Unsupervised Classification of λρ-μρ Attributes Derived From Well Log Data in the Barnett Shale

Wallet, Bradley C. (University of Oklahoma) | Altimar, Roderick P. (DrillingInfo) | Slatt, Roger M. (University of Oklahoma)


Using an unsupervised learning approach to data clustering, we allow our data to speak for themselves, providing insight into the underlying data distribution. We then look at the rock proprieties of the discovered clusters to better understand the nature of the data. Introduction The Barnett Shale is an organically-rich and thermally-mature rock deposited during the Mississippian time ( 340 Ma) in the Fort Worth Basin, characterized by low average permeability (70 nD) and porosity (6%) distributed in a variety of depositional facies (Deacon, 2011). The Barnett Shale is the primary source rock for oil and gas produced from the Paleozoic reservoir rocks in the basin (Jarvie et al., 2007). Clay, quartz, and carbonate are the main minerals in the Barnett Shale, and their content are highly variable over the whole section (Karastathis, 2007; Kale, 2009).