Speaker
Description
Conducting coreflooding experiments with high-resolution imaging has become a key approach for understanding the physical phenomena governing CO2 storage in underground porous geological formations. These experiments focus on displacement and trapping mechanisms of immiscible phases, which are predominantly capillary-dominated. A significant advantage of pore-scale imaging lies in its ability to capture the interface between any two phases, enabling the in-situ measurement of contact angles and curvature (capillary pressure). These measurements are crucial for quantifying how rock heterogeneity influences flow, as capillary forces are governed by wettability and pore space structure. Contact angles also serve as vital input parameters for pore-scale simulations, where pore-by-pore contact angles dictate the sequence of pore displacement. Additionally, experimental capillary pressure measurements validate the predictive capabilities of simulations.
However, both contact angle and curvature measurements are highly sensitive to image processing methods. Current approaches often derive these properties from segmented images, which are prone to high error rates. Segmentation methods eliminate intensity gradients that indicate phase transitions (interface boundary), leading to the loss of critical features needed to preserve the true geometry of interface boundaries. Furthermore, when interfaces are extracted on discrete surfaces with stair-step geometries, smoothing operations intended to improve surface quality can inadvertently introduce curvature errors. Even minor errors can significantly amplify the margin of error for contact angle measurements.
This study addresses these challenges by developing automatic, open-source tools to improve curvature and contact angle measurements. The approach enhances the extracted interface surface between phases by leveraging sharp intensity gradients from grayscale images. A neural network is trained to recognize interface geometries under various wetting conditions, such as minimal surfaces, concave menisci, and convex menisci. Training is conducted using synthetic images with analytically known curvatures and real two-phase flow images from rock samples. The tools also refine geometric contact angle measurements by analyzing three-phase loops (fluid-gas-solid contact regions) and extracting angles based on localized interface geometry, rather than relying solely on contact node information. By advancing these methodologies, this study improves the accuracy and reliability of key pore-scale measurements, providing more robust data for simulations and enhancing our understanding of multiphase flow in geological storage systems.
| Country | UAE |
|---|---|
| Student Awards | I would like to submit this presentation into the student poster award. |
| Acceptance of the Terms & Conditions | Click here to agree |
