Enhancing Petrographic Analysis Through Data Fusion

Abstract

Despite recent advances in characterization technology, there are still limitations in accuracy and throughput potential when characterizing textural relationships and diagenetic history of thin-bedded and lithic, clay-rich, and/or feldspar-rich reservoirs. These characteristics have large impacts on accurately assessing porosity and permeability profiles for upscaling into reservoir models. Several different methods, including scanning electron microscopy (QEMSCAN), mercury injection capillary pressure, and laser grain-size analysis are routinely used to quantify grain size, porosity, and mineralogy on thin-sections; however, each of these methods is associated with drawbacks on accuracy, time or both. One of those modalities alone is not adequate to describe most rock types. However, the continuous increases in computing power and the more common availability of advanced imaging technology and processing software has provided the opportunity for more accurate and statistically robust extraction of geological parameters from rock sample material. This presentation aims to show how fusing textural and mineralogical information derived from back-scattered electron (BSE), cathodolumenesce (CL) and energy-dispersive spectroscopy (EDS) improves rock characterization by obtaining measurements of grain size, shape, roundness, and composition in order to understand the diagenetic history Using the BSE image, grains can easily be distinguished from pore space and organics at nanometer resolutions. Some minerals, such as a pyrite, can be easily identified from a silicate or carbonate matrix due to the relationship of BSE intensity with atomic number. However, many minerals have a substantially different chemistry but indistinguishable BSE singles due to similar mean atomic numbers. One example is to compare quartz (SiO 2 ) to albite (NaAlSi 3 O 8 ), which have mean atomic numbers of 10 and 10.03 respectively. The BSE image alone can provide a statistical approximation of grain size distribution, shape statistics, and roundness measurements. Automation of those measurements is possible, but many methods are not robust. Grain boundaries are often visible, but automated grain separation may be deterred by numerous internal grain fractures or bad approximations on non-spherical grains when using a chamfer-distance map and watershed separation algorithm. Despite these limitations, the wealth of data extracted from BSE images provides the contextual backdrop for the workflow provided in this presentation.The CL image also has a resolution on the order of nanometers, but is more sensitive to chemistry and microstructure of the grains as the signal is recorded against the luminescence of the material caused by the bombarding electrons. The CL image specifically allows investigation of grain overgrowth which may be of the same material, a solid-solution with a different material, or a completely different material than the respective grain. Separation of the grain with the overgrowth and cement allows for better insight into the ...