Analysis of High-Resolution Digital Thin Section Images – Applications for Rock-Typing

Abstract

Rock-typing approaches to identify potential reservoir analogs have been in use throughout industry for over 30 years. Each approach seeks to leverage visual estimates of rock property characteristics (lithology, porosity, grain size, sorting, consolidation, shaliness, pore types and cements) from cuttings, thin section or poorly archived core by searching for matches of these parameters in a database containing physically measured values. Matches usually return additional values including total porosity, effective and absolute permeability, electrical properties, capillary pressure and geomechanical properties originally derived from fresh conventional core. The matching samples provide a foundation for the evaluation of a reservoir in which fresh conventional core is unavailable. However, visual estimation of the rock property characteristics used in the database search can be subjective and lead to biased results. High-resolution thin section images, coupled with machine learning technology, offer an objective alternative to visual estimation approaches. Multiple representative reservoir thin sections are created from the same samples on which physically measured petrophysical properties were determined. Each thin section is scanned in high-resolution using both plain and polarized light, and the resulting image files are used to create a training set, coupled with the physically measured core data, for the creation of a cognitive model. Once created, such a model can be quickly and efficiently applied to determine reservoir rock types using only representative thin sections from cuttings, undersized rotary sidewall cores or poorly archived conventional core.

AAPG Datapages/Search and Discovery Article #90327 © 2018 AAPG Middle East Region GTW, Digital Subsurface Transformation, Dubai, UAE, May 7-8, 2018