(540b) Generating Realistic 3D Volumes to Mimic Pore Structures in Unconventional Reservoir Rocks

Unconventional reservoirs are composed of chemically heterogeneous mixtures of minerals, organic components, and pores. In addition, the size of the organic and mineral components varies at the sub-micron scale while the pores are typically on the order of 10s to 100s of nanometers. This nanoscale heterogeneity poses challenges in measuring petrophysical properties that affect flow and data interpretation with reference to the changing rock structure. High resolution imaging using two-dimensional scanning electron microscopy (2D SEM) and three-dimensional focused ion beam SEM (3D FIB-SEM) is commonly used to characterize the nanoscale pores responsible for hydrocarbon storage and flow. Specifically, 3D FIB-SEM volumes are used to analyze pore connectivity and flow using Digital Rock Physics (DRP) simulations. Rock volumes are collected by imaging a 2D region of interest (ROI) followed by removal of the imaged layer. As this process is repeated, it generates a series of image slices which comprise a 3D rock volume (typically 1000 x 1000 x 1000 pixels) that is later processed to separate the pores used in DRP flow simulations. The time required for sample preparation, imaging, and data analysis limits the number of 3D FIB-SEM volumes collected per sample and the limited field-of-view (FOV) makes collecting a representative volume of these heterogenous nanostructured materials infeasible.

One solution to this problem involves generating digital pore volumes (DPVs), as this can reduce the amount of experimental data collected to effectively sample the flow properties using DRP. However, the ability to mimic amorphous nanostructures with rapidly changing sizes and shapes can be a challenge. Methods that rely on generating DPVs using spheres inadequately represent real pore structures, and methods that utilize orthogonal views from 3D laboratory data to generate 3D objects still require time-consuming 3D imaging and processing. Here, we discuss a methodology to realistically modulate pore structures captured from large FOV 2D SEM images to generate many 3D DPVs in a much shorter time frame, which can be used in DRP flow simulations. First, a 2D ROI seed is selected from a large FOV 2D SEM image which contains >100,000 pores. Morphological processes, such as erosion and dilation, in addition to probabilistic pore terminations and reseeding are used to modulate pore structures in the ROI seed and generate subsequent slices, allowing for realistic changes in pore shapes and sizes. Initial comparisons show that the 3D FIB-SEM volumes and DPVs are structurally similar, and DRP-predicted permeability results from DPVs are on the same order of magnitude as those from 3D FIB-SEM volumes of a similar porosity in the first slice. In summary, the method can generate 100s of realistic DPVs in the time frame required to prepare, collect, and segment a single 3D FIB-SEM image. This allows for more effective permeability sampling while also providing a tool to develop pore feature to permeability correlations, which can be challenging in a data-limited domain. Overall, better permeability sampling and developing permeability correlations can improve our understanding of flow in these structurally complex reservoir rocks. Further steps include developing methods that include additional processes, such as axial rotations and edge interpolation, as well as expanding the method to incorporate objects other than pores, such as organic matter.