Earth Resources Laboratory :: Abstracts

An Expert System for Well-to-Well Log Correlation

David J. Lineman

Submitted to the Department of Earth, Atmospheric, and Planetary Sciences on December 10, 1986 in partial fulfillment of the requirements for the degree of Doctor of Master of Science in Geophysics

Abstract

This thesis presents a robust method for correlation of geologic sequences known as "Dynamic Depth Warping" (DDW). This method uses dynamic programming to find an optimal depth matching between two sets of log data. The advantage of dynamic programming over conventional spectral correlation methods is the ease at which local geologic knowledge can guide the matching process both before and during the correlation.

The Dynamic Depth Warping algorithm is implemented in a LISP program called COREX. This program uses a knowledge-based system to integrate all of the information usually available to geologists for log correlation. For this thesis the data includes digitized wireline logs, simple lithologic information, seismic lines, interpreted dipmeter logs, and local geologic knowledge. The system uses rules stored in the knowledge base to analyze these data and impose geologic constraints on the correlation algorithm.

We use the knowledge base of COREX for two purposes. First, the program performs an initial match of the wells based on the lothologies present, the scale of the correlation, and the depositional environment under consideration. Second, using the initial match to establish tie lines, the knowledge base analyzes the local geologic structure and log quality, and translates this information into constraints on the DDW algorithm. This guidance allows meaningful correlations in areas that were previously too complex because of geologic structure or large data volume.

We demonstrate the method with synthetic examples in which the program successfully correlates across geologic structure and pinch-outs. We also apply the program to field examples from West Africa and Turkey. In both cases, the automated correlation agrees very well with hand correlations provided by geologists.