Vous êtes sur la page 1sur 9

Precise 3D Numerical Modeling

of Fracture Flow Coupled With X-Ray


Computed Tomography for Reservoir
Core Samples
N. Watanabe, T. Ishibashi, N. Hirano, and N. Tsuchiya, Tohoku University; Y. Ohsaki and T. Tamagawa,
Japan Petroleum Exploration Company Limited; and Y. Tsuchiya and H. Okabe,
SPE, Japan Oil, Gas and Metals National Corporation

Summary
The present study focuses on the feasibility of a precise 3D
numerical modeling coupled with X-ray computed tomography
(CT), which enables simple analysis of heterogeneous fracture
flows within reservoir core samples, as well as the measurement
of porosity and permeability. A numerical modeling was developed and applied to two fractured granite core samples. One of
the samples had an artificial single fracture (sample dimensions:
100 mm in diameter, 150 mm in length), and the other had natural multiple fractures (sample dimensions: 100 mm in diameter,
120 mm in length). A linear relationship between the CT value
and the fracture aperture (fracture-aperture calibration curve) was
obtained by X-ray CT scanning for a fracture-aperture calibration
standard while varying the aperture from 0.1 to 0.5 mm. With the
fracture-aperture calibration curve, 3D distributions of the CT
value for the samples (voxel dimensions: 0.350.350.50 mm3)
were converted into fracture-aperture distributions in order to
obtain fracture models for these samples. The numerical porosities reproduced the experimental porosities within factors of
approximately 1.3 and 1.1 for the single fracture and the multiple
fractures, respectively. Using the fracture models, a single-phase
flow simulation was also performed with a local cubic law-based
fracture-flow model for steady-state laminar flow of a viscous
and incompressible fluid. The numerically obtained permeabilities
were larger than the experimentally obtained permeabilities by
factors of approximately 2.2 and 2.7 for the single fracture and
the multiple fractures, respectively. However, these discrepancies
can be reduced to approximately 1.32.1 and 1.62.6, respectively,
by simply using the correction factor for the cubic-law equation
proposed by Witherspoon et al. (1980). Consequently, a precise
numerical modeling coupled with X-ray CT is essentially feasible.
Furthermore, the development of preferential flow paths (i.e.,
channeling flow) was clearly demonstrated for multiple fractures,
which is much more challenging to achieve by most other methods.
Further progress in modeling should enable the in-situ evaluation
of heterogeneous fracture flow within reservoir core samples, as
well as the clarification of the impacts of the heterogeneity on the
productivity of wells and, for example, the efficiency of recovery
by water-/gasflooding.
Introduction
Effective development of oil/gas fractured reservoirs requires a
good understanding of the fracture-flow characteristics in these reservoirs. Field and laboratory studies have suggested that fluid flow
through a rock fracture differs significantly from that through smooth
parallel plates because of the channeling flow in a heterogeneous

Copyright 2011 Society of Petroleum Engineers


This paper (SPE 146643) was revised for publication from paper IPTC 13164, first presented
at the International Petroleum Technology Conference, Doha, Qatar, 79 December 2009.
Original manuscript received for review 18 February 2010. Revised manuscript received for
review 7 December 2010. Paper peer approved 27 January 2011.

September 2011 SPE Journal

aperture distribution generated by rough surfaces (Cacas et al.


1990; Abelin et al. 1991; Dverstorp et al. 1992; Carmeliet et al.
2004; Johnson et al. 2006; Rowland et al. 2008; Watanabe et al.
2008, 2009; Nemoto et al. 2009). When channeling flow occurs
in a single fracture of granite, the area in which the flowing fluid
is present is expected to be only 5 to 20% of the total area of the
fracture at confining pressures of up to 100 MPa, with various features in the preferential flow paths (Watanabe et al. 2009). With this
in mind, we have recently developed a discrete fracture-network
model simulator, GeoFlow, in which fractures can have aperture
distributions with natural heterogeneities (Ishibashi et al. 2009).
3D fluid-flow simulations for fracture networks by GeoFlow have
demonstrated the development of 3D preferential flow paths in
these fracture networks and have suggested that the performance
of a production well can be strongly affected by the heterogeneous
fracture flow. Consequently, evaluating the heterogeneity of the
flow and clarifying its impact on the effective development of
fractured reservoirs are of considerable importance.
In order to address this problem, fluid-flow analyses for reservoir core samples may be the only direct approach to obtain
insights into heterogeneous fracture flows originating from natural
heterogeneities of a specific fractured reservoir. However, fracture
flows in naturally fractured rock core samples, which often contain
randomly distributed multiple fractures having different intensities,
are usually difficult to evaluate. Although fracture-flow analyses
by numerical fracture models with heterogeneous aperture distributions are effective (Watanabe et al. 2008, 2009; Nemoto et al.
2009), for most existing methods, a precise aperture determination
is much more challenging in the case of multiple fractures. Because
available reservoir core samples are limited and considerably valuable, the development of a 3D numerical modeling of a fracture
flow, which is coupled with a nondestructive precise aperture
determination for an in-situ reservoir core sample, is needed.
X-ray CT has been one of the most effective nondestructive
methods for analyzing internal structures in opaque materials,
such as fractures in rocks, which sometimes contain fluid flows
(Ketcham and Carlson 2001; Polak et al. 2003, 2004; Liu et al.
2006; Karpyn et al. 2007, 2008; Zhu et al. 2007). Moreover,
previous studies have demonstrated the possibility of a precise
fracture-aperture determination by X-ray CT, sometimes with a
fracture-flow simulation with the apertures (Pyrak-Nolte et al.
1997; Johns et al. 1993; Keller 1998; Keller et al. 1999; Montemagno and Pyrak-Nolte 1999; Bertels et al. 2001). Consequently,
a 3D numerical model analysis of fracture flow with an aperture
distribution by X-ray CT is one of the most favorable options for
reservoir core samples. However, the relevance of such methods
in previous studies remains unclear because of the lack of quantitative evaluation of the numerical results based on independently
obtained experimental results, such as experimentally determined
porosities and/or permeabilities. Furthermore, scant data have been
reported for multiple fractures.
Therefore, in the present study, we focused on the feasibility of a
precise 3D numerical modeling of fracture flow coupled with X-ray
CT for reservoir core samples to provide insight into heterogeneous
683

Sample

X-ray CT scanner

Fig. 1Third-generation medical X-ray CT scanner at the


JOGMEC TRC.

fracture flows, such as channeling flow, in reservoirs. A third-generation medical X-ray CT scanner, which provides a relatively high
voxel resolution with a voxel size of less than 0.1 mm3, was used
for precise imaging of fractured granite core samples with either
an artificial single fracture or natural multiple fractures. Fracture
models (i.e., aperture distributions) obtained by the CT images were
evaluated by comparison of the numerically and experimentally
determined porosities of the samples. In addition, a single-phase
fluid flow simulation was performed for the fracture models using
a local cubic law-based fracture-flow model. The fracture-flow
model was also evaluated using the experimentally determined
permeabilities of the samples.
Methods
X-Ray CT. In order to develop 3D models of fractured rock core
samples, it is necessary to convert CT images of the rock core
samples into the 3D fracture-aperture distributions within the samples using a fracture-aperture calibration curve (Johns et al. 1993;
Keller 1998; Bertels et al. 2001). Therefore, we first scanned a fracture-aperture calibration standard with a known fracture aperture in
order to prepare the calibration curve. We then scanned fractured
rock core samples to obtain input data for the modeling.
X-ray CT scanning was performed at the Japan Oil, Gas and
Metals National Corporation Technology and Research Center
(JOGMEC TRC), with the AquilionTM third-generation medical
scanner produced by Toshiba Medical Systems Corporation (Fig. 1).
All of the scans in the present study were performed using dry
samples at room temperature and atmospheric pressure with identical parameter values (Table 1). The CT scanner provides a 3D
distribution of the CT value (a measure of X-ray attenuation), with
a voxel resolution of 0.350.350.50 mm3, within several minutes
for the sample dimensions used in the present study. Although the
resolution is relatively low as compared with industrial microfocus
scanners, medical scanners have an advantage with respect to the
size and weight of the sample and equipment. It is much easier to
scan an entire reservoir core sample that is situated in a relatively

TABLE 1VALUES OF THE PARAMETERS OF THE X-RAY


CT SCANNER
Parameter

684

Value

Voltage

1 2 0 kV

Current

150 mA

Slice thickness

0.50 mm

Pixel resolution of slice

0.35 mm

Pixel matrix size

512 512

large and heavy core holder using a medical scanner. This will
provide a considerable advantage in the in-situ analysis of heterogeneous fracture flows in future studies.
The CT value at each voxel depends primarily on the density
of the material. The CT value increases with increasing density.
Because of this fundamental principle, we can detect the presence
of a fracture as an apparent density reduction, as compared with the
original density of the material, even if the fracture is smaller than
the voxel. Moreover, the apparent density reduction depends on the
pore volume, which in turn depends on the fracture aperture. Consequently, we can convert the CT value into the fracture aperture
by obtaining a relationship between the CT value and the fracture
aperture (a fracture-aperture calibration curve). It should be noted
that we used the minimum CT value caused by the fracture when
preparing the calibration curve, rather than the integrated value
as shown in the literature (Johns et al. 1993; Keller 1998; Bertels
et al. 2001). This was done because the integrated value can be
calculated only for samples having fractures with known locations
and orientations, whereas naturally fractured rock core samples
often contain fractures with unknown locations and orientations.
The fracture-aperture calibration standard with a known fracture aperture was cylindrical granite having a simulated single
fracture with a unique aperture. The fracture was prepared by
cutting a piece of cylindrical granite, 100 mm in diameter and 45
mm in length. The granite was Inada granite, which was quarried
in Ibaraki, Japan, and had a matrix permeability of 1019 to 1018
m2. The fracture surfaces were polished and had a mean roughness of approximately 20 to 30 m. The aperture was created by
inserting a thickness gauge between the surfaces. The calibration
standard was scanned with varying apertures of from 0.1 to 0.5
mm by changing the thickness of the gauge.
Because it is important to develop a 3D numerical modeling not
only for a single fracture but also for multiple fractures, considering that reservoir core samples usually contain multiple fractures,
we used two fractured granite core samples, which contained either
a single fracture or multiple fractures, with known porosities and
permeabilities to evaluate the relevance of the modeling. One of the
samples (Sample 1) contained an artificial single fracture, and the
other (Sample 2) contained natural multiple fractures. The diameter
and length of the sample were 100 mm and 150 mm for Sample
1, and 100 mm and 120 mm for Sample 2, respectively. Sample
1 was prepared using the following procedure. A 200-mm cubic
Inada granite block was fractured using a wedge to create a tensile
fracture. The block was then fixed with concrete to achieve a mated
fracture. Finally, the block was cored and cut to the prescribed
dimensions. Sample 2 was prepared by cutting a drill core, which
was taken from granite bedrock at a depth of approximately 4500
to 4600 m in Hokkaido, Japan, to length.
Before the present study, the permeabilities and porosities of
both of the samples were determined at confining pressures from
3 to 34 MPa. The porosities were determined using a helium
porosimeter with a core holder, by reference to Boyles law. The
permeabilities were determined for straight flow geometry in the
axial direction of the samples, vertically from bottom to top, using
a steady-state air permeameter with a core holder, based on the
modified form of Darcys law for air permeability (API 1956). All
of the equipment used to determine the porosities was produced by
Core Laboratories Inc. As shown later herein, nonlinear decreases
in the porosity and, correspondingly, in the permeability with
increasing confining pressure were observed, most likely because
of the common behavior of the nonlinear closer for a fracture by
increasing contacting asperities.
Fracture and Fluid-Flow Modeling. Once a fracture-aperture
calibration curve is obtained, we can convert the CT values at
each voxel into fracture apertures, providing 3D fracture models of
the samples. Moreover, we can simulate the fluid flow within the
fracture models using existing fracture flow models for a variable
aperture fracture. Because the porosities and permeabilities can
be calculated for the models, comparisons between the numerical
and experimental results provide quantitative evaluations of the
models.
September 2011 SPE Journal

y = 0.35 mm

Voxel

z
=

0.

50

Fracture
aperture, a

x = 0.35 mm
Fig. 2Smooth parallel-plate single fracture for a voxel of
0.350.350.50 mm3 used in the fracture modeling.

In the fracture modeling, we assumed a smooth parallel-plate


single fracture with a unique aperture and orientation for each
voxel (Fig. 2), although, in reality, the voxels might contain not
only a single fracture, but also multiple fractures, with various
apertures and orientations. This is the same relationship between
fracture and voxel as seen in the determination of the fractureaperture calibration curve. Therefore, the CT values of the samples
were directly converted into fracture apertures with the aperture
calibration curve. Note that, in Fig. 2, the maximum fracture aperture in a voxel was y (0.35 mm), which means that even if the
fracture-aperture calibration curve provides a fracture aperture of
> 0.35 mm at a voxel, the fracture aperture is changed to 0.35 mm.
The calibration curve can also provide negative apertures at points
of the rock matrix. The negative apertures are changed to zero apertures. The porosity of a fracture model, , can be described as

M 1

i=0
M 1

i=0

ai x z

x yz

100,

. . . . . . . . . . . . . . . . . . . . . . . . . . (1)

where M is the total number of voxels, ai is the fracture aperture in


the ith voxel, and x, y, and z are the voxel dimensions.
For the modeling of a single-phase fluid flow in three dimensions (x-y-z coordinates), we used a local-cubic-law-based (LCLbased) fracture flow model for a steady-state laminar flow of a
viscous and incompressible fluid. The equation of continuity for
the Darcy flow can be described as follows (Watanabe et al. 2008,
2009; Nemoto et al. 2009; Brown 1987; Mourzenko et al. 1995;
Zimmerman and Bodvarsson 1996; Ge 1997; Oron and Berkowitz
1998; Yeo et al. 1998; Pyrak-Nolte and Morris 2000; Brush and
Thomson 2003; Konzuk and Kueper 2004):

voxel, although the fracture has a specific orientation in the fracture


modeling. Note that A and k are set to zero only at points with
zero apertures, which means that fluid flow can occur between a
fracture and the rock matrix when the rock matrix has nonzero
local aperture by grain-scale cracks.
By solving the finite-difference form of Eq. 2 at the boundary
conditions described next, we determined the flow-rate distributions
and permeabilities for the single-phase flow of the fluid having a
dynamic viscosity of 1103 Pas and a density of 1103 kg/m3.
The permeability measurements of the samples were performed
with straight flow geometry in the axial direction of the samples
vertically from bottom to top. Therefore, the boundary conditions
were given such that the same flow geometry was achieved where
a constant fluid potential difference of 1 MPa was applied between
the inflow and outflow boundaries, and the nonflow condition was
applied at the other boundaries. The permeability of the model,
K, which includes contributions of both the rock matrix and the
fracture parts, can be determined based on Darcys law:
K=

where P is the fluid potential difference, Q is the flow rate, and r


and L are the radius and length of the model (i.e., the dimensions
of the sample to be modeled), respectively.
Results and Discussion
Fracture-Aperture Calibration Curve. In the X-ray CT scanning for the fracture-aperture calibration standard, fracture signals
were observed as relatively large drops in the CT value (Fig. 3).
However, it was difficult to visually determine the fracture signal
for the fracture aperture of 0.1 mm because of the variation of the
CT value at the rock matrix. The variation could be caused by the
presence of rock-forming minerals having different densities, as
described later herein. The fracture signals for the fracture aperture
of 0.1 mm were therefore detected by automatically searching
the minimum CT value in the profile. In addition to the fracture
apertures of 0.1 to 0.5 mm, the mean CT value of the rock matrix
was calculated in order to determine the value of the zero fracture
aperture. In this calculation, the CT values near the sample edge,
from the edge to a distance of 5 mm radially inward from the
edge, were not used in order to exclude the elevated CT values
owing to the artifact generated by beam hardening (Ketcham and
Carlson 2001).
The relationship between the CT value and the fracture aperture
was linear (Fig. 4) and was well-fitted by the least-squares method
with the following equation:
ct = 2189a + 1690, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (6)

Ak P Ak P Ak P
= 0, . . . . . . . . . . (2)
+
+
x x y y z z

A = wa,

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3)

a2
,
12

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)

k=

where w and a are the local width and the local aperture of the
fracture, respectively. In the present study, w was assumed to
be equivalent to x in Fig. 2 (no significant difference between
x and z), and the product Ak was assumed to be isotropic at a
September 2011 SPE Journal

3000

CT value (Hounsfield units)

where A is the cross-sectional area in which the fluid flows, k is the


local permeability, is the dynamic viscosity, and P is the fluid
potential defined by P = p + gz, in which p is the fluid pressure,
is the density of the fluid, and g is the gravitational acceleration.
In the LCL, A and k in Eq. 2 are given as follows:

QL
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (5)
r 2 P

Sample
edge

Sample
edge

2500
2000
1500
1000
500

Fracture signal

0
-500

-1000

100

200
300
Position (pixels)

400

500

Fig. 3Example of the fracture signal observed in the X-ray CT


scanning for the fracture-aperture calibration standard.
685

Sample 1(Single fracture )

CT value (Hounsfield units)

2500

2000

1500
A

1000

500

0
0.0

0.1

0.2

0.3

0.4

0.5

0.6
(a)

Fracture aperture (mm)


Fig. 4Relationship between the CT value and the fracture
aperture as determined by X-ray CT scanning for the fractureaperture calibration standards. The data point and the error
bar indicate the arithmetic mean and the standard deviation
by calculations from 100 data, respectively. The solid line
indicates the linear-regression curve obtained by the leastsquares method.

Sample 2 (Multiple fractures)


H

where ct is the CT value and a is the fracture aperture. Moreover,


the coefficient of determination R2 by the linear fit was 0.99,
indicating good linearity of the fracture-aperture calibration curve
between fracture apertures of from zero to 0.5 mm. The CT value
at the intercept (a = 0) was 1,690, which was quite similar to the
mean CT value of granite (1,680) by Keller (1998), and showed
the threshold between zero and nonzero fracture apertures in the
modeling. The relatively large deviation of the CT value from the
regression curve was observed at the fracture aperture of 0.1 mm.
However, this may have been caused by the difficulty in exactly
creating the smallest fracture aperture, and the actual fracture
aperture could be slightly greater than 0.1 mm. Therefore, the linear relationship was determined to hold even at fracture apertures
smaller than 0.1 mm.

686

(b)

-1000

1000

2000

3000

4000

CT value (Hounsfield units)


Fig. 5CT images of (a) Sample 1 and (b) Sample 2.

Fracture Models. With the fracture-aperture calibration curve


described by Eq. 6, fracture modeling was performed for the
fractured granite core samples. In the modeling, the CT values
near the sample edges were not used because of the artifact, as
-12

10

Sample 1 (Single fracture)


Sample 2 (Multiple fractures)
-13

Permeability (m2)

CT Images of Fractured Granite Core Samples. As described


previously, the CT images of the samples also exhibited variations in the CT value for the rock matrix and elevated values at
the sample edges (Fig. 5). A wide variation in the CT value was
observed, as indicated by the light grey regions, where the CT
values were approximately 1,600 to 2,500. This variation could be
responsible for the rock-forming minerals with different densities
because the shape and distribution appeared similar to those of
minerals. However, the aperture-calibration curve with a CT value
of 1,690 at the zero fracture aperture suggests that this variation
has a less significant effect in the aperture determination because
most CT values in the variation were larger than 1,690 and were
recognized as the rock matrix in the modeling.
In the CT image of Sample 1, the single fracture was observed
as a relatively darker gray curve in the vertical direction (Fig.
5a). Because black areas indicate signals of air, the gray fracture
signal suggests that the apertures were smaller than the voxel size,
which results in the unclear imaging of the fracture shape and
aperture. Similarly, in the CT images of Sample 2, fractures were
not observed clearly (Fig. 5b). This is a limitation in imaging by
a medical X-ray CT scanner. Using a medical X-ray CT simply
as an imaging tool usually provides unclear information regarding
the fracture flow when a fracture has smaller apertures than the
voxel size. However, as shown later herein, fractures with such
small apertures have significantly larger permeabilities compared
with the rock matrix of 1019 to 1018 m2 (Fig. 6), and can therefore
play a major role in fluid migration in the reservoirs. Although
recent microfocus X-ray CT enables pore-scale imaging, a number of limitations remain regarding the available size and weight
of samples and equipment, and more effective use of the medical
X-ray CT, as described in the present study, is preferable.

10

Numerical results

-14

10

-15

10

-16

10

10

15

20

25

30

35

40

45

50

Confining pressure (MPa)


Fig. 6Comparison of the numerical and experimental permeabilities of the samples. The experimental permeabilities were
measured at confining pressures of from 3 to 35 MPa and were
used to predict the permeabilities at atmospheric pressure
(solid lines), which were compared with the numerical permeabilities generated by the fracture-flow models of the samples
at atmospheric pressure.
September 2011 SPE Journal

Frequency (%)

8
6
4
2
0
0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Fracture aperture (mm)


Sample 2 (Multiple fractures)

10

Frequency (%)

8
6
4
2
0
0.00

0.05

0.10 0.15
0.20 0.25
Fracture aperture (mm)

0.30

0.35

Fig. 7Fracture aperture histograms of the fracture models


of (a) Sample 1 and (b) Sample 2. Note that the zero fracture
aperture is excluded in the histograms.

noted previously. Consequently, the dimensions of the models were


slightly smaller than the sample dimensions. The dimensions of the
models were 90 mm in diameter and 150 mm in length for Sample
1, and 90 mm in diameter and 120 mm in length for Sample 2. The
fracture models, which are described by Eq. 4 as 3D distributions
of the permeability, are presented in the next section, together with
the results of the fluid-flow simulation. This section focuses only
on the results obtained in the fracture-aperture determination using
the calibration curve and the evaluation of these results through
comparison with experimentally obtained porosities.
According to fracture-aperture histograms of the fracture models for the samples (except for zero fracture aperture), most of the
fracture apertures were smaller than the voxel dimensions (i.e.,
< 0.35 mm), as expected from the CT images (Fig. 7). The samples
appeared to have fracture apertures larger than the voxel dimensions (i.e., 0.35 mm), as shown in the right column of the histograms. However, these frequencies were approximately 1% for
Sample 1 and were smaller than 1% for Sample 2. These frequencies are quite small for such larger fracture apertures, and most
of the fracture apertures were less than 0.2 mm. The histograms
appeared to be highly skewed and to have long tails; the log-normal
distribution, rather than the Gaussian distribution, might be better
suited for describing the fracture-aperture distributions, although
the distributions were not perfectly log-normal because of the
cutoff caused by the presence of the zero fracture aperture. This
type of fracture-aperture distribution has been reported for single
fractures of granite by aperture determinations based on fracture
surface mapping in previous studies (Watanabe et al. 2008, 2009;
Nemoto et al. 2009), which suggests that the present fracture-aperture determination could provide qualitatively reasonable results
even for fracture apertures smaller than the voxel dimensions.
September 2011 SPE Journal

The numerical porosities were 1.3 and 6.3% for Samples 1 and
2, respectively. Because the X-ray CT scanning was performed at
atmospheric pressure, the numerical results corresponded to the
samples at that condition. Although we obtained the porosities of
the samples, these results were obtained at confining pressures of
from 3 to 34 MPa. Therefore, we compared the numerical porosities with the experimental porosities at atmospheric pressure; these
were inferred by extrapolation of the experimental results (Fig. 8).
Based on the extrapolation, the experimental porosities at atmospheric pressure were determined to be approximately 1.0 and
5.5% for Samples 1 and 2, respectively. Consequently, the numerical porosities could predict the experimental porosities within factors of approximately 1.3 and 1.1 for Samples 1 and 2, respectively.
This discrepancy may have been caused primarily by the variation
in the CT value of the rock matrix. Because points with CT values
of <1,690 may be present (Figs. 3 and 5), the fracture-aperture
calibration curve could provide nonzero fracture apertures even for
the rock matrix. The variation of the CT values is unavoidable and
may limit the degree of accuracy in the modeling.
Fluid-Flow Models. Using the fracture models, the single-phase
fluid-flow modeling was performed for the fractured granite core
samples. In addition to the original voxel resolution, voxel resolutions equal to one-half and one-third of the original voxel resolution were used to examine the possibility of the reduction of computational cost, because the total number of voxels obtained by the
original voxel resolution can be enormous (for example, more than
1.0107 in the present study). The data for the voxel resolutions
equal to one-half (voxel dimensions: 0.700.701.00 mm3) and
one-third (voxel dimensions: 1.051.051.50 mm3), respectively,
of the original resolution were created so that the pore volume of
the original voxel resolution could be preserved.
The numerical permeabilities obtained using the original voxel
resolution were 4.31014 m2 and 1.91013 m2 for Samples 1 and
2, respectively (Fig. 6). In the same manner as in the preceding
section, the experimental permeabilities at atmospheric pressure
were inferred by the extrapolation, yielding permeabilities of
approximately 2.01014 m2 and 7.01014 m2 for Samples 1 and
2, respectively. The numerical permeabilities were larger than the
experimental permeabilities by factors of approximately 2.2 and
2.7 for Samples 1 and 2, respectively. One possible reason for
this discrepancy is the degree of accuracy in the fracture model.
However, this discrepancy is inevitable, as noted previously. The
other possible reason is the limitation of the cubic-law equation
(Brown 1987). Even on the local scale, the fracture may not be
perfectly modeled by smooth parallel plates. The deviations from
the ideal smooth parallel plates may only result in an apparent
10
Sample 1 (Single fracture)
Sample 2 (Multiple fractures)

9
8
Porosity (%)

Sample 1 (Single fracture)

10

Numerical results

6
5
4
3
2
1
0

10

15

20

25

30

35

40

45

50

Confining pressure (MPa)


Fig. 8Comparison of the numerical and experimental porosities of the samples. The experimental porosities were measured at confining pressures of from 3 to 34 MPa and were used
to infer the porosities at atmospheric pressure (solid lines),
which were compared with the numerical porosities.
687

10

-11

Permeability distribution

Flow rate distribution

Permeability (m )

Sample 1 (Single fracture)


Sample 2 (Multiple fractures)
10

10

10

-12

-13

-14

0.0

0.2

0.4

0.6

0.8 1.0 1.2 1.4


Voxel size (mm3)

1.6

1.8

2.0

Fig. 9Numerical permeabilities of the models of the samples


at the original voxel resolutions and one-half and one-third the
original voxel resolutions, where the voxel sizes are 0.061, 0.49,
and 1.65 mm3, respectively.

reduction in the flow rate, which may be incorporated into the flow
model (Eq. 2) by the correction factor proposed by Witherspoon
et al. (1980):
1 Ak P 1 Ak P 1 Ak P
+
= 0,
+
x f x y f y z f z

(a)

(b)

Fig. 10. (a) Permeability distribution and (b) flow-rate distribution for the model of Sample 1 with a single fracture. The flowrate distribution demonstrates the formation of preferential
flow paths within the single fracture having the permeability
distribution. The boundary conditions were given so that the
straight flow geometry, in the z (blue axis) direction, vertically
from bottom to top, was achieved, where the constant fluid
potential difference was applied between the inflow and outflow
boundaries and the nonflow condition was applied at the other
boundaries.

Permeability distribution

Flow rate distribution

. . . . . . . . . . . . . . . . . . . . . . . . (7)
where f is the correction factor, which is reported to vary from
1.04 to 1.65, depending on conditions. Considering the range of
the correction factor, the discrepancy in the permeability can be
improved to factors of approximately 1.32.1 and 1.62.6 for
Samples 1 and 2, respectively. However, the optimum value for
the present model may be larger (i.e., approximately 2.2 to 2.7),
which should be addressed in future studies.
The data for the one-half and the one-third voxel resolutions
were created by using the following procedure. At each point, a
local pore volume V was calculated by the following equation:

Fracture 1

Fracture 2

(a-1)

(b-1)

Permeability distribution

V = i = 0 ai x z ,
N 1

Flow rate distribution

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (8)

where N is the number of neighboring voxels that are combined


when creating a larger voxel (N equals 8 and 27 for the one-half and
the one-third resolutions, respectively), ai is the fracture aperture
in the ith voxel, and x and z are the original voxel dimensions.
After this calculation, a modified fracture aperture am was assigned
to that point by the following equation:

Fracture 2
Fracture 3
Fracture 4

am =

V
,
X Z

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (9)

where X and Z are the voxel dimensions after changing the voxel
resolution (X and Z are 2x and 2z for the one-half resolution
and 3x and 3z for the one-third resolution, respectively). The
numerical permeability increased with decreasing voxel resolution
for both samples (Fig. 9), perhaps because of reduction of points
having the zero aperture. However, the permeabilities for the onehalf and the one-third resolutions (the voxel sizes are 0.49 mm3
and 1.65 mm3, respectively) were within one order of magnitude
of those for the original resolution (0.061 mm3). Furthermore,
the fluid-flow paths formed in the models with the original voxel
resolutions, which are shown later herein (Figs. 10 and 11),
were not significantly changed by changing the voxel resolution.
688

(a-2)

(b-2)

Fig. 11. (a-1) and (a-2) Permeability distributions and (b-2)


and (b-2) flow-rate distributions from two different angles for
the model of Sample 2 with multiple fractures. The flow-rate
distribution demonstrates the formation of preferential flow
paths within the multiple fractures having the permeability
distribution. The boundary conditions were given so that the
straight flow geometry, in the z (blue axis) direction, vertically
from bottom to top, was achieved, where the constant fluid
potential difference was applied between the inflow and outflow
boundaries and the nonflow condition was applied at the other
boundaries.
September 2011 SPE Journal

Consequently, it may be possible to use the one-half and the onethird resolutions for rough analyses.
The present study demonstrated that even the present simple
numerical modeling could reproduce the experimental results
for the fractured granite core samples, with relatively small
discrepancies, even for the multiple fractures having apertures
smaller than the voxel dimensions. Consequently, a precise numerical modeling of the fracture flow coupled with X-ray CT for reservoir core samples is essentially feasiblealthough further studies
are required to obtain an improved fracture-flow modeland
would be an effective method for analyzing heterogeneous fracture flow, as demonstrated next. In the permeability distributions
of Sample 1, relatively larger permeabilities, as shown in color,
were distributed in a single plane because Sample 1 had a single
fracture (Fig. 10a). Furthermore, the flow-rate distribution clearly
demonstrated the formation of preferential flow paths (i.e., channeling flow) within the single fracture (Fig. 10b), which is logical
and agrees well with the literature (Watanabe et al. 2008, 2009;
Nemoto et al. 2009). In contrast, the permeability distribution
of Sample 2 was different from that of Sample 1 (Figs. 11a and
11b). Relatively larger permeabilities were distributed over the
entire volume, indicating that several fractures were present within
Sample 2. At least four fractures (Fractures 1 through 4) were easily identified, and these fractures appeared to form a network. The
flow-rate distribution clearly demonstrated a 3D channeling flow
in the fracture network. This phenomenon is expected in fractured
reservoirs (Ishibashi et al. 2009). The detailed characteristics of
this phenomenon should vary depending on conditions such as
different fracture and aperture distributions, stress and temperature
conditions, and fluid types. Analyses performed with varying these
conditions are therefore necessary, and this problem may be easily
addressed only by approaches such as that proposed in the present
study. In future studies, in-situ X-ray CT scanning (Polak et al.
2003, 2004; Liu et al. 2006; Zhu et al. 2007) and a multiphaseflow model for a variable-aperture fracture (Murphy and Thomson
1993) should be performed in order to achieve more valuable
analyses of heterogeneous fracture flow of entire reservoir core
samples under in-situ conditions, and clarification of the impacts
of the heterogeneity on the productivity of wells and the efficiency
of recovery by water-/gasflooding, for example.
Conclusions
Effective development of oil/gas fractured reservoirs requires a
clear understanding of the fracture-flow characteristics in these
reservoirs. Field and laboratory studies have suggested that fluid
flow through a rock fracture differs from that through smooth
parallel plates because of the channeling flow in a heterogeneous
aperture distribution by rough surfaces. For fracture networks of
reservoirs, it has been suggested that the development of 3D preferential flow paths affects the performance of production wells.
Consequently, it is of considerable importance to evaluate the heterogeneity and its impact on the effective development of fractured
reservoirs. In order to address this problem, fluid-flow analyses
for reservoir core samples may be the only direct approach for
obtaining insight into the heterogeneous fracture flow originating from natural heterogeneities of a specific fractured reservoir.
However, fracture flow in naturally fractured rock core samples,
which often contain multiple fractures, is usually very difficult to
evaluate. Although fracture-flow analyses by numerical fracture
models with heterogeneous aperture distributions are effective,
a precise aperture determination using most existing methods is
much more challenging for multiple fractures. Because available
reservoir core samples are limited and are of considerable value,
a numerical fracture-flow modeling coupled with a nondestructive
precise aperture determination is required.
Therefore, the present study focused on the feasibility of a
precise 3D numerical modeling coupled with X-ray CT, which
enables simple analyses of the heterogeneous fracture flow within
reservoir core samples, as well as measurements of porosity and
permeability. The numerical modeling developed in the present
study was applied to two fractured granite core samples, which
September 2011 SPE Journal

had either an artificial single fracture or natural multiple fractures.


The present fracture models could provide numerical porosities,
which reproduced the experimental porosities within factors of
approximately 1.3 and 1.1 for the single fracture and multiple fractures, respectively. On the other hand, the single-phase fluid-flow
simulation, with the local-cubic-law-based fracture-flow model,
provided numerical permeabilities that were larger than the experimental permeabilities by factors of approximately 2.2 and 2.7 for
the single fracture and multiple fractures, respectively. However,
these discrepancies can be reduced to approximately 1.32.1 and
1.62.6, respectively, using the correction factor for the cubic-law
equation proposed by Witherspoon et al. (1980). Consequently, a
precise numerical modeling coupled with X-ray CT is essentially
feasible. Furthermore, the development of preferential flow paths
(i.e., channeling flow) was clearly demonstrated for multiple fractures, which is much more challenging to achieve by most other
methods. Further progress in this type of numerical modeling
should enable more valuable analyses of heterogeneous fracture
flow of entire reservoir core samples under in-situ conditions and
should clarify the impacts of the heterogeneity on the development
of fractured reservoirs.
Nomenclature
a = local fracture aperture, L
ai = fracture aperture in the ith voxel, L
am = modified fracture aperture after changing the voxel
resolution, L
A = cross-sectional area in which the fluid flows, L2
ct = CT value, Hounsfield unit
f = correction factor for the cubic law equation
g = gravitational acceleration, LT2
k = local permeability, L2
K = permeability, L2
L = length of the model, L
M = total number of voxels
N = number of neighboring voxels that are combined
when creating a larger voxel
P = fluid potential, ML1T2
Q = flow rate, L3T1
r = radius of the model, L
V = local pore volume, L3
w = local fracture width, L
x, y, z = spatial coordinates in the Cartesian coordinate
system, L
P = fluid potential difference, ML1T2
x, y, z = voxel dimensions in the Cartesian coordinate
system, L
X, Y, Z = voxel dimensions in the Cartesian coordinate system
after changing the voxel resolution, L
= dynamic viscosity, ML1T1
= circle ratio
= density of the fluid, ML3
= porosity, volume %
Subscript
i = spatial index
Acknowledgments
The authors would like to thank Kimio Watanabe at RichStone,
Limited (currently at Renergies, Limited) for the coding of the
modeling algorithm. The present study was supported in part by
the Ministry of Education, Culture, Sports, Science and Technology (MEXT) through a Grant-in-Aid for Scientific Research on
Innovative Areas, No. 22107502.
References
Abelin, H., Birgersson, L., Moreno, L., Widen, H., gren, T., and Neretnieks, I. 1991. A Large-Scale Flow and Tracer Experiment in Granite:
689

2. Results and Interpretation. Water Resour. Res. 27 (12): 31193135.


doi: 10.1029/91WR01404.
API RP 27, Recommended Practice for Determining Permeability of Porous
Media, third edition. 1952. Washington, DC: API.
Bertels, S.P., DiCarlo, D.A., and Blunt, M.J. 2001. Measurement
of Aperture Distribution, Capillary Pressure, Relative Permeability, and in Situ Saturation in a Rock Fracture using Computed
Tomography Scanning. Water Resour. Res. 37 (3): 649662. doi:
10.1029/2000WR900316.
Brown, S.R. 1987. Fluid flow through rock joints: The effect of surface
roughness. Journal of Geophysical Research 92 (B2): 13371347. doi:
10.1029/JB092iB02p01337.
Brush, D.J. and Thomson, N.R. 2003. Fluid flow in synthetic rough-walled
fractures: Navier-Stokes, Stokes and local cubic law simulations. Water
Resour. Res. 39 (4): 1085. doi: 10.1029/2002WR001346.
Cacas, M.C., Ledoux, E., de Marsily, G., Tillie, B., Barbreau, A., Durand,
E., Feuga, B., and Peaudecerf, P. 1990a. Modeling Fracture Flow With
a Stochastic Discrete Fracture Network: Calibration and Validation 1.
The Flow Model. Water Resour. Res. 26 (3): 479489. doi: 10.1029/
WR026i003p00479.
Carmeliet, J., Delerue, J.-F., Vandersteen, K., and Roels, S. 2004. Threedimensional liquid transport in concrete cracks. Int. J. Numer. Anal.
Meth. Geomech. 28 (78): 671687. doi: 10.1002/nag.373.
Dverstorp, B., Andersson, J., and Nordqvist, W. 1992. Discrete fracture network interpretation of field tracer migration in sparsely fractured rock.
Water Resour. Res. 28 (9): 23272343. doi: 10.1029/92WR01182.
Ge, S. 1997. A governing equation for fluid flow in rough fractures. Water
Resour. Res. 33 (1): 5361. doi: 10.1029/96WR02588.
Ishibashi, T., Watanabe, N., Hirano, N., Okamoto, A., and Tsuchiya, N.
2009. Development of Discrete Fracture Network Model Simulator,
GeoFlow, for Evaluation of Three Dimensional Channeling Flow. Paper
IPTC 13143 presented at the International Petroleum Technology Conference, Doha, Qatar, 79 December. doi: 10.2523/13143-MS.
Johns, R.A., Steude, J.S., Castainer, L.M., and Roberts, P.V. 1993. Nondestructive Measurements of Fracture Aperture in Crystalline Rock
Cores Using X Ray Computed Tomography. J. Geophys. Res. 98 (B2):
18891900. doi: 10.1029/92JB02298.
Johnson, J., Brown, S., and Stockman, H. 2006. Fluid flow and mixing in
rough-walled fracture intersections. J. Geophys. Res. 111: B12206. doi:
10.1029/2005JB004087.
Karpyn, Z.T., Alajmi, A., Radaelli, F., Halleck, P.M., and Grader, A.S.
2008. X-ray CT and hydraulic evidence for a relationship between
fracture conductivity and adjacent matrix porosity. Engineering Geology 103 (34): 139145. doi: 10.1016/j.enggeo.2008.06.017.
Karpyn, Z.T., Grader, A.S., and Halleck, P.M. 2007. Visualization of
fluid occupancy in a rough fracture using micro-tomography. Journal
of Colloid and Interface Science 307 (1): 181187. doi: 10.1016/
j.jcis.2006.10.082.
Keller, A. 1998. High resolution, non-destructive measurement and characterization of fracture apertures. Int. J. Rock Mech. Min. Sci. 35 (8):
10381050. doi: 10.1016/S0148-9062(98)00164-8.
Keller, A.A., Roberts, P.V., and Blunt, M.J. 1999. Effect of fracture aperture
variations on the dispersion of contaminants. Water Resour. Res. 35 (1):
5563. doi: 10.1029/1998WR900041.
Ketcham, R.A., and Carlson, W.D. 2001. Acquisition, optimization and
interpretation of X-ray computed tomographic imagery: applications to
the geosciences. Comp. Geosci. 27 (4): 381400. doi: 10.1016/S00983004(00)00116-3.
Konzuk, J.S., and Kueper, B.H. 2004. Evaluation of cubic law based models
describing single-phase flow through a rough-walled fracture. Water
Resour. Res. 40: W02402. doi: 10.1029/2003WR002356.
Liu, J., Sheng, J., Polak, A., Elsworth, D., Yasuhara, H., and Geader, A.
2006. A fully-coupled hydrological-mechanical-chemical model for
fracture sealing and preferential opening. Int. J. Rock Mech. Min. Sci.
43 (1): 2336. doi: 10.1016/j.ijrmms.2005.04.012.
Montemagno, C.D., and Pyrak-Nolte, L.J. 1999. Fracture Network
Geometry versus Single Fractures: Measurement of Fracture Network
Geometry with X-ray Tomography. Phys. Chem. Earth (A) 24 (7):
575579.
Mourzenko, V.V., Thovert, J.-F., and Adler, P.M. 1995. Permeability of a
Single Fracture; Validity of the Reynolds Equation. J. Phys. II France
5 (3): 465482. doi: 10.1051/jp2:1995133.
690

Murphy, J.R. and Thomson, N.R. 1993. Two-Phase Flow in a Variable Aperture Fracture. Water Resour. Res. 29 (10): 34533476. doi:
10.1029/93WR01285.
Nemoto, K., Watanabe, N., Hirano, N., and Tsuchiya, N. 2009. Direct
measurement of contact area and stress dependence of anisotropic flow
through rock fracture with heterogeneous aperture distribution. Earth
Planet. Sci. Lett. 281 (12): 8187. doi: 10.1016/j.epsl.2009.02.005.
Oron, A.P. and Berkowitz, B. 1998. Flow in rock fractures: The local cubic
law assumption reexamined. Water Resour. Res. 34 (11): 28112825.
doi: 10.1029/98WR02285.
Polak, A., Elsworth, D., Liu, J., and Grader, A. S. 2004. Spontaneous switching of permeability changes in a limestone fracture with net dissolution.
Water Resour. Res. 40: W03502. doi: 10.1029/2003WR002717.
Polak, A., Grader, A.S., Wallach, R., and Nativ, R. 2003. Chemical diffusion between a fracture and the surrounding matrix: Measurement by
computed tomography and modeling Water Resour. Res. 39 (4): 1106.
doi: 10.1029/2001WR000813.
Pyrak-Nolte, L.J. and Morris, J.P. 2000. Single fractures under normal
stress: The relation between fracture specific stiffness and fluid flow.
Int. J. Rock Mech. Min. Sci. 37 (12): 245262. doi: 10.1016/S13651609(99)00104-5.
Pyrak-Nolte, L.J., Montemagno, C.D., and Nolte, D.D. 1997. Volumetric
imaging of aperture distributions in connected fracture networks. Geophys. Res. Lett. 24 (18): 23432346. doi: 10.1029/97GL02057.
Rowland, J.C., Manga, M., and Rose, T.P. 2008. The influence of poorly
interconnected fault zone flow paths on spring geochemistry. Geofluids
8 (2): 93101. doi: 10.1111/j.1468-8123.2008.00208.x.
Watanabe, N., Hirano, N., and Tsuchiya, N. 2008. Determination of
aperture structure and fluid flow in a rock fracture by high-resolution
numerical modeling on the basis of a flow-through experiment under
confining pressure. Water Resour. Res. 44: W06412. doi: 10.1029/
2006WR005411.
Watanabe, N., Hirano, N., and Tsuchiya, N. 2009. Diversity of channeling
flow in heterogeneous aperture distribution inferred from integrated
experimental numerical analysis on flow through shear fracture in
granite. J. Geophys. Res. 114: B04208. doi: 10.1029/2008JB005959.
Witherspoon, P.A., Wang, J.S.Y., Iwai, K., and Gale, J.E. 1980. Validity
of Cubic Law for Fluid Flow in a Deformable Rock Fracture. Water
Resour. Res. 16 (6): 10161024. doi: 10.1029/WR016i006p01016.
Yeo, I.W., de Freitas, M.H., and Zimmerman, R.W. 1998. Effect of shear
displacement on the aperture and permeability of a rock fracture.
Int. J. Rock Mech. Min. Sci. 35 (8): 10511070. doi: 10.1016/S01489062(98)00165-X.
Zhu, W.C., Liu, J., Elsworth, D., Polak, A., Grader, A., Shen, J.C., and Liu,
J.X. 2007. Tracer transport in a fractured chalk: X-ray CT characterization and digital-image-based (DIB) simulation. Transp. Porous Med.
70 (1): 2542. doi: 10.1007/s11242-006-9080-5.
Zimmerman, R.W. and Bodvarsson, G.S. 1996. Hydraulic conductivity of
rock fractures. Transport in Porous Media 23 (1): 130. doi: 10.1007/
BF00145263.

Noriaki Watanabe is an assistant professor in the Graduate


School of Environmental Studies (GSES) at Tohoku University,
Japan. His research interests are in the areas of hydrogeology and reservoir engineering. He holds a Ph.D. degree from
Tohoku University. Takuya Ishibashi is a Ph.D. student in the
GSES at Tohoku University. His research interests are in the
areas of hydrogeology and reservoir engineering. He holds an
M.S. degree in environmental studies from Tohoku University.
Nobuo Hirano is an associate professor in the GSES at Tohoku
University. His research interests are in the areas of mineralogy and geochemistry. He holds a Ph.D. degree from Tohoku
University. Noriyoshi Tsuchiya is a professor in the GSES, Tohoku
University. His research interests are in the area of geofluid science. He holds a Ph.D. degree from Tohoku University. Yutaka
Ohsaki is a research manager in the Fracture Technology
Group, Development and Engineering Laboratory, JAPEX
(Japan Petroleum Exploration Company Limited) Research
Center. His research interests are in the area of rock mechanics. Tetsuya Tamagawa is a senior research manager in the
Fracture Technology Group, Development and Engineering
Laboratory, JAPEX Research Center. His research interests are
September 2011 SPE Journal

fluid flow in fractures, geomechanical modeling, and estimation of fracture distribution using seismic attributes. He holds
an M.S. degree in mining engineering from Kyushu University,
Japan. Yoshihiro Tsuchiya is a researcher in the Research and
Analysis Group, Technical Department, JOGMEC (Japan Oil,
Gas and Metals National Corporation). His research interests
are in the area of core analyses with X-ray computed tomography. Hiroshi Okabe is a deputy director in the Planning Division,
Business Strategy Department, JOGMEC. He has worked for

September 2011 SPE Journal

Waseda University, Japan, as an IOR/EOR project manager;


for Kumamoto University, Japan, as a visiting professor; and
for Imperial College London. He has worked in reservoir engineering, petrophysical, and geostatistical research projects in
several different countries. His research interests are multiphase
flow through porous media, including pore-scale modeling,
petrophysics (CCAL/SCAL), and reservoir characterization. He
holds a Ph.D. degree from Imperial College, London. He is an
active SPE member.

691

Vous aimerez peut-être aussi