Artificial intelligence techniques to the interpretation of geophysical measurements

Abstract

SummaryIntegration of geology and geophysics thinking requires a common earth model, that accommodates, with errors, all the features from the geophysics interpretation products, topological rules from geology, field mapping and drill logs, and the link between the two via physical rock property estimates. Automation of intelligent search, inference engines to this problem involves 200+ individual processes, to yield a family of plausible models. The true limiting factor is managing technical complexity and communicating to the geoscience team, not compute power.All models are wrong and are destined to be replaced as further data or insights are gleaned. So, taking too much time trying to create the ultimate “correct” interpretation is wasteful, hence the need for automation.An example of a simple pattern recognition technique to locate kimberlite pipes that have a magnetic and/or gravity response is given. This uses a range of pipe diameters, depths, topography, density and susceptibility, and variable directions of remanent magnetization vectors, in an automated manner.