Speaker
Description
Microscopic multiphase fluid dynamics in porous media form the basis of various macroscopic phenomena in geological and industrial applications. Dynamic X-ray micro-CT enables us to study how fluid distributions evolve in 3D at the pore scale in opaque samples without interfering with the system, and has thus become a key tool for in-situ visualization of dynamic multiphase flow processes in porous media. However, a key challenge in this method is the relatively low achievable time resolution due to the time needed to acquire a sufficient number of projections (radiographs) from multiple angles to reconstruct a 3D volume. If a dynamic CT dataset is treated as a time sequence of independent 3D volumes (further called frames), improving the time resolution comes at the cost of low-quality images.
To improve the achievable time resolution in micro-CT imaging of flow in porous media without reducing image quality, we introduce here a novel reconstruction methodology named DYRECT [Goethals et al. 2025]. Rather than reconstructing the 3D geometry of the sample for each global time frame, this technique specifically aims to retrieve local changes, pinpointing these events in space and time. This can be stored as a memory-efficient dataset of parameters, irrespective of the original frame rate, that describe how each voxel in the sample changes over time. This representation is inherently coupled to the discrete and sparse nature of pore-scale fluid dynamics, thereby integrating the image analysis phase into the CT reconstruction.
Figure 1 illustrates how this event-based concept changes the analysis of dynamic CT scans. The novelty lies in the reconstruction of the transition map, which in this case represents the arrival time of brine displacing oil from the pores in a sandstone. The DYRECT reconstruction technique iteratively improves this transition map to produce a solution that is most consistent with the experimentally acquired projection data. This is how local events can be reconstructed individually with temporal accuracy towards the projection level instead of the typical 360° CT frame level. The technique was tested on smooth scans with high angular resolution, typical fast scanning protocols at synchrotrons and lab-CT. The technique pinpoints events with temporal precision better than a tenth of a 360° rotation. For this precision level, there was no significant dependency on flow direction compared to the CT viewing angle.
In its simplest form, the presented single-transition time model applies best to irreversible displacement dynamics with non-mixing fluids and other dynamics like emerging fractures. More advanced dynamics like dissolution fronts and intermittent flow pathways require alternative time models to capture these complex details without relying on frame-based methods and heavy post-processing. This will enhance the analysis of existing and future dynamic CT scans, to develop better models for fluid dynamics in porous media.
| References | Goethals, W., Bultreys, T., Berg, S., Boone, M. N., & Aelterman, J. (2025). DYRECT Computed Tomography: DYnamic Reconstruction of Events on a Continuous Timescale. IEEE Transactions on Computational Imaging. |
|---|---|
| Country | Belgium |
| Acceptance of the Terms & Conditions | Click here to agree |
