Determination of Lithology from Well Logs Using a Neural Network
Samuel J. Rogers, J. H. Fang, C. L. Karr, David A. Stanley · AAPG Bulletin · 1992
ABSTRACT We have developed a computer program to automatically determine lithologies from well logs using a back-propagation neural network. Unlike a conventional serial computer, a neural network is a computational system composed of nodes (sometimes called neurons, neurodes, or units) and the connections between these nodes. Neural computing attempts to emulate the functions of the mammalian brain, thus mimicking thought processes. The neural network approach differs from previous pattern recognition methods in its ability to “learn” from examples. Unlike conventional statistical methods, this new approach does not require sophisticated mathematics and a large amount of statistical data. This paper discusses the application of neural networks to a pattern recognition problem in geology: the determination of lithology from well logs. The neural network determined the lithologies (limestone, dolomite, sandstone, shale, sandy and dolomitic limestones, sandy dolomite, and shale sandstone) from selected well logs in a fraction of the time required by an experienced human log analyst.