Abstract
The pore scale morphology of a porous medium determines the fluid transport
through that medium. It can be described by parameters, such as porosity, pore size distribution
and surface area, which can be computationally determined from synchrotron
based X-ray tomography (SXCT) data. The uncertainties in the parameters extracted
from a tomogram can increase when it is compromised with artifacts, is segmented improperly,
or if the volume of the sample or the measurement resolution is low. Similarly,
if 3D data are used for pore scale simulations, such as NMR relaxation or lattice Boltzmann
methods, the results can become unreliable if the structural disparity between a
reconstructed tomogram and a real sample is large. In this thesis, I attempted to decrease
this disparity, especially because the samples were natural porous media, such
as Bryozoan limestone and North Sea Basin chalk, where errors in a small scale study
can have profound implications when the results are applied to larger scale systems.
The sources of the disparity between a physical sample and tomography data already
begin with imaging. As most imaging systems, tomograms are prone to various
artifacts. A reconstructed tomogram often contains ring artifacts, which have no physical
significance but that introduce false pores or solids when the image is segmented. I
developed an algorithm that successfully suppresses these artifacts and can be operated
automatically over large SXCT datasets. Another hindrance for an SXCT based quantitative
study at nanometer length scales is that the imaging field of view (FOV) is limited
to a small volume of the sample. It was found that a larger volume with a lower image
resolution is more representative of a high porosity bryozoan limestone, whereas for
medium porosity chalk, the extracted properties show a higher dependence on image
resolution than sample volume. This indicates that high resolution tomography is not
always needed for some rock types. A straightforward application of 3D tomograms
for a pore scale NMR simulation gave a false estimation of available surfaces for the
suspended particles, that resulted in erroneous interfacial interactions. To solve this
problem, I introduced a local surface area correction scheme, where an effective marching
cubes surface is provided to a particle as soon as it comes in contact with a solid
interface. The local surface area modulated the annihilation probability individually
for millions of particles digitally suspended in a computational model of a rock sample.
Such an adaptive correction narrowed the gap between numerically and analytically obtained
NMR responses. This confirms the need for conditioning the data for a reliable
physical characterization.
through that medium. It can be described by parameters, such as porosity, pore size distribution
and surface area, which can be computationally determined from synchrotron
based X-ray tomography (SXCT) data. The uncertainties in the parameters extracted
from a tomogram can increase when it is compromised with artifacts, is segmented improperly,
or if the volume of the sample or the measurement resolution is low. Similarly,
if 3D data are used for pore scale simulations, such as NMR relaxation or lattice Boltzmann
methods, the results can become unreliable if the structural disparity between a
reconstructed tomogram and a real sample is large. In this thesis, I attempted to decrease
this disparity, especially because the samples were natural porous media, such
as Bryozoan limestone and North Sea Basin chalk, where errors in a small scale study
can have profound implications when the results are applied to larger scale systems.
The sources of the disparity between a physical sample and tomography data already
begin with imaging. As most imaging systems, tomograms are prone to various
artifacts. A reconstructed tomogram often contains ring artifacts, which have no physical
significance but that introduce false pores or solids when the image is segmented. I
developed an algorithm that successfully suppresses these artifacts and can be operated
automatically over large SXCT datasets. Another hindrance for an SXCT based quantitative
study at nanometer length scales is that the imaging field of view (FOV) is limited
to a small volume of the sample. It was found that a larger volume with a lower image
resolution is more representative of a high porosity bryozoan limestone, whereas for
medium porosity chalk, the extracted properties show a higher dependence on image
resolution than sample volume. This indicates that high resolution tomography is not
always needed for some rock types. A straightforward application of 3D tomograms
for a pore scale NMR simulation gave a false estimation of available surfaces for the
suspended particles, that resulted in erroneous interfacial interactions. To solve this
problem, I introduced a local surface area correction scheme, where an effective marching
cubes surface is provided to a particle as soon as it comes in contact with a solid
interface. The local surface area modulated the annihilation probability individually
for millions of particles digitally suspended in a computational model of a rock sample.
Such an adaptive correction narrowed the gap between numerically and analytically obtained
NMR responses. This confirms the need for conditioning the data for a reliable
physical characterization.
| Original language | English |
|---|
| Publisher | Department of Chemistry, Faculty of Science, University of Copenhagen |
|---|---|
| Publication status | Published - 2016 |