Print this pageField Scale Well Log Correlation Using an Expert 
System
R. A. Startzman, T. B. Kuo, S. A. Wong
Lateral changes in formation characteristics usually observed in well logs
can cause difficulty in correlating consistent lithofacies. A computer system
can be useful in this work by processing large amounts of information in a short
time. Besides shortening the work requirement in a large field study, computer
processing may also result in unique and useful correlations not achieved using
conventional and manual approaches.
Early computer approaches were derived from methods used to correlate traces
that were laterally closely spaced such as those from seismic sections and
dipmeters. These methods have been almost entirely numerically based. As lateral
spacing between traces increases, the amount of geological inference needed to
achieve acceptable correlation also increases. Thus correlation becomes
increasingly dependent on expert opinion and less dependent on numbers per se.
We have developed a prototype expert system capable of using at least a moderate
amount of geological inference to correlate traces between a pair of field-scale
wireline logs.
The system attempts to emulate a working geologist's or petroleum engineer's
approach to log correlation. Following digitization of all traces from suites of
logs run in each well, the first interpretation step is to identify individual
lithofacies or zones. The next step is to characterize each zone in sufficient
detail to enable correlation with the same zone in another well. Correlation is
then accomplished by application of a set of zonal comparison rules intended to
reflect the geological environment. Having achieved a tentative correlation, the
system then reexamines the original zonation scheme and attempts to
automatically improve the correlation until improvement is no longer possible.
This paper describes our experience to date with the system
. It shows how it
works with log data on a Symbolics Model 3640 minicomputer with KEE software.
The paper also discusses problems we encountered and the potential for
developing expert systems software for log correlation.
AAPG Search and Discovery Article #91038©1987 AAPG Annual Convention, Los Angeles, California, June 7-10, 1987.