. Dynamic programming is then used to trace the path of minimum total difference, designated by a P-matrix. These P-matrices may reveal certain geologic structures which are helpful not just in correlation but also in structural interpretation.

AAPG Search and Discovery Article #91004 © 1991 AAPG Annual Convention Dallas, Texas, April 7-10, 1991 (2009)