 |
| Titel |
Effective structural descriptors for natural and engineered radioactive waste confinement barriers |
| VerfasserIn |
Laurent Lemmens, Bart Rogiers, Mieke De Craen, Eric Laloy, Diederik Jacques, Marijke Huysmans, Rudy Swennen, Janos L. Urai, Guillaume Desbois |
| Konferenz |
EGU General Assembly 2017
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 19 (2017) |
| Datensatznummer |
250149716
|
| Publikation (Nr.) |
 EGU/EGU2017-14094.pdf |
|
|
|
|
|
| Zusammenfassung |
| The microstructure of a radioactive waste confinement barrier strongly influences its flow and
transport properties. Numerical flow and transport simulations for these porous media at the
pore scale therefore require input data that describe the microstructure as accurately as
possible. To date, no imaging method can resolve all heterogeneities within important
radioactive waste confinement barrier materials as hardened cement paste and natural
clays at the micro scale (nm-cm). Therefore, it is necessary to merge information
from different 2D and 3D imaging methods using porous media reconstruction
techniques.
To qualitatively compare the results of different reconstruction techniques, visual
inspection might suffice. To quantitatively compare training-image based algorithms, Tan et
al. (2014) proposed an algorithm using an analysis of distance. However, the ranking of the
algorithm depends on the choice of the structural descriptor, in their case multiple-point or
cluster-based histograms.
We present here preliminary work in which we will review different structural descriptors
and test their effectiveness, for capturing the main structural characteristics of radioactive
waste confinement barrier materials, to determine the descriptors to use in the analysis of
distance. The investigated descriptors are particle size distributions, surface area distributions,
two point probability functions, multiple point histograms, linear functions and two point
cluster functions. The descriptor testing consists of stochastically generating realizations
from a reference image using the simulated annealing optimization procedure introduced by
Karsanina et al. (2015). This procedure basically minimizes the differences between
pre-specified descriptor values associated with the training image and the image being
produced. The most efficient descriptor set can therefore be identified by comparing the
image generation quality among the tested descriptor combinations. The assessment of the
quality of the simulations will be made by combining all considered descriptors. Once the set
of the most efficient descriptors is determined, they can be used in the analysis of
distance, to rank different reconstruction algorithms in a more objective way in future
work.
Karsanina MV, Gerke KM, Skvortsova EB, Mallants D (2015) Universal Spatial
Correlation Functions for Describing and Reconstructing Soil Microstructure. PLoS ONE
10(5): e0126515. doi:10.1371/journal.pone.0126515
Tan, Xiaojin, Pejman Tahmasebi, and Jef Caers. "Comparing training-image based
algorithms using an analysis of distance." Mathematical Geosciences 46.2 (2014): 149-169. |
|
|
|
|
|
|