Sie sind auf Seite 1von 4

GEOPHYSICAL RESEARCH LETTERS, VOL. 29, NO. 8, 1237, 10.

1029/2001GL013414, 2002

Influence of coordination number and percolation probability


on rock permeability estimation
U. Fauzi
Earth Physics Laboratory, Department of Physics, Institute of Technology Bandung, Indonesia

A. Hoerdt and F. M. Neubauer


Institute of Geophysics and Meteorology, University of Cologne, Germany
Received 9 May 2001; revised 12 January 2002; accepted 12 January 2002; published 26 April 2002.

[1] Estimation of permeability from image sections can be


conducted by means of Local Porosity Theory (LPT) and the
effective medium approximation (EMA). Using these approaches,
the coordination number that is important for rock modeling is
included. In this paper, we extend the EMA approach by
considering the percolation probability as an additional
parameter. Both equations show that permeability will be higher
as coordination number as well as percolation probability
increases. The permeability increases most rapidly close to the
percolation threshold. Several thin sections of sandstone were
created by digital image analysis. Two Point Correlation Functions
were applied to estimate porosity and specific surface area.
Permeability distribution as an input to EMA was created from the
local permeability. The effective permeability was then calculated
from the EMA equation iteratively. The result shows that
considering coordination number and percolation probability will
improve permeability estimation.
I NDEX TERMS: 5114
Physical Properties of Rocks: Permeability and porosity; 5112
Physical Properties of Rocks: Microstructure; 5194 Physical
Properties of Rocks: Instruments and techniques

shape of the channels, heterogeneity and anisotropy of the samples


[Koplik et al., 1984]. Doyen [1988] used the same method as
Koplik et al. [1984], however, he distinguished the pore and throat
sizes using an erosion operation. He pointed out that the EMA
predictions of electrical and hydraulic transport properties are
accurate to within a factor of 3 over a porosity range of about
5% 22%.
[4] In this paper we use specific surface area rather than pore or
throat size as a parameter to calculate hydraulic permeability. We
extend this method by considering percolation probability, which
has significant influences on transport properties. The influence of
coordination number and percolation probability on permeability
estimation is also studied using LPT.

1. Introduction

where keq: the equivalent transport properties, f(k): the probability


function of transport properties k, and z: the coordination number
(number of throats meeting at a node). For a continuum model, z/2
should be replaced by the dimension of the system [King, 1989;
Sahimi, 1995]. Equation (1) can be extended by adding the
percolation probability (l). The equation then can be divided into
percolates (kp) and non-percolating blocks(kb) [Hilfer, 1991], i.e.:

[2] The effective medium approximation (EMA) is widely


used to describe various physical properties of disordered media
[Berryman, 1995; Koplik, 1981]. Kirkpatrick [1973] proposed an
effective medium approximation to describe conducting properties of heterogeneous media. This approach was compared with
direct numerical calculation for several lattices [Koplik, 1981]
and several probability distribution functions [David et al.,
1990]. The method is excellent for non-critical random linear
networks independent of coordination number, network size and
shape [Koplik, 1981]. It works well for a uniform distribution
even when defined with wide ranges of random variables [David
et al., 1990; King, 1989]. The EMA is relevant, with a
maximum error of 7%, in the worst case of peak-like distributions. The narrower the distribution is, the better the agreement
between EMA and numerical calculation. The EMA is not suited
for describing the conducting properties of heterogeneous media
with decreasing exponential distributions [David et al., 1990]. In
the worst case, the EMA shows an error of as much as 30%
[King, 1989].
[3] The EMA was applied to estimate transport properties of
rocks, i.e.: their electrical and hydraulic conductances, by Koplik
et al. [1984] and Doyen [1988]. Koplik et al. [1984] found that the
ratio of estimated to measured permeability is about an order of
magnitude. This overestimate may be due to non-percolation, and

2. Theoretical Background
[5] The EMA equation derived for random resistor networks by
Kirkpatrick [1973] reads:
Z

keq  k

 f k dk 0
k 2z  1 keq

keq  kp k

 lf k dk
kp k 2z  1 keq
Z
keq  kb k

 1  lf k dk 0

kb k 2z  1 keq

Because kb = 0. Equation (2) leads to:


Z

keq  kp k

 lf k dk
kp k 2z  1 keq

keq

1  lf k dk 0
2  1 keq

z

3
Simplifying with
Z

Copyright 2002 by the American Geophysical Union.


0094-8276/02/2001GL013414

78 - 1

f (k)dk = 1, we get:

1
2zl1
kp k
z
1
2
B

 C
f k @
Adk 0
kp k 2z  1 keq
0

keq 

78 - 2

FAUZI ET AL.: INFLUENCE OF COORDINATION NUMBER AND PERCOLATION PROBABILITY


The local specific surface area (S) then may be estimated from the
image section from the slope of the TPCF at its origin [Berryman
and Blair, 1987] or by polynomial fitting. The discretized version
of equation (7) reads:
T m; n

1
N

fij fim; jn

10

0 if i; j in solid
1 if i; j in pore space

11

1ii max
1 j j max

where:

fi j

with imax = L*  m, jmax = L*  n, N = imax jmax, for 0  m,n 


(L*  1). The isotropic average of the TPCF is then expressed as:
T V
Figure 1.

Coordination number versus porosity.

2.1. Several Limiting Cases


[6] If all cells percolate, i.e. l = 1, equation (4) will tend to the
equation (1). For l = 1 and z = 2 this equation will give the lower
bound (equation 5)
Z
5
keq f k kp k dk
and it gives the upper bound for l = 1 and z = 1 (6).
keq R

1
f k dk
kp k

If l = 2/z it reaches the percolation threshold, since keq will be


zero. If f(k) is set to 1 then this equation reduces to the simple
Kozeny-equation.
2.2. Local Permeability
[7] The local permeability distribution ( f(k)) and the local
permeability (kp) as input parameters for equation (4) are created
with the help of local porosity theory. An image section is divided
into several cells having cell size L*. This cell size L* is obtained
from the extremum of the information functions which are a
function of the local porosity distribution [Hilfer, 1992]. An
intermediate length scale is also expected to exist for the local
specific surface area. The characteristic length may also be
estimated from the first minimum of the two-point correlation
functions.
[8] The local permeability is calculated from the Kozeny
equation, where the local porosity and the local specific surface
area are estimated from the two-point correlation functions (TPCF)
[Debye et al., 1957; Berryman and Blair, 1987]. The TPCF is
defined as [Berryman and Blair, 1987]:
Z
1
T r
f xf x rd 3 r
7
V

2V
1 X
T V; q
2V 1 l0

12



pl
pl
; V sin 4V
where T V; q
T V cos 4V
is the bilinear interpolation
given in Press et al. [1986]. We used z as the integer value of the
(L*/2). The local hydraulic conductance is then expressed as:
hk

f3k L*2
th f Sk2 L

13

where fk: local porosity of the k-th cell, tk: local tortuosity of the
k-th cell., Sk: local specific surface area of the k-th cell, f: shape
factor, L*: cell size, and L: length of the cell.
2.3. Coordination Number
[9] The coordination number (z) is defined as the number of
branches meeting at a node. This parameter is not directly
measurable from a single planar cross section, since it needs a
3-dimensional model. Doyen [1988], however, defined a pseudo
two-dimensional coordination number as the total number of
throats divided by the total number of pore elements in a thin
section. For the largest porosity, however, he assumed that z = 6.
Yuan [1981] pointed out that 3-dimensional packing has an
average value of coordination number of 6. An empirical correlation between coordination number and porosity can expressed as
[Dullien, 1992]:
1  f 1:072  0:11937z 0:0043z 2

14

This equation shows that coordination number increases as


porosity increases as also shown by Doyen [1988] for pseudocoordination number. Figure 1 shows this tendency.


f x

0 if x in solid
1 if x in pore space

T 0 0 

S
4

Figure 2. Digital image processing equipments for rock permeability estimation.

FAUZI ET AL.: INFLUENCE OF COORDINATION NUMBER AND PERCOLATION PROBABILITY

Figure 3.

78 - 3

Rock image and local permeability distribution.

[10] It is shown that coordination number increases as porosity


becomes larger.
2.4. Percolation Probability
[11] Percolation probability indicates whether a cell percolates
or blocks. It equals 1 for percolating cells and 0 otherwise. The
average percolation probability is an important parameter, which is
not easy to measure. A three-dimensional model is needed to
estimate this percolation probability. Percolation probability for a
sample may be anisotropic (Hilfer, pers. comm.). Hilfer [1991]
studied several special models to estimate percolation probability.
The percolation probability is constant for a simple uniform
capillary model. It depends on the percolation threshold for a
consolidation model and on porosity for centered pore model. As
an adjustable parameter the percolation probability may be
expressed as a function of porosity [Hansen et al., 1993; Haslund
et al., 1994]:

Figure 5. Permeability as a function of percolation probability.


The measured permeability is taken from Fauzi [1997].

found a critical porosity. They considered this critical porosity to


modify the Kozeny-Carman relation.

3. Experimental Methods
l afb

15

where a and b are positive constant parameters, and f is porosity. It


is clear that percolation probability increases as porosity goes up.
Based on the permeability-porosity data, Mavko and Nur [1997]

[12] Digital image processing is used to study the influence of


coordination number and percolation probability for permeability
estimation. Figure 2 shows the procedure of estimating permeability using the above approaches.
[13] First, a thin section was cut and placed on an optical
microscope. The scanning procedure was done using a digital
camera mounted on an optical microscope that is connected
directly to a PC. The size of each pixel is approximately 10 mm.
A binary image is created after selecting a threshold. After an
appropriate threshold range is chosen, each pixel is set to one
(pore) if it is within the range or to zero (matrix) if it is outside the
range. The LPT and EMA are then applied to estimate permeability
as a function of coordination number and percolation probability.
[14] We applied the TPCF to calculate porosity and specific
surface area. The parameters obtained are then used as input
parameters to LPT. The LPT formula applied to estimate permeability reads as follows [Hilfer, 1992]:


 S
k f;

Figure 4. Permeability as a function of coordination number. The


measured permeability is taken from Fauzi [1997].

3

 
t
f
 


 S l f; S  pc
f th S2 l f;

16

where k is permeability in, f is a shape factor which depends on the


cross-sectional shape of the capillaries, th is hydraulic tortuosity, S
 is the average local
is the average local specific surface area, f
porosity, pc is percolation threshold, l is the percolation
probability, and t = 1 in the effective medium approximation
[Hilfer, 1992].
[15] The shape factor varies between 2 (for circular shape) and 3
(for an infinite rectangular disk) [Georgi and Menger, 1994].
Hydraulic tortuosity th is defined as the square of the ratio of
the effective average path length to the shortest distance along the
direction of flow [Dullien, 1992]. The percolation threshold
depends on the choice of Bravais lattice. In the effective medium
approximation pc = 2/z [Hilfer, 1992; Sahimi, 1995], where z is the
coordination number of the lattice of measurement cells.

78 - 4

FAUZI ET AL.: INFLUENCE OF COORDINATION NUMBER AND PERCOLATION PROBABILITY

[16] Estimation of permeability as a function coordination


number and percolation probability by means of EMA is conducted
as follows. The binary image was divided into several cells with
the size L* obtained from local porosity theory or two point
correlation functions [Fauzi et al., 1996]. Figure 3 show the binary
images and local permeability distribution for sample used in this
study.
[17] The black color refers to pores and white indicates matrix.
The corresponding permeability distribution shows that the high
porosity areas are related to high permeability. The estimated
permeability as a function of coordination number is shown in
Figure 4.
[18] Permeability goes up as coordination number rises. The
increase of permeability as a function of coordination number
calculated from EMA is slower than using LPT. The calculated
permeability is close to the measured air permeability for very low
coordination number. It does not mean that the low coordination
number necessarily belongs to the above sample, since the measured permeability is due to three-dimensional properties and the
image considers only two-dimensional case.
[19] Influence of the percolation probability on permeability
estimation is shown in Figure 5:
[20] Percolation probability significantly influences permeability calculation using LPT as well as with EMA. Permeability will
be larger as percolation probability increases. The permeability
calculated by the LPT approach gives a rapid increase for low
percolation probability and increases more slowly at higher percolation probability.

4. Conclusion and Discussion


[21] The influences of coordination number and percolation
probability can be studied by means of LPT and EMA with the
help of image analysis. The results show the permeability increases
rapidly at small coordination number as well as at low percolation
probability. We can see clearly that the influences of coordination
number and percolation probability are not linear. It may represent
the complexity of transport mechanism.
[22] Three-dimensional rock modeling and reconstruction are
very useful in determining coordination number and percolation
probability. In turn they will improve the permeability estimation
of real rocks.
[23] Acknowledgments. This work was partly supported by URGE
project and RUT VI from the Republic of Indonesia, and also IGM
University of Cologne-Germany. We are grateful to reviewers and Prof.
K. Vozoff for their constructive comments and suggestions.

Boger, F., J. Feder, T. Jssang, and R. Hilfer, Microstructural sensitivity of


local porosity distributions, Physica A, 187, 55 70, 1992.
David, C., Y. Gueguen, and G. Pampoukis, Effective medium theory and
network theory applied to the transport properties of rock, J. Geoph. Res.,
95, 6993 7005, 1990.
Debye, P., H. R. Anderson, and H. Brumberger, Scattering by an inhomogeneous solid II: The correlation function and its application, J. Appl.
Phys., 28, 679 683, 1957.
Doyen, P. M., Permeability, conductivity and pore geometry of sandstone,
J. Geoph. Res., 93, 7729 7740, 1988.
Dullien, F. A., Porous media: Fluid transport and pore structure, Academic,
San Diego, New York, 1992.
Fauzi, U., A. Hoerdt, F. M. Neubauer, and K. Vozoff, Permeability estimation of rocks using local porosity theory: 58th annual meeting, European
Association fo Geoscientists & Engineers, Amsterdam, ext. abstract vol.,
1996.
Fauzi, U., Untersuchungen zur Charakterisierung der Porengeometrie
von Gesteinen zur Abschatzung der hydraulischen Permeabilitat and
des Formationsfaktors mit Hilfe von Bildanalysen, Dissertation, Institut fur Geophysik und Meteorologie, Universitat zu Koln, Germany,
1997.
Fauzi, U., Influence of coordination number on permeability estimation,
proceedings of HAGI annual meeting, Surabaya, Indonesia, 1999.
Georgi, D. T., and S. T. Menger, Reservoir quality, porosity and permeability relationship, in Stratigraphie und Lithologie in der geophysikalischen Exploration, edited by L. Dresen, J. Fertig, H. Ruter, and W.
Budach, 12th Mintrop Seminar, Munster, 1994.
Hansen, B. D., E. Haslund, R. Hilfer, and B. Nst, Dielectric dispersion
measurements of salt water saturated porous glass compared with local
porosity theory, Material Research Society Proceedings, 290, 185 190,
1993.
Haslund, E., B. D. Hansen, R. Hilfer, and B. Nst, Measurement of local
porosities and dielectric dispersion for a water saturated medium, J. App.
Phys., 76, 5473 5480, 1994.
Hilfer, R., Geometric and dielectric characterization of porous media, Phys.
Rev. B, 44, 60 75, 1991.
Hilfer, R., Local porosity theory for flow in porous media, Phys. Rev. B.,
45, 7115 7121, 1992.
King, P. R., The Use of Renormalization for calculating effective permeability, Transport in Porous Media, 4, 37 58, 1989.
Kirkpatrick, S., Percolation and Conduction, Rev. in Modern Phys., 45,
574 588, 1973.
Koplik, J., C. Lin, and M. Vermette, Conductivity and permeability from
microgeometry, J. Appl. Phys., 56, 3127 3131, 1984.
Koplik, J., On the effective medium theory of random linear networks,
J. Phys. C., 14, 4821 4837, 1981.
Mavko, G., and A. Nur, The effect of percolation threshold in KozenyCarman relation, Geophysics, 62, 1480 1482, 1997.
Press, W. H., B. P. Flannery, S. A. Teukolsky, and W. T. Vetterling, Numerical Recipes, Cambridge University Press, 1986.
Sahimi, M., Flow and transport in porous media and fractured rock, VCH
Verlagsgesellschaft mbH., Germany, 1995.
Yuan, H. H., Influence of pore coordination on petrophysical parameters:
56th Annual Fall Technical Conf. and Exh., Soc. Pet. Eng., paper 10074,
1981.

References
Berryman, J. G., and S. C. Blair, Kozeny-Carman relations and image
processing methods for estimating Darcys constant, J. Appl. Phys., 62,
2221 2228, 1987.
Berryman, J. G., Mixture theories for rock properties, in Handbook of physical constants, edited by A. J. Ahrens, pp. 205 228, AGU, New York, 1995.



U. Fauzi, Earth Physics Laboratory, Department of Physics, Institute of


Technology Bandung, Indonesia.
A. Hoerdt and F. M. Neubauer, Institute of Geophysics and Meteorology,
University of Cologne, Germany.

Das könnte Ihnen auch gefallen