Expert Systems in Exploration: Can They Be Cost-Effective?

R. P. J. Lyon

PROSPECTOR is the best-known application of an expert system in exploration. Others exist for gamma-ray well logging analysis but are in general company-restricted and not in the open literature. PROSPECTOR, however, is comprised of a large set of elegant subprograms, each designed for a specific goal--generally for hard minerals. The program is expensive, costly to run, and requires a mainframe (usually a LISP machine) for operation. Recently, a microcomputer-based version (u-PROSPECTOR) has become available, but it still follows the formal artificial intelligence (AI) syntax of PROSPECTOR.

In the Remote Sensing Laboratory at Stanford, we have been experimenting with "low-end" ($100-$1,000) AI programs. The development of these has been driven by (1) the explosion of availability of microcomputers and (2) the realization by developers that the marketplace has many more Fortran and C-language machines available than the dedicated (and expensive) LISP units.

This paper will discuss those commercially available, low-priced AI shells as applied to several of the simplest exploration problems--spectral pattern recognition and "texture" in radar imagery--and extrapolate their usefulness to more complex decision-making steps in exploration practice.

AAPG Search and Discovery Article #91038©1987 AAPG Annual Convention, Los Angeles, California, June 7-10, 1987.