3D Porous media reconstruction for natural and engineered radioactive waste confinement barriers – A multiscale-multiphase approach

Lemmens Laurent


Huysmans Marijke, (KUL), Marijke.Huysmans@ees.kuleuven.be

SCK•CEN Mentor

Rogiers Bart
+32 14 33 31 23

SCK•CEN Co-mentor

Laloy Eric
+32 14 33 32 19

Expert group

Engineered and Geosystems Analysis

PhD started


Short project description

Heterogeneities at the pore scale influence flow and reactive transport processes in porous media such as those relevant in natural (clay) and engineered (cement-based materials) barriers in a radioactive waste disposal system. Flow properties of porous media are controlled by the geometry of their pore space (two-phase system, i.e. solid or pore). Reactive transport properties additionally depend on the spatial distribution of different minerals within the solid matrix (multiphase system) and their specific characteristics. Therefore, 3D reconstruction of the pore and solid structure is crucial as input for pore-scale modelling of flow and reactive transport processes and to upscaling towards effective properties allowing for multiscale modelling over a large continuum of scales (i.e. ranging from the micrometre scale to the scale of a repository).

Reconstruction of porous structures requires accurate imaging of the pore space and mineral phases at different scales. Recent advances in 2D imaging techniques such as broad-ion beam scanning electron microscopy (BIB-SEM), focused ion-beam serial cross-sectioning (FIB-SEM) in combination with 3D multiple x-ray computed tomography (CT) scans with different resolutions (from nano- to medical CT) have made this possible. The use of 2D images for computer-based 3D reconstruction of porous media together with the integration of 2D and 3D images with different scales and resolutions, are therefore the subject of intensive research, aiming at reconstructions at a scale representative to determine effective properties. A first step in an upscaling workflow is to integrate 2D (training image for small-scale heterogeneity) and 3D (training images for larger scale heterogeneity) images and is recently introduced by Claes (2015) in the µCT to medical CT range for a two-phase system (solid – pore). Gerke et al. (2015) present a workflow for merging µCT and FIB-SEM/SEM images, for a four-phase system, assuming independency of the structures at different scales. These workflows are scale-independent and can in theory address any kind of scale range. Therefore in the proposed research, the scale of information will be enhanced spatially (e.g. by going from FIB-SEM to medical CT imaging), with the inclusion of different phases in the solid matrix.

In addition, the choice of an optimal algorithm for performing porous media modelling depends strongly on the characteristics of the porous media. Although multiple point statistics (MPS) or different kinds of two-point correlation functions were used in the workflows mentioned above, other complex porous media as clay and concrete may require other techniques as solid texture synthesis approaches, plurigaussian simulations, process-based simulations, … Here also, extension to multiphase solid matrices is a huge challenge. Hence, accurate 3D multiscale and multiphase porous media reconstruction based on 2D and/or 3D data involves many scientific and computational challenges.



Claes, S. (2015). Pore classification system and upscaling strategy in travertine reservoir rocks. PhD thesis, KU Leuven.
Gerke, K., Karsanina, M., & Mallants, D. (2015). Universal Stochastic Multiscale Image Fusion: An Example Application for Shale Rock. Scientific Reports, Accepted.


To model heterogeneous porous structures in 3D at different scales, the objectives of this PhD are:

  1. To develop a 3D porous media reconstruction methodology that is able to (1) handle and integrate heterogeneities at different scales (in terms of sample size and resolution) based on 2D and 3D imaging techniques ranging from the nm-µm scale (FIB-SEM and BIB-SEM) over mm-cm scale (petrography of thin sections, µCT) to the dm-m scale (medical CT scans), and (2) handle complex multi-phase media such as, e.g., Boom Clay and different types of cement-based materials. The proposed methodology will, however, be generic and applicable to any kind of porous media system at any kind of scale.
  2. To apply and validate the reconstruction methodology for one or more of the above mentioned materials. Validations are based on statistical comparisons of generated and measured pore structure properties and on comparison of measured effective properties (permeability, diffusivity) with pore scale simulations.

To achieve this, the study is subdivided into the following work packages:

  1. Gathering 2D and 3D images of different resolutions and sample sizes, from existing studies and by performing new measurements. These images should encompass all main characteristics and facies of the studied porous media, and allow for testing different workflows and algorithms. Moreover, determining the REV of the different facies types is important in order to determine the relevant sample size for each resolution used in the study. Effective properties (diffusivity and permeability) at a representative scale will be measured using existing techniques at SCK•CEN.
  2. Reviewing and testing different reconstruction algorithms, focusing on their computational efficiency and conditioning capabilities. Amongst other methods, MPS approaches, solid texture synthesis and plurigaussian simulation algorithms will be investigated. The method(s) presenting an optimal tradeoff between efficiency and accuracy will be identified.
  3. Reviewing, testing and improving upon existing 2D to 3D simulation workflows, that either simulate directly in 3D or use sequential 2D simulations, depending on the capabilities of the algorithms selected in step 2. With respect to the existing approaches, a more data-driven workflow will be developed for more generic applicability.
  4. Reviewing, testing and improving upon existing multi-scale simulation approaches. Starting from the currently used ad hoc methods, the potential of integrating secondary data and performing block conditioning within the multi-scale problem will be studied, as these methodological developments might provide a more generic solution. Moreover, the effect of using the raw continuous (potentially multi-variate) imaging data (instead of the processed segmented images) in a multi-scale framework will be investigated as well.
  5. Finalizing the proposed general methodology for performing 3D multi-scale multi-phase porous media reconstruction, and demonstrating its efficiency and accuracy for the selected case studies (Boom Clay, and different types of concrete). Flow and transport is simulated with an in-house developed lattice Boltzmann solver to derive effective properties.

This PhD thesis introduces several innovative aspects:

  • Accurate and efficient 3D multiphase porous media reconstruction presents a huge challenge, which will be tackled by looking at algorithms used in different fields, identifying the optimal workflows, and combining them into a new accurate and efficient approach.
  • Multiscale porous media reconstruction, with conditioning between different scales, has only been explored for the two-phase case. A proper geostatistical multiscale framework that makes use of secondary data or block conditioning has not been used in this context to date, and should enable the extension to multiphase porous media.
  • Complex pore geometry and spatial distribution of different minerals in natural and engineered radioactive waste confinement barriers have seldom been investigated in the framework of porous media reconstruction. Together with the state-of-the-art imaging techniques, the first two innovative aspects will enable us to address such complex media as well.