GEOSCAN Search Results: Fastlink


TitleQuantitative analysis of statistical properties of organic-rich mudstone using large field-of-view SEM images
AuthorBizhani, MORCID logo; Ardakani, O HORCID logo; Knapp, L J; Akai, T
SourceJournal of Natural Gas Science and Engineering vol. 95, 104238, 2021 p. 1-13,
Alt SeriesNatural Resources Canada, Contribution Series 20210296
Mediapaper; digital; on-line
File formatpdf; html
NTS83K/05; 83K/06; 83K/07; 83K/08
Lat/Long WENS-118.0000 -116.0000 54.5000 54.2500
Subjectsfossil fuels; sedimentology; Science and Technology; Nature and Environment; Upper Devonian; bedrock geology; lithology; sedimentary rocks; mudstones; shales; petroleum resources; hydrocarbons; gas; oil; condensate; scanning electron microscope analyses; statistical analyses; pore structure; pore size; Duvernay Formation; Western Canada Sedimentary Basin; Phanerozoic; Paleozoic; Devonian
Illustrationslocation maps; geoscientific sketch maps; tables; photomicrographs; bar graphs; plots; schematic representations
ProgramEnergy Geoscience Clean Energy Resources - Decreasing Environmental Risk
Released2021 09 10
AbstractThis paper presents a quantitative and statistical analysis of large field-of-view (FOV) scanning electron microscope (SEM) images of organic-rich shale samples. The samples are from the Upper Devonian Duvernay Formation at the onset of the condensate hydrocarbon window. Our data set contains 12 mosaic SEM images which were each obtained by stitching 100+ high-resolution SEM images. The goal was to establish a basis for deriving meaningful statistical properties of the pore space that can be used for characterization and possibly up-scaling purposes.
The results show that pores smaller than 100 nm in diameter are the predominant type in the studied Duvernay shale samples in terms of the number of pores. However, chord length distribution analysis shows that decompression/desiccation cracks can heavily skew statistical properties such as the first moment of the distribution. Two characteristic length scales were computed by using the two-point correlation function to show the disparity of scales in the same sample imaged at different resolutions. Our analysis indicate that there is a scale dependency on the computed properties of the pore space, and statistical convergence cannot be claimed without having a multi-scale approach for characterization.
Summary(Plain Language Summary, not published)
Unconventional resources, including shale oil and gas, have become the primary focus of the energy industry. Comparing to conventional reserves, shales have a much smaller pore size, which makes them hard to characterize. Scanning Electron Microscopy (SEM) is a powerful visualization method for studying organic-rich shales. However, extracting meaningful quantitative data from SEM images is not straightforward and suffers from a few drawbacks. In this study, we attempt to improve data extraction from large SEM images. In a given SEM image, separating different structures is called segmentation and is usually done using thresholding methods. We present an improved method for segmentation through the use of machine learning methods. We present several probabilistic functions to better study the properties of organic-rich shales. Ultimately, this study makes use of computational power to reduce image data into useful quantitative data for the characterization of porous media.

Date modified: