Commonly Used Porous Building Materials: Geomorphic Pore Structure and Fluid Transport

Article (PDF Available)inJournal of Materials in Civil Engineering 25(12):1803-1812 · December 2013with 107 Reads
DOI: 10.1061/(ASCE)MT.1943-5533.0000706
Cite this publication
Abstract
Knowledge of microscopic geomorphic structures is critical to understanding transport processes in porous building materials. X-ray scans were obtained of a variety of commonly used porous building materials to both qualitatively and quantitatively evaluate their pore structures. The specimens included natural materials (two sandstones and a limestone) and engineered materials (three types of concretes and a brick). Scanned images were processed to reconstruct the geomorphic structures of these materials. Random walk analyses were performed on the reconstructed pore structures to estimate macroscopic transport properties (including tortuosity, specific surface, and permeability). The effective porosity and permeability of these materials were also experimentally determined and compared to computed values. Calibration of the threshold pixel value used in the postprocessing of X-ray images against measured effective porosity appears to be a more appropriate method of selecting this value than the typical approach, which employs selection based solely on observed histograms. The resulting permeabilities computed by using a calibrated threshold pixel value compare better with the measured permeabilities. This study also demonstrates that the relatively homogeneous and heterogeneous pore structures associated with the natural and engineered building materials under investigation can be captured by X-ray tomography. (C) 2013 American Society of Civil Engineers.
Commonly Used Porous Building Materials: Geomorphic
Pore Structure and Fluid Transport
Liang Bo Hu, A.M.ASCE1; Cabot Savidge2; Donna M. Rizzo3; Nancy Hayden, M.ASCE4;
James W. Hagadorn5; and Mandar Dewoolkar, M.ASCE6
Abstract: Knowledge of microscopic geomorphic structures is critical to understanding transport processes in porous building materials.
X-ray scans were obtained of a variety of commonly used porous building materials to both qualitatively and quantitatively evaluate their pore
structures. The specimens included natural materials (two sandstones and a limestone) and engineered materials (three types of concretes and
a brick). Scanned images were processed to reconstruct the geomorphic structures of these materials. Random walk analyses were performed
on the reconstructed pore structures to estimate macroscopic transport properties (including tortuosity, specific surface, and permeability).
The effective porosity and permeability of these materials were also experimentally determined and compared to computed values.
Calibration of the threshold pixel value used in the postprocessing of X-ray images against measured effective porosity appears to be a
more appropriate method of selecting this value than the typical approach, which employs selection based solely on observed histograms.
The resulting permeabilities computed by using a calibrated threshold pixel value compare better with the measured permeabilities.
This study also demonstrates that the relatively homogeneous and heterogeneous pore structures associated with the natural and engineered
building materials under investigation can be captured by X-ray tomography. DOI: 10.1061/(ASCE)MT.1943-5533.0000706.© 2013
American Society of Civil Engineers.
CE Database subject headings: Porous media; Construction materials; Permeability.
Author keywords: Porous building materials; X-ray; Computed tomography; Transport; Permeability.
Introduction
The characterization of geomorphic pore structures (e.g., pore size,
shape, and connectivity) of natural and engineered building mate-
rials is important for studies of the transport, fate, and remediation
of chemical and biological contaminants. It is also of great practical
importance to understand the relationship between the geomorphic
pore structures and the material properties relevant to fluid transport
within them. Although transport in porous media has been inten-
sively investigated in the context of multiphase flow theory, many
models are still empirical and involve parameters that are difficult
to determine (Richards 1931;Brooks and Corey 1964). Each flow
is driven by a pressure gradient; saturation dependent capillary
pressure functions and relative permeability functions are needed
to solve the system equations. Such models are often further
reduced to a nonlinear diffusion equation for the process of unsatu-
rated flow of the wetting fluid in rigid porous building materials
(Hall and Hoff 2002;Lockington and Parlange 2003;El Abd et al.
2005). These models usually involve nonlinear diffusivities that de-
pend on saturation. Despite their merits and limitations, all of these
approaches require the quantitative characterization of pore struc-
tures to understand and model the transport process in porous me-
dia. Although traditional macroscopic laboratory experiments can
provide average estimates for properties such as porosity, many
pore structure details cannot be easily obtained with these methods.
In addition, the heterogeneity of pore structures plays a prominent
role in the transport process, and therefore, quantifying this prop-
erty field is an emerging need.
Microfocus X-ray computed tomography has widely been used
in a variety of scientific research and engineering studies to obtain
three-dimensional (3D) images of opaque materials (White et al.
2006). This technique allows the reconstruction of the porous
material microstructure to characterize and quantify the geomor-
phic structure by image analysis of a series of X-ray scans. In multi-
phase transport processes, it has emerged as an effective tool for
the direct detection of phase evolution (Cnudde and Jacobs 2004;
Roels and Carmeliet 2006) and for entities that are difficult to
evaluate, such as interfacial surface area and common phase curves
(Willson et al. 2004;Zhao and Ioannidis 2007). Nonetheless, ef-
fective estimation methods for macroscopic transport properties,
including permeability and diffusivity, are still of great interest
in many fields such as geophysics, geology, and civil and environ-
mental engineering. Many simulation methods have focused on
the determination of the diffusion coefficient and permeability.
Coker et al. (1996) extracted several different correlation func-
tions to statistically characterize the pore space morphology and
relevant pore space length and time scales from the X-ray image
1Assistant Professor, Dept. of Civil Engineering, Univ. of Toledo,
2801 West Bancroft St., Toledo, OH 43606 (corresponding author). E-mail:
Liangbo.Hu@utoledo.edu
2Engineer, New England Research, Inc., 331 Olcott Dr., Suite L1, White
River Junction, VT 05001. E-mail: csavidge@ner.com
3Professor, Univ. of Vermont, 33 Colchester Ave., Burlington,
VT 05405. E-mail: Donna.Rizzo@uvm.edu
4Professor, Univ. of Vermont, 33 Colchester Ave., Burlington,
VT 05405. E-mail: Nancy.Hayden@uvm.edu
5Tim and Kathryn Ryan Curator of Geology, Denver Museum of
Nature and Science, 2001 Colorado Blvd., Denver, CO 80205. E-mail:
Whitey.Hagadorn@dmns.org
6Associate Professor, Univ. of Vermont, 33 Colchester Ave., Burlington,
VT 05405. E-mail: Mandar.Dewoolkar@uvm.edu
Note. This manuscript was submitted on December 30, 2011; approved
on October 2, 2012; published online on October 4, 2012. Discussion per-
iod open until May 1, 2014; separate discussions must be submitted for
individual papers. This paper is part of the Journal of Materials in Civil
Engineering, Vol. 25, No. 12, December 1, 2013. © ASCE, ISSN 0899-
1561/2013/12-1803-1812/$25.00.
JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / DECEMBER 2013 / 1803
J. Mater. Civ. Eng. 2013.25:1803-1812.
Downloaded from ascelibrary.org by Colorado University At Boulder on 03/23/15. Copyright ASCE. For personal use only; all rights reserved.
of a Fontainebleau sandstone sample. Similarly, in the Lattice
Boltzmann method (Maier et al. 1998;OConnor and Fredrich
1999;Fredrich et al. 2006), the discrete Boltzmann equation is
solved to simulate the flow of a Newtonian fluid and the permeabil-
ity is obtained through the correlation between the pressure field
and the flow velocity field (White et al. 2006). Manwart et al.
(2002) and Piller et al. (2009) used a conventional computational
fluid dynamics approach to obtain the permeability tensor for a
laminar flow through porous media. A possible alternative method
proposed by Nakashima and Watanabe (2002), which is relatively
easy to implement computationally, uses a random walk algorithm
to estimate relevant transport properties. This technique can be used
to simulate the Brownian motion of a molecule as it travels in a
liquid or a gas, and thus, allows for the extraction of the diffusive
properties of the particle in the medium.
One important aspect, when employing the preceding X-ray im-
age analysis, is its applicability in commonly used porous building
materials. In combination with several of the computed tomogra-
phy (CT) scanning technical factors, the complex and possibly
heterogeneous material structure presents a challenge for estimat-
ing macroscopic transport properties with acceptable levels of ac-
curacy. This paper quantifies the geomorphic pore structures of
various natural and engineered porous building materials by
using X-ray CT. Similar methods have been applied only to arti-
ficial porous systems [e.g., an assembly of beads in the study by
Nakashima and Watanabe (2002)] and certain cementitious materi-
als [e.g., cement paste in the study by Promentilla et al. (2009)]. In
this work, a variety of natural and engineered building materials
including sandstones, limestones, bricks, and concretes were inves-
tigated and their transport properties were estimated. X-ray CT was
employed to reconstruct the geomorphic pore structure of the ma-
terials. The reconstructed geomorphic pore structure was sub-
sequently used to estimate the transport properties. This was
achieved by using a random walk-based model. The estimated per-
meabilities were compared with experimental measurements.
Whereas the conventional segmentation processes typically involve
an empirical determination of the threshold pixel values based
solely on observed histograms, an alternative method based on cal-
ibration against experimentally measured effective porosity was
also used in the present work. One of the principal goals of this
study was to investigate the effects resulting from this calibration
on the estimation of other transport properties.
Methods
Materials and Experimental Setup
A variety of natural and engineered building materials were
investigated, including three types of concretes, two types of sand-
stone, one limestone, and one brick. The concrete included:
(1) 3,000 psi concrete, produced according to a standard mixing
procedure (Derucher et al. 1994) to have a compressive strength
of 3,000 psi; (2) 5,000 psi concrete, produced to have a com-
pressive strength of 5,000 psi; and (3) D04 concrete, supplied by
Edgewood Chemical Biological Center (ECBC), Maryland. The
stone building materials included a sample of arenitic sandstone,
hereafter termed Ohio sandstone,a sample of an arkosic sand-
stone, hereafter termed Arkose sandstone,and a sample of a pack-
stone, hereafter termed Indiana limestone.A generic type of solid
brick with sand to granule sized stone aggregate was also obtained
from a hardware store for this experiment. Cylindrical core spec-
imens, approximately 12 mm in diameter, were extracted for X-ray
CT scanning. As discussed in the subsequent sections, additional
smaller and larger Ohio sandstone specimens were used to examine
sample size effects. For CT scanning, specimens were cored from
the larger specimens used for laboratory macroscopic permeability
testing. Because the properties associated with the surfaces of con-
crete and brick could differ substantially from the properties of their
interior, the same interior surface samples were used for both CT
scans and macroscopic analyses.
All X-ray scanning was performed with a microfocus X-ray CT
scanner (Skyscan 1172). Specimens were scanned at 74 keV
(133 μA), with approximately 500600 equally spaced images col-
lected across 180° of rotation. Three frames were averaged for each
image to improve the signal-to-noise ratio; ring artifacts were cor-
rected by using a random movement algorithm; X-ray attenuation
slices were generated by using a modified filtered backprojection of
shadow images. Resulting stacks of cross-sectional images had
voxel size dimensions ranging from 1.5 to 7 μm, depending on
the size of the scanned specimens.
Sample X-Ray Images and Image Processing
A typical original cross-sectional X-ray CT image is a grayscale
image with pixel values varying from 0 (black) to 255 (white).
Shades of gray represent the range of X-ray attenuation through
the sample, with black pixels representing the pore space (air)
and the white or gray pixels representing the solid phase. A seg-
mentation process is needed to convert the original grayscale image
to black and white (B&W) for further analyses. The most common
procedure involves the selection of a threshold pixel value. All pix-
els with values less than the threshold are converted to 0(black),
whereas values above the threshold are converted to 1(white),
resulting in a B&W image. In practice, the colors are often reversed
for a more conventional visualization by using black for solids and
white for pores.
Selection of the threshold pixel value is a critical factor. It is
often selected by using a histogram of pixel values in which
two peaks can be identified and distinguished from one another.
In practice, most geologic and engineered building materials are
comprised of materials whose X-ray contrast are similar and/or
overlap one another. In an ideal data set, the minimum point (val-
ley) between the peaks is selected as the threshold (Lu et al. 2006).
Alternatively, Nakashima and Watanabe (2002) used the average of
these two peak values for the threshold in CT scans of an assembly
of beads. In these applications, a clear distinction between glass
beads and pores was readily visible prior to any processing; and
the two histogram peaks were well separated. However, in scanned
images of real building materials, especially cementitious materi-
als, it is often difficult to distinguish the histogram peak values rep-
resenting the solids and pores, and the separation between the peaks
can be very small. As a result, the processing can be quite sensitive
to the selection of the threshold. The characteristics of an image are
dictated by the physical nature of the porous material, although the
processing procedure, quality and limitations of the experimental
device, and setup are all potentially significant.
The comparison of pore structure between sandstone and con-
crete is a good example of this feature. A typical cross-sectional
image of an Ohio sandstone specimen is shown in Fig. 1(a) after
the contrast is enhanced for a better illustration. The corresponding
cropped B&W image [Fig. 1(b)] has colors reversed for more con-
ventional visualization (white pixels representing the pore space).
The boundaries of the sandstone grains are distinguishable from the
pores. Fig. 2(a) shows an enhanced X-ray portion of a 5,000 psi
concrete specimen. It clearly reveals the coarse aggregates (lightest
shade of gray) and small pores (darkest gray) between these aggre-
gates. Most of the space between the coarse aggregates may be
1804 / JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / DECEMBER 2013
J. Mater. Civ. Eng. 2013.25:1803-1812.
Downloaded from ascelibrary.org by Colorado University At Boulder on 03/23/15. Copyright ASCE. For personal use only; all rights reserved.
finer sand to clay sized aggregates or cement. As a result, its final
B&W image critically depends on the selected threshold. In this
case, the threshold determines the fraction of the pore space among
the fine aggregates. The B&W image of Fig. 2(b) has a high
selected pixel threshold (25), and therefore a larger porosity.
X-ray images of the rest of the investigated materials are provided
in Fig. 3.
Quantification of porosity in this case may be obtained directly
from the pixel histogram. Fig. 4shows the histogram of the original
X-ray image (Fig. 2) of the 5,000 psi concrete in which most of the
pixels have values less than 100. A closer look at the histogram
distribution reveals two peaks. The first peak, pixel value 0, rep-
resents the pores. The second peak, at approximately 30, has the
largest population. Consequently, the selection of a threshold value
between 0 and 30 would delineate the pore structure and the poros-
ity. The porosity may be plotted as a function of the selected thresh-
old value (Fig. 5), resulting in the cumulative distribution function
of the histogram. Changing the threshold from 15 (the midpoint
between 0 and 30) to 20 would result in a difference of almost
0.10 in terms of absolute value of porosity.
Because of this potential discrepancy, it is often recommended
to experimentally measure the porosity and adjust the contrast
threshold until the porosity (estimated by using the image alone)
reasonably matches the measured porosity (Promentilla et al.
2009). This is the approach adopted in this study. The measured
porosity is the effective material porosity when the connectivity
of the pores has been considered. Therefore, the calculated porosity
must be based on those connected pores only, which must be evalu-
ated through 3D reconstruction rather than from a two-dimensional
(2D) slice. Computationally, this is achieved by using MATLAB
(V7.7) functions to identify those connected pores (numerically
represented as ones) in a 3D matrix. In this study, both criteria were
used for the segmentation process, including: (1) effective porosity
calibration-based thresholding, and (2) histogram-based threshold-
ing. In the subsequent section, the transport properties were simu-
lated by using the distinctive pore structures resulting from the
different thresholding methods. Comparison between these estima-
tions allows the effectiveness of each method to be assessed.
The 3D material pore structure can subsequently be recon-
structed with volume rendering and is useful for simulating the
material properties of these commonly used building materials.
The memory required to process the 3D matrix is significantly
greater than the 2D matrix. As a consequence, the size of the re-
constructed 3D volume on which random walk simulations are per-
formed is usually much smaller than the original scanned image. In
the present study, a stack of 250 scans are used to generate a
2503-voxels cube with a size of ð900 μmÞ3, or a little less than
1mm3. The selected size of the matrix is close to the upper limit
afforded with a conventional PC with 3 GB of RAM. All of the
general estimation methods discussed in the following sections
are readily applicable to 2D images, which often provide better
demonstration examples. All computation was implemented by
using MATLAB V7.7.
Determination of Porosity and Specific Surface Area
The Ohio sandstone was rather homogeneous even at a relatively
small scale; in contrast, the presence and distribution of large
Fig. 1. (a) Enhanced sample X-ray image of an Ohio sandstone specimen (5×5mm in diameter), in which a cropped area is outlined in the square
(contrast enhanced); (b) final B&W image after image processing
Fig. 2. (a) Enhanced sample X-ray image of a portion of a 5,000 psi concrete specimen; (b) final B&W image after image processing
JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / DECEMBER 2013 / 1805
J. Mater. Civ. Eng. 2013.25:1803-1812.
Downloaded from ascelibrary.org by Colorado University At Boulder on 03/23/15. Copyright ASCE. For personal use only; all rights reserved.
aggregates rendered 5,000 psi concrete inhomogeneous, at least at a
macroscopic specimen scale. Quantitatively, two properties that
can be directly computed from the X-ray images, porosity and
specific surface area, may be used to examine the heterogeneity
of the studied materials.
Porosity may be computed by counting the number of white
pixels (ones) in the matrix representing the B&W 2D image, or
in the 3D array representing a volume. The specific surface area
discussed in this work is defined as the surface area of the solid
pore interface per unit total (bulk) volume of porous material,
ðS=VÞt. This quantity is inversely related to the ratio of the surface
area of solid pore interface to the volume of pores, termed surface to
volume of pores, ðS=VÞp, by porosity ϕ,ðS=VÞp¼1=ϕðS=VÞt.
The surface to volume of pores may be considered characteristic
of the pore size; the specific surface area considered here solely
describes the amount of solid pore interface area in a given volume
of porous material.
The solid pore interfacial surface area can be directly obtained
by identifying solid pore interfaces in a 3D array and calculating the
surface area represented. In 2D images, the specific surface area is
reduced to the ratio of the perimeter of the solid pore interface to the
total area.
The heterogeneous nature of the porous materials studied in
this work was evaluated by using the distribution of porosity
Fig. 3. Portion of an enhanced sample X-ray image (left) and its cor-
responding B&W image after image processing (right): (a) 3,000 psi
concrete; (b) D04 concrete; (c) Arkose sandstone; (d) Indiana lime-
stone; (e) brick
0 20 40 60 80 100
0
2
4
6
8
10
12 x 104
Pixel value
Number of pixels
Fig. 4. Histogram of an original scan of 5,000 psi concrete specimen
[Fig. 3(a)]
0 20 40 60 80 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Threshold
Porosity
Fig. 5. Total porosity in relation to the selected threshold, i.e., the cu-
mulative function of the histogram in Fig. 4
1806 / JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / DECEMBER 2013
J. Mater. Civ. Eng. 2013.25:1803-1812.
Downloaded from ascelibrary.org by Colorado University At Boulder on 03/23/15. Copyright ASCE. For personal use only; all rights reserved.
and specific surface area. Estimates were obtained based on the
histogram-based thresholding. The distribution of porosity and
specific surface area in Ohio sandstone and 5,000 psi concrete
along the vertical (z) direction of a cylindrical specimen are plotted
in Figs. 6and 7, respectively. At each location along z, 64 estima-
tions are made in a volume of 2,000 ×2,000 ×250 pixels, which is
divided into 8×8cubes [each cube of 2503pixels has a volume of
ð900 μmÞ3]. A 2503-pixel array (cube) was chosen because it is
nearly the maximum 3D array size that the computer used in this
research can handle, while simultaneously being efficient at run-
ning a walker (particle) to produce meaningful random walk sim-
ulations. The distributions of porosity and specific surface are more
continuous and vary significantly less for Ohio sandstone than
5,000 psi concrete. The average porosity was calculated to be
0.22 with an SD of 0.03 for Ohio sandstone. The average porosity
and SD for 5,000 psi concrete were 0.17 and 0.08, respectively.
Ohio sandstone also has an average specific surface area of 8.31 ×
104m1with an SD of 1.13 ×104m1, whereas 5,000 psi con-
crete has an average specific surface area of 7.96 ×104m1with
an SD of 2.88 ×104m1. The variation along the cylindrical core
of the concrete specimen clearly manifests the heterogeneous
nature of concrete.
Estimation of Tortuosity and Permeability
There are several ways to estimate other transport properties such
as permeability by using the reconstructed porous structure from
the X-ray CT images. Here, the random walk algorithm discussed
by Nakashima and Watanabe (2002) was adopted because of its
simplicity and relative computational efficiency. The relationship
of random walk simulation to the stochastic Wiener process, a pro-
cess similar to Brownian motion, has been established in probabil-
ity and statistics theory (Feller 1968). A detailed discussion and
review of its correlation to the diffusion coefficient can be found
in the studies by Nakashima and Watanabe (2002) and Anta et al.
(2008). In the following, a B&W image from one of the CT scans
(Fig. 1) is used to illustrate the essential components of this ap-
proach. A particle (or walker) is initially placed at random inside
a pore (Fig. 8). It subsequently migrates randomly along discrete
pixels over time, which is denoted using a dimensionless integer
time, τ. Usually, a large number of these particles must be used
and the average of the square distances of all of these random par-
ticles are computed as mean-square displacement < rðτÞ2>,asa
function of τ. A general 3D formulation may be expressed by the
following equation:
<rðτÞ2>¼1
nX
n
i¼1
riðτÞ2
¼1
nX
n
i¼1xiðτÞxð0Þ2þ½yiðτÞyð0Þ2
þ½ziðτÞzð0Þ21Þ
where xi,yi, and zi=coordinates of the ith particle; and n=number
of particles. In 2D cases, only xiand yineed to be considered.
If this random walk is completely unrestricted, i.e., in a void
without any solid particles present, < rðτÞ2>is proportional to
τ(solid line in Fig. 9), and the proportionality constant will reflect
the diffusion coefficient of the particle in the free space without
solids (water diffusivity in bulk water). However, because real
porous media contain solids (black pixels in Fig. 8), < rðτÞ2>
is reduced because the solids act as obstacles. The change in the
function < rðτÞ2>provides a measure for the diffusivity in real
porous materials; thus, tortuosity may be obtained by comparing
the gradient to that of a free random walk < rðτÞ2>free . According
to Nakashima and Kamiya (2007), the geometrical tortuosity may
00.1 0.2 0.3 0.4 0.5 0.6
0
5
10
15
20
Porosity
Z (mm)
5,000 psi concrete
Ohio sandstone
Fig. 6. Porosity profiles for Ohio sandstone and 5,000 psi concrete
Fig. 7. Specific surface area profiles for Ohio sandstone and 5,000 psi
concrete
Fig. 8. Sample 2D random walk trajectory in the pore space of Ohio
sandstone (the solid circle indicates the initial location of the particle)
JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / DECEMBER 2013 / 1807
J. Mater. Civ. Eng. 2013.25:1803-1812.
Downloaded from ascelibrary.org by Colorado University At Boulder on 03/23/15. Copyright ASCE. For personal use only; all rights reserved.
be represented as the ratio between these gradients, D0=D, and a
useful expression can be obtained:
D
D0¼
drðτÞ2
dτ
a2¼1
4a
9ffiffiffiffiffi
6π
pS
Vpffiffi
τ
pþc2τð2Þ
where c2=data fitting constant; a=lattice constant of the simple
lattice, i.e., the dimension of a CT pixel. This equation can be used
to estimate ðS=VÞpvia data fitting. Furthermore, permeability can
also be estimated based on the Kozeny-Carman equation:
k¼αϕ
ðD0
DÞðS
VÞ2
p¼αϕ3
ðD0
DÞðS
VÞ2
tð3Þ
where α=correction factor and is often assumed to be 1 when
details of the pore geometry are unknown. It was found that
ðS=VÞpis readily available from the images; therefore, the use
of Eq. (2) to obtain the estimates of ðS=VÞpis not an absolute ne-
cessity, especially when the curve fitting does not yield a satisfac-
tory match for the proposed form of the function. The porosity in
the preceding equation is typically the total porosity. In this study,
this estimation is also evaluated by replacing the total porosity with
the effective porosity in the subsequent sections.
In random walk simulation, the distance a particle travels is con-
fined to the area (for 2D) or the space (for 3D) of the pores. For
example, for a 100 ×100 matrix representing a porous medium
in Eq. (1), the maximum possible square displacement for each
walker is the longest distance, 100 ffiffi
2
p. Consequently, the mean
square displacement for all of the walkers is bounded, that is,
<rðτÞ2>will approach an upper limit over time. Even for
<rðτÞ2>free, obtained from simulation in a free space, it can con-
tinue to grow proportionally to τonly if the random walk is allowed
in an infinite free space. This implies that in computing <rðτÞ2>free,
it is more appropriate to use the free space of the same size instead
of an infinite free space.
In addition to the porosity, specific surface area, tortuosity, and
permeability, the connected pores may be identified in any given
2D or 3D image. For example, MATLAB permits identification
of the pixel connectivity, defined according to whether faces (6-
connectivity), edges (18-connectivity), or corners (28-connectivity)
touch those of adjacent pixels. This capability allows one to obtain
the continuous pore space in a 3D cube, whereas the rest of the pore
space may be considered isolated. The volume fraction of the con-
nected pores is referred to as the effective porosity (Promentilla
et al. 2009), which is more similar to that measured in the labora-
tory than the total porosity.
Results and Discussion
The aforementioned 3D transport properties for each material were
estimated by using the reconstructed porous structure based on two
different thresholding criteria, discussed previously. In the simula-
tions based on the calibration-based thresholding criterion, the
threshold was adjusted iteratively until the effective porosity
matched that measured in the experiments. Subsequently, in the
image analysis and random walk simulations, all estimates were
based on a volume of 2,000 ×2,000 ×250 voxels divided into
8×8cubes, each having 2503voxels (0.93 mm3). The transport
properties are estimated for each material and summarized in
Table 1. The measured permeability is obtained in the laboratory
by using ASTM standard D5084-00 (2002). These tests were per-
formed on specimens of 50 mm in diameter and 50 mm in height.
The effective porosity was estimated for the same specimen after its
permeability was determined. Each specimen was initially oven
dried and weighed. The permeability test was conducted as a con-
tinuous water flow was introduced to the specimen for permeability
measurement. Subsequently, the specimen was weighed and the
weight difference was converted to the volume of water to deter-
mine the volume of pore space occupied by water. This porosity
was considered to be the effective porosity because it is reasonable
to presume that the water only enters the connected pores. For each
material, five specimens were tested for both permeability and
effective porosity measurements, and the average results are re-
ported here. Simulations were performed in each cube and resulted
in 64 estimates that were averaged and reported in Table 1.
In addition to the experimentally measured effective porosity
and permeability, Table 1includes the estimated total porosity,
0 200 400 600 800 1000
0
100
200
300
400
500
600
700
800
900
1000
τ
<r2>
Unrestricted diffusion (free space)
Restricted diffusion (CT scan)
Fig. 9. Mean-square displacement of the random walk for a total of
2,000 particles, as shown in Fig. 8
Table 1. Estimates of Transport Properties with Experimental Effective Porosity-Based Thresholding
Material
Computed
porosity
(%)
Computed
effective
porosity
(%)
Measured
effective
porosity
(%)
Computed
tortuosity
Computed
specific
surface area
(m1)
Computed
permeability
(ρ-based)
(m2)
Computed
permeability
(ρe-based)
(m2)
Measured
permeability
(m2)
Ohio sandstone 15.0 11.1 11.5 6.5 6.3×1041.3×1013 0.8×1013 1.4×1013
Arkose sandstone 13.3 5.8 7.5 9.1 6.5×1047.2×1014 1.9×1014 2.1×1014
Indiana limestone 15.0 7.4 7.5 8.9 7.5×1048.6×1014 2.6×1014 6.4×1014
Brick 33.6 30.0 31.1 3.1 7.4×1041.3×1011 1.1×1011 9.2×1013
D04 concrete 17.4 11.1 11.0 7.7 8.2×1041.4×1013 6.0×1014 1.9×1014
3,000 psi concrete 17.1 11.5 13.2 10.0 8.4×1041.1×1013 6.6×1014 4.5×1014
5,000 psi concrete 20.3 15.9 17.3 11.6 9.5×1042.0×1013 1.6×1013 4.2×1014
1808 / JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / DECEMBER 2013
J. Mater. Civ. Eng. 2013.25:1803-1812.
Downloaded from ascelibrary.org by Colorado University At Boulder on 03/23/15. Copyright ASCE. For personal use only; all rights reserved.
effective porosity, tortuosity, specific surface area, and permeabil-
ity. In essence, the effective porosity was calibrated to the measured
porosity by adjusting the threshold pixel value; however, the esti-
mated effective porosity did not exactly match the measured be-
cause of the discrete nature of the pixel values. The seventh
column shows the permeability estimates via Eq. (3) by using
the total porosity (ρ). The differences between the estimated and
measured permeability were generally within one order of magni-
tude for most of the materials, with the exception of brick. This
level of accuracy may be regarded as satisfactory, considering
the complex geomorphic structures of real building materials,
which usually exhibit a wide range for permeability (Savidge
2010). Permeability estimates for the natural and homogeneous ma-
terials, sandstones and limestone, were particularly close to the
measured values. Among concretes, the differences for 5,000 psi
and D04 concretes were greater than the variations for 3,000 psi
concrete.
The eighth column in Table 1shows the permeability estimates
via Eq. (3) by using the effective porosity (ρe). The difference was
not significant, but for cementitious materials, it can be considered
a minor improvement. Estimates for brick are also challenging as a
result of its heterogeneous nature. The brick sample possessed a
large porosity and the simulated permeability was considerably
higher than the measurement. Figs. 10 and 11 show the effective
porosity field for Ohio sandstone and 3,000 psi concrete, respec-
tively, by using the 64 (8×8) data points and by shading such that
each cell is a bilinear interpolation of the values at its four vertices.
The 3,000 psi concrete was clearly heterogeneous, with the pres-
ence of large aggregates indicated by the near zero effective poros-
ity. Figs. 12 and 13 show the estimated permeability field for Ohio
sandstone and 3,000 psi concrete, respectively. The estimates of
permeability (in log scale) ranged between 11 and 16 orders
of magnitude (Fig. 13). In stark contrast, the homogeneous nature
of Ohio sandstone (Fig. 12) showed a very narrow range of the
estimated permeability.
1 2 3 4 5 6 7 8
1
2
3
4
5
6
7
8
X [mm]
Y [mm]
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
Fig. 10. (Color) Ohio sandstone: effective porosity field
1 2 3 4 5 6 7 8
1
2
3
4
5
6
7
8
X [mm]
Y [mm]
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
Fig. 11. (Color) 3,000 psi concrete: effective porosity field
Fig. 12. (Color) Ohio sandstone: estimated permeability field (log m2)
Fig. 13. (Color) 3,000 psi concrete: estimated permeability field
(log m2)
JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / DECEMBER 2013 / 1809
J. Mater. Civ. Eng. 2013.25:1803-1812.
Downloaded from ascelibrary.org by Colorado University At Boulder on 03/23/15. Copyright ASCE. For personal use only; all rights reserved.
The second criterion examined in this study was to select the
threshold between solid and pore based on the histogram using
the mean value of the two peaks, as discussed in the preceding sec-
tion. When the threshold was adjusted to match the porosity meas-
urement, as presented in Table 1, it was expected to yield better
estimates of the porosity than histogram-based thresholding; it is
interesting to compare the estimates of other transport properties
using these two different criteria. The subsequent simulations
for these properties followed the same procedure as already out-
lined after the porous structure was reconstructed. Results are pre-
sented in Table 2and suggest that thresholding based solely on
pixel histogram may not result in an accurate estimation for poros-
ity. Especially for the cementitious materials examined herein, ρe
was very small; as a consequence, the permeability estimated using
ρewas nearly equal to zero. However, estimates using ρwere suf-
ficiently close, partially because their counterparts based on the first
thresholding criterion (Table 1) generally overestimated the per-
meability; with smaller estimated porosity, estimation for per-
meability actually improved slightly.
Considering that the seven tested materials may be classified
into two categories: natural (sandstone and limestone) and engi-
neered (brick and concrete), a direct comparison can be made be-
tween the estimates and the experimental results using the total
porosity, ρ(Fig. 14), and the effective porosity, ρe(Fig. 15). Com-
paring the estimates in Fig. 14 with those in Fig. 15, it appears that
calibration-based thresholding is more precise than histogram-
based thresholding. In general, calibration-based thresholding
slightly overestimates the permeabilities, whereas histogram-based
thresholding underestimates the permeabilities. For natural materi-
als such as those in this study, both thresholding techniques seem
adequate if estimates are made using the total porosity. If the
effective porosity is used, calibration-based thresholding provides
better estimates for both natural and engineered materials.
Clearly, the purpose of X-ray CT scanning is to reproduce the
geomorphic structure of opaque materials as accurately as possible.
However, the quality of the original scans can be affected by
material composition and X-ray experimental techniques, in addi-
tion to the subsequent image processing; such variations may
largely dictate the accuracy of the reconstructed material structure,
which serves as a basis for further simulations, and consequently,
strongly influences the simulation results. Similarly, each simula-
tion method has different stengths and weaknesses as a function
of the validity of its assumptions and its specific computa-
tional demands; in general, these have a lesser impact than the
reconstruction of the material nature. The current numerical analy-
sis indicates the potential necessity of applying multiple segmen-
tation techniques including thresholding, region growing, classifier
method, or clustering (Pham et al. 2000) for reliably reconstructing
porous materials. Finally, the methodology used to characterize the
Table 2. Estimates of Transport Properties with Histogram-Based Thresholding
Material
Computed
porosity
(%)
Computed
effective
porosity
(%)
Measured
effective
porosity
(%)
Computed
tortuosity
Computed
specific
surface area
(m1)
Computed
permeability
(ρ-based)
(m2)
Computed
permeability
(ρe-based)
(m2)
Measured
permeability
(m2)
Ohio sandstone 15.0 11.1 11.5 6.5 6.3×1041.3×1013 1.4×1013 1.4×1013
Arkose sandstone 8.9 2.3 7.5 11.5 4.6×1043.4×1014 1.0×1015 2.1×1014
Indiana limestone 10.3 2.3 7.5 11.6 5.4×1044.2×1014 4.3×1015 6.4×1014
Brick 16.7 11.3 31.1 6.0 4.5×10414 ×1013 9.2×1013 9.2×1013
D04 concrete 8.8 0.0 11.0 15.8 4.6×1044.7×1014 01.9×1014
3,000 psi concrete 6.7 0.1 13.2 28.6 3.9×1041.5×1014 3.6×1017 4.5×1014
5,000 psi concrete 8.3 0.5 17.3 44.6 4.6×1042.7×1014 1.9×1016 4.2×1014
Fig. 14. Permeability, estimated by using ρversus measured
Fig. 15. Permeability, estimated by using ρeversus measured
1810 / JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / DECEMBER 2013
J. Mater. Civ. Eng. 2013.25:1803-1812.
Downloaded from ascelibrary.org by Colorado University At Boulder on 03/23/15. Copyright ASCE. For personal use only; all rights reserved.
materials of concern also depends on the associated scale of the
primary observations relative to the grain or aggregate size of
the studied materials, especially when localized estimates are
not adequate for field level or macro scale characterizations.
Because the scan system employed in these experiments has a
fixed field of view (4,000 ×4,000 pixels for a slice), the resolution
of the image pixels depends on the sample size of the specimen. To
explore the effects of resolution on the property estimates, three
Ohio sandstone cores with diameters of 5, 12.6, and 24 mm,
corresponding to X-ray spatial resolutions of 1.5, 3.67, and
6.27 μm, respectively, were scanned to generate identically sized
images (4,000 ×4,000 pixels). Thresholds were selected to match
the measured porosity, and subsequently, the other properties were
obtained (a summary is provided in Table 3).
The wide range of tortuosity values computed from the random
walk suggests that coarser resolution (i.e., larger mean voxel size)
tends to be inadequate when detecting possible connected pores.
This effect was also reported in the study by Promentilla et al.
(2009), in which the results are based on images (matrices) of
different sizes representing the same specimen. The results of
specific surface area did not show a monotonic correlation with
the resolution, partially because the porosity was nearly identical
and did not reflect an effect on the pore connectivity. The
permeability estimates in this case were mostly affected by the
difference in tortuosity estimates and were understandably greater
for finer resolution, smaller sized voxel dimensions.
Conclusions
Seven commonly used porous building materials were examined by
using X-ray CT. The materials included three seemingly homo-
geneous natural stones (two sandstones and a limestone) and four
heterogeneous engineered materials (three concretes and a brick).
Scanned images of each building material were processed to recon-
struct their geomorphic structures and two approaches were exam-
ined for thresholding in the analysis of these samples. Random
walk simulations were performed on the reconstructed pore struc-
tures to compute properties that are relevant to transport phenom-
ena, such as tortuosity, specific surface, and permeability. The
computed porosities and permeabilities were compared to the
measured values.
The calibration-based method for permeability analysis seemed
to provide better estimates than the histogram-based method,
especially when the effective porosity was used in Eq. (3).
Presently, the latter is more commonly used, but may not be appro-
priate for cementitious and heterogeneous building materials such
as concrete and brick. The reconstructed geomorphic structures of
these heterogeneous engineered materials varied greatly when
the thresholds for image analysis were selected based on the
histograms.
Measured and computed permeabilities were generally within
one order of magnitude of each other when the thresholding
was based on measured effective porosity. The computed perme-
abilities compared better with the measured permeabilities when
the thresholding was based on the measured effective porosity
than the histogram, particularly for the heterogeneous, artificial
materials.
Finally, the relatively homogeneous and heterogeneous pore
structures of commonly used natural and engineered building
materials can be captured by X-ray tomography.
Acknowledgments
Support for this work was provided by Defense Threat Reduction
Agency, HDTRA1- 08-C-0021. We thank Ms. Lindsay Meador for
help in the laboratory.
References
Anta, J. A., Mora-Sero, I., Dittrichc, T., and Bisquert, J. (2008). Interpre-
tation of diffusion coefficients in nanostructured materials from
random walk numerical simulation.Phys. Chem. Chem. Phys., 10(30),
44784485.
ASTM. (2002). Soil and rock; building stones.D5084-00, West
Conshohocken, PA.
Brooks, R. H., and Corey, A. T. (1964). Hydraulic properties of porous
media.Hydrology papers, Colorado State University, Fort Collins,
CO.
Cnudde, V., and Jacobs, P. J. S. (2004). Monitoring of weathering
and conservation of building materials through non-destructive
x-ray computed microtomography.Environ. Geol., 46(34),
477485.
Coker, D. A., Torquato, S., and Dunsmuir, J. H. (1996). Morphology and
physical properties of Fontainebleau sandstone via a tomographic
analysis.J. Geophys. Res., 101(B8), 1749717506.
Derucher, K., Korfiatis, G., and Ezeldin, A. (1994). Materials for civil and
highway engineers, 3rd Ed., Prentice Hall, NJ.
El Abd, A. E. G., Czachor, A., Milczarek, J. J., and Pogorzelski, J. (2005).
Neutron radiography studies of water migration in construction porous
materials.IEEE Trans. Nucl. Sci., 52(1), 299304.
Feller, W. (1968). An introduction to probability theory and its applica-
tions, Wiley, New York.
Fredrich, J. D., DiGiovanni, A. A., and Noble, D. R. (2006). Predicting
macroscopic transport properties using microscopic image data.
J. Geophys. Res., 111(B3), B03201.
Hall, C., and Hoff, W. W. (2002). Water transport in bricks, stones and
concrete, Spon Press, New York.
Lockington, D. A., and Parlange, J. Y. (2003). Anomalous water absorp-
tion in porous materials.J. Phys. D, 36(6), 760767.
Lu, S., Landis, E. N., and Keane, D. T. (2006). X-ray microtomographic
studies of pore structure and permeability in portland cement concrete.
Mater. Struct., 39(6), 611620.
Maier, R. S., Kroll, D. M., Kutsovsky, Y. E., Davis, H. T., and Bernard, R. S.
(1998). Simulation of flow through bead packs using the lattice
boltzmann method.Phys. Fluids,10(1),6074.
Manwart, C., Aaltosalmi, U., Koponen, A., Hilfer, R., and Timoten, J.
(2002). Lattice-boltzmann and finite-difference simulations for the
permeability for three-dimensional porous media.Phys. Rev. E,
66(1), 016702.1016702.11.
MATLAB, V7.7 [Computer software]. MathWorks, Inc., Natick, MA.
Nakashima, Y., and Kamiya, S. (2007). Mathematica programs for the
analysis of three-dimensional pore connectivity and anisotropic tortuos-
ity of porous rocks using x-ray computed tomography image data.
J. Nucl. Sci. Technol., 44(9), 12331247.
Nakashima, Y., and Watanabe, Y. (2002). Estimate of transport proper-
ties of porous media by microfocus x-ray computed tomography
and random walk simulation.Water Resour. Res., 38(12), 1272.
OConnor, R. M., and Fredrich, J. T. (1999). Microscale flow
modelling in geologic materials.Phys. Chem. Earth A, 24(7),
611616.
Pham, D. L., Xu, C., and Prince, J. L. (2000). Current methods in
medical image segmentation.Ann. Rev. Biomed. Eng., 2, 315337.
Table 3. Parameters Estimated for Ohio Sandstone of Different Sizes
Resolution (μm) 1.5 3.67 6.27
Sample diameter (mm) 5 12.5 24
Porosity (%) 11.5 11.2 11.8
Tortuosity 4.6 21.6 58.3
Specific surface area (m1)4.5×1042.8×1044.0×104
Permeability (m2)21 ×1014 8.1×1014 1.8×1014
JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / DECEMBER 2013 / 1811
J. Mater. Civ. Eng. 2013.25:1803-1812.
Downloaded from ascelibrary.org by Colorado University At Boulder on 03/23/15. Copyright ASCE. For personal use only; all rights reserved.
Piller, M., Schena, G., Nolich, M., Favretto, S., Radaelli, F., and Rossi, E.
(2009). Analysis of hydraulic permeability in porous pedia: From high
resolution x-ray tomography to direct numerical simulation.Transp.
Porous Media, 80(1), 5778.
Promentilla, M., Sugiyama, T., Hitomi, T., and Takeda, N. (2009). Quan-
tification of tortuosity in hardened cement pastes using synchrotron-
based x-ray computed microtomography.Cem. Concr. Res., 39(6),
548557.
Richards, L. A. (1931). Capillary conduction of liquids through porous
mediums.Phys., 1(5), 318333.
Roels, S., and Carmeliet, J. (2006). Analysis of moisture flow in porous
materials using microfocus x-ray radiography.Int. J. Heat Mass Tran.,
49(2526), 47624772.
Savidge, C. (2010). Characterization of porous building materials
for agent transport predictions using artificial neural networks.
M.S. thesis, University of Vermont, Burlington, VT.
White, J. A., Borja, R. I., and Fredrich, J. T. (2006). Calculating the
effective permeability of sandstone with multiscale lattice boltzmann/
finite element simulations.Acta Geotech., 1(4), 195209.
Willson, C. S., Stacey, R. W., Ham, K., and Thompson, K. E. (2004).
Investigating the correlation between residual nonwetting phase
liquids and pore-scale geometry and topology using synchrotron x-ray
tomography.Proc. SPIE, U. Bonse, ed., Vol. 5535, The International
Society for Optics and Photonics (SPIE), Bellingham, WA, 101111.
Zhao, W. S., and Ioannidis, M. A. (2007). Effect of napl film stability on
the dissolution of residual wetting napl in porous media: A pore-scale
modeling study.Adv. Water Resour., 30(2), 171181.
1812 / JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / DECEMBER 2013
J. Mater. Civ. Eng. 2013.25:1803-1812.
Downloaded from ascelibrary.org by Colorado University At Boulder on 03/23/15. Copyright ASCE. For personal use only; all rights reserved.
  • Article
    Full-text available
    The path integral for classical statistical dynamics is used to determine the properties of one-dimensional Darcy flow through a porous medium with a correlated stochastic permeability for several spatial correlation lengths. Pressure statistics are obtained from the numerical evaluation of the path integral by using the Markov chain Monte Carlo method. Comparisons between these pressure distributions and those calculated from the classic finite-volume method for the corresponding stochastic differential equation show excellent agreement for Dirichlet and Neumann boundary conditions. The evaluation of the variance of the pressure based on a continuum description of the medium provides an estimate of the effects of discretization. Log-normal and Gaussian fits to the pressure distributions as a function of position within the porous medium are discussed in relation to the spatial extent of the correlations of the permeability fluctuations.
  • Preprint
    Full-text available
    The path integral for classical statistical dynamics is used to determine the properties of one-dimensional Darcy flow through a porous medium with a correlated stochastic permeability for several spatial correlation lengths. Pressure statistics are obtained from the numerical evaluation of the path integral by using the Markov chain Monte Carlo method. Comparisons between these pressure distributions and those calculated from the classic finite-volume method for the corresponding stochastic differential equation show excellent agreement for Dirichlet and Neumann boundary conditions. The evaluation of the variance of the pressure based on a continuum description of the medium provides an estimate of the effects of discretization. Log-normal and Gaussian fits to the pressure distributions as a function of position within the porous medium are discussed in relation to the spatial extent of the correlations of the permeability fluctuations.
  • Article
    Tortuosity is a parameter that plays a significant role in the characterization of complex porous media systems and it has a significant impact on many engineering and environmental processes and applications. Flow in porous media, diffusion of gases in complex pore structures and membrane flux in water desalination are examples of the application of this important micro-scale parameter. In this paper, an algorithm was developed and implemented as a MATLAB code to compute tortuosity from three-dimensional images. The code reads a segmented image and finds all possible tortuous paths required to compute tortuosity. The code is user-friendly, easy to use and computationally efficient, as it requires a relatively short time to identify all possible connected paths between two boundaries of large images. The main idea of the developed algorithm is that it conducts a guided search for connected paths in the void space of the image utilizing the medial surface of the void space. Once all connected paths are identified in a specific direction, the average of all connected paths in that direction is used to compute tortuosity. Three-dimensional images of sand systems acquired using X-ray computed tomography were used to validate the algorithm. Tortuosity values were computed from three-dimensional images of nine different natural sand systems using the developed algorithm and compared with predicted values by models available in the literature. Findings indicate that the code can successfully compute tortuosity for any unconsolidated porous system irrespective of the shape (i.e., geometry) of particles.
  • Article
    An industrial lightweight concrete developed for pipe insulation is characterized using X-ray computed tomography to determine the size distribution of air voids within the paste fraction. Strategies to address practical difficulties are presented, particularly discerning air voids in paste from air voids within aggregate, and combining air void size distribution data collected at different resolutions. To address the first issue, a step-wise approach was employed to reduce the problem to a series of easier-to-solve two-phase grayscale threshold problems. The first step isolates a significant number of light weight aggregate particles on 2D slices using an active contour line approach and an effective threshold identified to discern between air void and solid voxels. The threshold is applied globally to the 3D data set, and the air void voxels removed. Next, a significant number of rectangular regions of interest from 2D slices are used to identify a second threshold between the solid portions of the lightweight aggregate and the paste. Finally, a 3D shape based criterion is used to isolate air voids in paste. Given the trade-off between resolution and sample size, it is necessary to conduct multiple scans using different resolutions and sample diameters in order to cover the full range of the air void size distribution. A power law curve is used to unite distributions collected at different resolutions.
  • Article
    Full-text available
    Three representative rock types (limestone, sandstone, and shale) and glauconite samples collected from Ewekoro Quarry, eastern Dahomey Basin in Nigeria were characterized using low field 2 MHz and 20 MHz Nuclear Magnetic Resonance (NMR) techniques. NMR T2 relaxation time decay measurement was conducted on disc samples under partial water-saturation and full water-saturation conditions using CPMG spin-echo routine. The T2 relaxation decay was converted into T2 distribution in the time domain to assess and evaluate the pore size distribution of the samples. Good agreement exists between water content from T2 NMR distributions and water imbibition porosity (WIP) technique. Results show that the most useful characteristics to discriminate the different facies come from full saturation NMR 2 MHz pore size distribution (PSD). Shale facies depict a quasi-unimodal distribution with greater than 90% contribution from clay bound water component (T2s) coupled to capillary bound water component (T2i) centred on 2 ms. The other facies with well connected pore structure show either bimodal or trimodal T2 distribution composed of the similar clay bound water component centred on 0.3 ms and quasi-capillary bound water component centred on 10 ms. But their difference depends on the movable water T2 component (T2l) that does not exist in the glauconite facies (bimodal distribution) while it exists in both the sandstone and limestone facies. The basic difference between the limestone and sandstone facies is related to the longer T2 coupling: T2i and T2l populations are coupled in sandstone generating a single population which convolves both populations (bimodal distribution). Limestone with a trimodal distribution attests to the fact that carbonate rocks have more complex pore system than siliclastic rocks. The degree of pore connectivity is highest in sandstone, followed by limestone and least in glauconite. Therefore a basic/quick NMR log run on samples along a geological formation can provide precise lithofacies characterization with quantitative information on pore size, structure and distributions.
  • Article
    Full-text available
    The restricted platform facies dolomites are widely distributed in the northern and northeastern Guizhou, China, and they are the main lithology of karst groundwater and a primary target for locating drinking water resources in the province of Guizhou. The geological history and deposition environment of the restricted platform dolomite facies vary significantly across the province, making it more difficult to evaluate the dissolution and water-bearing capacity of karstic dolomites. This paper explores a quantitative approach to assessing the solubility of four typical geological strata using analytical hierarchy process (AHP) modeling. Properties including CaO/MgO, clast content, porosity, permeability and fracture density are investigated and used as the major criteria in the AHP analysis to render the rankings for solubility among the studied formations. Hydrogeological characteristics are also discussed. Overall the presented results demonstrate the feasibility of AHP as a modeling tool to provide a quantitative approach when a wide range of properties need to be considered. The explored approach can have strong implications for the exploitation and utilization of karst groundwater in Guizhou, China.
  • Article
    The lattice Boltzmann method (LBM) is used to simulate viscous fluid flow through a column of glass beads. The results suggest that the normalized velocity distribution is sensitive to the spatial resolution but not the details of the packing. With increasing spatial resolution, simulation results converge to a velocity distribution with a sharp peak near zero. A simple argument is presented to explain this result. Changes in the shape of the distribution as a function of flow rate are determined for low Reynolds numbers, and the large-velocity tail of the distribution is shown to depend on the packing geometry. The effect of a finite Reynolds number on the apparent permeability is demonstrated and discussed in relation to previous results in the literature. Comparison with velocity distributions from NMR (nuclear magnetic resonance) spectroscopy finds qualitative agreement after adjusting for diffusion effects in the NMR distributions.
  • Article
    Full-text available
    The transport properties (porosity, surface-to-volume ratio of the pore space, diffusion coefficient, and permeability) of a porous medium were calculated by image analysis and random walk simulation using the digital image data on the pore structure of a bead pack (diameter 2.11 mm). A theory developed for laboratory experiments of nuclear magnetic resonance was applied to the random walk simulation. The three-dimensional data set (2563 voxels) of the bead pack was obtained by microfocus X-ray computed tomography at a spatial resolution of 0.053 mm. An original cluster labeling program, Kai3D.m, was used to estimate the porosity and surface-to-volume ratio. The surface-to-volume ratio and diffusion coefficient were calculated by an original random walk program, RW3D.m. The calculations were completed on a personal computer in reasonable time (
  • Article
    We presentstudy of the morphologynd bulk physical properties of  Fontainebleau sndstone vian X ray tomographicnalysis. Synchrotron-bsed X ray tomographic techniques provide us with a high-resolution (7.5 um), three- dimensional digitized representation of the sandstone that leaves the smple intact and unaltered. To estimatewide spectrum of bulk properties of the Fontainebleau sndstone specimen, we extract from this image a number of different correlation functions that statistically characterize the pore-space morphology and relevant pore-space length and time scales. These statistical measures are obtainable from lineal, plane, nd/or volume measurements and include the porosity, specific surface, two-point and three-point probability functions, lineal-path function, chord-length distribution function, pore-size distribution function, and coarseness. The pore- size distribution function, in particular, containscertain level of connectedness informationndccordingly cn only be obtained fromthree-dimensional representation of the smple. Mny bulk properties of the sndstone, such as the mean survival time r (obtainable from Nuclear Magnetic Resonance relaxation studies), fluid permeability k, effective electrical and thermal conductivities, and effective elastic moduli, cn be estimated using theforementioned statistical correlation functions. Specifically, the electrical conductivity (or, equivalently, the formation fctor F), mean survival time, and fluid permeability are determined using rigorous bounds. The mean survival timend fluid permeabilityre also found using direct simulation techniquesnd cross-property relations, respectively. One such cross-property relation for k depending on r and F givespermeability estimate that is withinfactor of 2 of the experimental result.
  • Article
    Full-text available
    Understanding of the transport properties of porous rocks is important for safe nuclear waste disposal because harmful contaminated groundwater can migrate along pore spaces over long distances. We developed three original Mathematica® version 5.2 programs to calculate the transport properties (porosity, pore connectivity, surface-to-volume ratio of the pore space, and anisotropic tortuosity of the pore structure) of porous rocks using three-dimensional (3-D) 8-bit TIFF or BMP X-ray computed tomography (CT) images. The pre-processing program Itrimming.nb extracts a 3-D rectangular region of interest (ROI) from the raw CT images. The program Clabel.nb performs cluster-labeling processing of the pore voxels in the ROI to export volume, surface area, and the center of gravity of each pore cluster, which are essential for the analysis of pore connectivity. The random walk program Rwalk.nb simulates diffusion of non-sorbing species by performing discrete lattice walks on the largest (i.e., percolated) pore cluster in the ROI and exports the mean-square displacement of the non-sorbing walkers, which is needed to estimate the geometrical tortuosity and surface-to-volume ratio of the pore. We applied the programs to microfocus X-ray CT images of a rhyolitic lava sample having an anisotropic pore structure. The programs are available at http://www.jstage.jst.go.jp/browse/jnst/44/9/ and http://staff.aist.go.jp/nakashima.yoshito/progeng.htm to facilitate the X-ray CT approach to groundwater hydrology.
  • Article
    Wettability profoundly affects not only the initial distribution of residual NAPL contaminants in natural soils, but also their subsequent dissolution in a flowing aqueous phase. Under conditions of preferential NAPL wettability, the residual NAPL phase is found within the smaller pores and in the form of continuous corner filaments and thick films on pore walls. Such films expose a much greater interfacial area for mass transfer than would be exposed by the same amount of non-wetting NAPL. Importantly, capillary and hydraulic continuity of NAPL filaments and thick films is essential for sustaining NAPL–water counterflow during the course of NAPL dissolution in flowing groundwater—a mechanism which maintains and even increases the interfacial area for mass transfer. Continued dissolution results in gradual thinning of the NAPL films, which may become unstable and rupture causing disconnection of the residual NAPL in the form of clusters. Using a pore network simulator, we demonstrate that NAPL film instability drastically modifies the microscopic configuration of residual NAPL, and hence the local hydrodynamic conditions and interfacial area for mass transfer, with concomitant effects on macroscopically observable quantities, such as the aqueous effluent concentration and the fractional NAPL recovery with time. These results strongly suggest that the disjoining pressure of NAPL films may exert an important, and hitherto unaccounted, control on the dissolution behaviour of a residual NAPL phase in oil wet systems.