| dc.description.abstract | The physical properties of shale are fundamentally controlled by its microstructure. Connectivity of various components in shale is an important property that governs the transport of mass, energy and momentum. Quantifying connectivity of components is a critical aspect to understand the microstructure of shales. Scanning electron microscope (SEM) imaging technique is a popular technique to capture the microstructure of materials. Before quantifying connectivity of components captured in the SEM image, different components in SEM images need to be identified and segmented. In the first part of this study, an automated SEM-image segmentation workflow involving feature extraction followed by machine learning is developed and tested on SEM images of shale. The proposed segmentation workflow is an alternative to classical threshold-based and object-based segmentation. Four components, namely pore/crack, pyrite, organic/kerogen, and rock matrix including clay, calcite and quartz, are automatically identified and segmented. The performance of the automated SEM-image segmentation workflow, quantified in terms of overall F1 score, on the validation dataset was higher than 0.9. In the second part of this study, five different connectivity-quantification metrics, namely two-point statistical function (S2), two-point cluster function (C2), cluster size distribution, travel times computed using fast marching method (FMM), and Euler’s number, are tested on SEM images of shale. First, the relationships between the connectivity and the responses of the five connectivity-quantification metrics are determined and validated by statistical analysis on a synthetic dataset of binary images, which contains six types of connectivity from the lowest to the highest. Second, such relationships are directly applied to quantify the connectivity of organic/kerogen and pore/crack components in the SEM images of shale. | en_US |
