Dongwon Lee, SFB 1313 doctoral researcher at the Institute of Applied Mechanics (CE) (research project B05), will defend his dissertation:
Title: "Advancing XRCT Techniques: From Enhanced Segmentation to Improved Temporal Resolution and Advanced Micromodel Fabrication for Pore-Scale Studies"
Date: 17 October 2024
Time: 3 pm CET
Venue: Pfaffenwaldring 9, 3.141, 70569 Stuttgart, Campus Vaihingen.
Abstract
This study explores innovative strategies to overcome spatial and temporal constraints inherent in lab-based X-ray computed tomography (XRCT) for pore-scale studies of porous media. The research focuses on three main aspects: enhancement of segmentation techniques, improvement of temporal resolution, and integration of advanced micromodel fabrication with XRCT imaging. Firstly, an exploration is conducted into the refinement of segmentation workflows tailored for micro-fracture networks within Carrara marble XRCT datasets, which are often characterized by low-contrast imaging and ambiguous boundaries due to apertures below the spatial resolution limit. Through a meticulous examination of various methodologies, including machine learning-based algorithms, significant advancements in computation time and accuracy are demonstrated compared to conventional segmentation workflows. Notably, machine learning methods exhibit superior performance, even in scenarios where images are contaminated with noise, showcasing their potential for enhancing segmentation outcomes. Secondly, the challenge of temporal limitations in XRCT imaging, especially during the study of dynamic processes within porous media, is addressed. Conventional XRCT technologies often encounter a trade-off between image quality, including spatial resolution, and scanning time. To mitigate this constraint, innovative workflows leveraging machine learning algorithms are proposed to augment temporal resolution. By capturing pore space alterations during phenomena such as Enzyme Induced Calcite Precipitation (EICP) with heightened fidelity, these approaches offer invaluable insights into the dynamic fluid flow dynamics that govern porous media behavior. Thirdly, an integrated methodology that combines 3D micromodel fabrication with XRCT imaging techniques is introduced. This comprehensive approach enables the design, fabrication, and validation of 3D micromodels that faithfully replicate the pore-scale characteristics of natural porous media. Leveraging stochastic reconstruction algorithms and advanced 3D printing technologies, highly detailed micromodels with unprecedented spatial resolutions are created. These micromodels serve as invaluable tools for validating numerical simulations, elucidating pore-scale phenomena, and advancing the understanding of fluid dynamics in complex yet well-controlled porous media systems. The outlook suggests further advancements through the integration of multi-modal techniques, machine learning, and expansion of training datasets to overcome current limitations, offering unprecedented insights into complex fluid flow phenomena within porous media.