Expert Geoscience Machine of Physics-augmented Computer Vision for Digital Rock Typing
Omar Al-Farisi
When an expert geoscientist spends months or years classifying rocks and quantifying its chemical and physical properties using thin section and core plugs, an expert geoscience machine using μCT images only needs few days to achieve the same.
Rock type is rocks with similar properties. Geoscientists perform conventional and special core analysis measurements, including grain density, porosity, permeability, capillary pressure, and relative permeability, then analyzing the measurements to produce the rock types. Geoscientists are looking for patterns in the data and then linking these patterns to lithofacies grouped out of the thin section analysis, a work that can span over several months or years to produce results of varying quality. We propose an Expert Geoscience Machine (EGM) by utilizing a machine learning approach of physics- augmented computer vision that can help geoscientists be more efficient, consistent, and produce higher-quality rock classes.
