Speaker
Description
Scope and Objective
Digital Rock technologies integrate 3D micro-CT imaging with computer vision and physics simulators to expedite and reduce the cost of rock property prediction. Key challenges here are the hardware-imposed limit on resolution and artefacts generated during the reconstruction of 2D projections to a 3D image. Consequently, boundaries between two phases appear as diffused zones rather than sharp edges, resulting in a further loss of resolution and decreased accuracy of property predictions, particularly in tight formations. This study presents a workflow to integrate techniques from continuum scale fluid flow to super resolve diffused interfaces into first-order accurate sharp approximations.
Approach and Methodology
A spatial fuzzy c-means algorithm is employed to determine the probabilities of each voxel's classification as either pore or mineral. Voxels situated at the interface exhibit low classification probability towards pore or mineral and are identified as uncertain interfacial voxels. The partial porosity of these interfacial voxels is calculated based on their grayscale values and the representative grayscale values of pore and mineral classes.
A method for piecewise iso-surface reconstruction is employed to generate interfaces that adhere to this partial porosity constraint. This iso-surface is a first-order approximation of the true interface. Super-resolution is performed by subdividing the original voxel into sub-voxels and classifying them based on their relative position to the iso-surface interface.
Results and Conclusion
To ensure robustness, the workflow was rigorously tested with a registered, multi-resolution dataset consisting of glass beads and clastic sandstone. Low and high-resolution image pairs, both focusing on the same field of view, were generated. The low-resolution image underwent super-resolution to achieve parity with its high-resolution counterpart, which served as the ground truth.
Validation metrics used included porosity, calculated with voxel counts; permeability, simulated using a multiple-relaxation time lattice Boltzmann simulation; and mercury intrusion capillary pressure curves, simulated via the maximal included sphere method. Each metric was computed for all images within the dataset and subjected to comparative analysis. Results indicated that the super-resolved image exhibited a deviation within 3 porosity units, 20% for permeability, and 4% for MICP curves. The close match for porosity and permeability metrics signifies the method's proficiency in accurately super-resolving the volume and tortuosity of the domain. Additionally, the close match for MICP curves validates the method's ability to preserve the spatial distribution and arrangement of grains within the samples.
Value
Digital Rock technologies offer significant advancements by providing geological information faster than conventional measurement techniques, at a substantially reduced cost. The proposed workflow enhances the precision of predictions without requiring extensive training data, which is typically unavailable for new formations. Additionally, this methodology enhances workflow efficiency by enabling accurate property predictions utilizing low-resolution images, which are quicker and more cost-effective to obtain. Furthermore, it broadens the applicability of Digital Rock technologies by overcoming hardware limitations to include tight rock formations that cannot be reliably resolved using micro-CT imaging techniques.
| Country | India |
|---|---|
| Acceptance of the Terms & Conditions | Click here to agree |
