A Particle-Based Image Segmentation Method for Phase Separation and Interface Detection in PIV Images of Multiphase Flow in Porous Media
Abstract
Multiphase flow in porous media is ubiquitous in various industrial and environmental processes, including enhanced oil recovery, ground water remediation and carbon capture and storage. A comprehensive understanding of the fundamental mechanisms in multiphase flow in porous media is highly critical in order to guide and improve those many applications. In that regard, pore-scale velocity measurement using the advanced technique called microscopic particle image velocimetry (micro-PIV) represents a key component in such research efforts. In micro-PIV, tracer particles are seeded in one or more phases of the flow, illuminated by a light source (e.g., laser) and then imaged by a camera. By performing proper phase discrimination and image cross-correlation, detailed velocity fields in one or more phases can be obtained, which has helped a great deal further our understanding of various pore scale events, such as Haines jumps and shear-induced recirculation. To conduct PIV measurement in multiphase flows, one crucial step and a big challenge is to perform proper phase segmentation based on images of discrete tracer particles. Additionally, fluid-fluid interface detection in such images is important as the interfaces and their dynamics are not only critical for understanding many momentum and mass transfer processes, including dissolution, exsolution and reactions, but also key to more fully understanding fluid migration from a fluid mechanics perspective. Despite numerous methods proposed previously for image segmentation, including color, intensity, feature shape and feature size separations, none of them is able to produce satisfying segmentation results for PIV images. To that end, we report a novel method for precise phase separation and interface identification based on particle detection and triangular meshing. In this method, tracer particles in all seeded phases are first identified, and the coordinates of the particle locations are then used to construct a 2D unstructured mesh, where the triangular grids provide the basis for phase separation and the outer edges provide basis for fluid-fluid interface detection. This method has proven especially powerful to process PIV images and quantify flow and interface dynamics in multiphase flow in porous media.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMH015...04L
- Keywords:
-
- 1807 Climate impacts;
- HYDROLOGY;
- 1832 Groundwater transport;
- HYDROLOGY;
- 1865 Soils;
- HYDROLOGY;
- 3994 Instruments and techniques;
- MINERAL PHYSICS