Você está na página 1de 5

GEOPHYSICAL RESEARCH LETTERS, VOL. 33, L02318, doi:10.

1029/2005GL025134, 2006

Permeability-porosity relationship: A reexamination of the


Kozeny-Carman equation based on a fractal pore-space
geometry assumption
Antonio Costa1,2
Sezione di Napoli, Istituto Nazionale di Geofisica e Vulcanologia, Naples, Italy
Received 4 November 2005; revised 15 December 2005; accepted 20 December 2005; published 31 January 2006.
[1] The relationship between permeability and porosity is knowledge of size distribution and spatial arrangement of
reviewed and investigated. The classical Kozeny-Carman the pore channels in the porous medium. For instance, two
approach and a fractal pore-space geometry assumption are porous systems can have the same porosities but different
used to derive a new permeability-porosity equation. The permeabilities. One of the most widely accepted and sim-
equation contains only two fitting parameters: a Kozeny plest model for the permeability-porosity relationship is the
coefficient and a fractal exponent. The strongest features Kozeny-Carman (KC) model [Kozeny, 1927; Carman,
of the model are related to its simplicity and its capability 1937] which provides a link between media properties
to describe measured permeability values of different and flow resistance in pore channels. This model represents
non-granular porous media better than other models. an attempt to describe permeability k in terms of porosity f
Citation: Costa, A. (2006), Permeability-porosity relationship: starting from first principles. However because of the
A reexamination of the Kozeny-Carman equation based on a complexity and the large number of related parameters no
fractal pore-space geometry assumption, Geophys. Res. Lett., 33, simple single functions can exist. Because of its vital
L02318, doi:10.1029/2005GL025134. importance in many fields, from industrial manufacturing
processes to the success of hydrocarbon recovery processes,
many semi-empirical corrections have been proposed [e.g.,
1. Introduction Panda and Lake, 1994] to improve the estimation of
[2] Estimation of permeability is of pivotal importance permeability.
for the description of different physical processes, such as [5] In the following, we firstly re-derive the basic KC
hydrocarbon recovery, fluid circulation in geothermal sys- equation then, in the Section 3, we extend it by incorporat-
tems and degassing from vesiculating magmas. Mainly due ing the concept of fractal effective cross-sectional area. The
to the intricate geometry of the connected pore space and to starting point of the KC equation is the comparison between
the complexity of porous media, it has been very difficult to the average fluid velocity U as given by Darcy’s law and
formulate satisfactory theoretical models for permeability. that obtained from the Poiseuille formula for capillary tube.
One of the most largely used methods remains the Kozeny- Darcy’s law was obtained empirically and defines the
Carman approach. permeability k as a proportionality coefficient in the rela-
[3] In this study we briefly review the Kozeny-Carman tionship between U and pressure gradient dp/dx:
model. Then, using the hypothesis of a fractal pore-space
geometry and the empirically based Archie law, we refor- k dp
U ¼ ð1Þ
mulate that model without introducing the concept of h dx
hydraulic radius and obtain a new simple permeability-
porosity equation. As an application, we used the obtained where h is the dynamic viscosity of the fluid. The
equation to predict permeability of fiber mat systems and of relationship (1) is valid at relatively low velocity when
vesicular rocks. inertial terms are negligible and thus laminar flow prevails.
On the other hand, we consider the general Poiseuille
equation in a straight channel with a generic cross-sectional
2. Permeability-Porosity Relationship shape [Mortensen et al., 2005]:
[4] Permeability of porous media is usually expressed as
function of some physical properties of the interconnected A 1 dp
Uchannel ¼  ð2Þ
pore system such as porosity and tortuosity. Although it is a h dx
natural to assume that permeability values depend on
porosity, it is not simple to determine which is the appro- where A is the generic cross-sectional area of the
priate relationship since this would require a detailed micro-channel and a is a dimensionless geometric factor.
For instance, for cylindrical cross-sections a = 8p and
A = p R2 (R is the tube radius), for elliptical cross-
1
Centre for Environmental and Geophysical Flows, Department of sections a = 4p (a/b + b/a) and A = pab (a and b are
Earth Sciences, University of Bristol, UK. the ellipse axes). In general a can be expressed in terms
2
Also at Osservatorio Vesuviano, Instituto Nazionale di Geofisica e
Vulcanologia, Naples, Italy.
of C = P 2/A where P is the perimeter of the boundary
confining the fluid [Mortensen et al., 2005]. Classical
Copyright 2006 by the American Geophysical Union. derivations of the KC model [e.g., Paterson, 1983; Panda
0094-8276/06/2005GL025134$05.00 and Lake, 1994] consider as a starting point the case of a

L02318 1 of 5
L02318 COSTA: PERMEABILITY-POROSITY RELATIONSHIP L02318

circular cross-section in (2), i.e. A/a = R2/8. They extend we assume that a fractal formulation can be adapted to
and adapt formula (2) to the equivalent channel model of describe even non-granular porous media such as volcanic
the porous body considering effects of (i) the effective vesicular rocks.
tortuous path, (ii) the effective volume and (iii) introdu- [7] In our derivation we consider that pore cross-sectional
cing the concept of effective radius R: area A is described by a fractal with dimension DS. In accord
with theory [Dullien, 1979; Bayles et al., 1989; Nigmatullin et
R2 f 1 dp
U ¼ c ð3Þ al., 1992] the sum of total area of pores with areas less than
8 t h dx ‘‘a’’ for a fractally fragmented system is:
where c was introduced as an empirical geometrical  
DS DS ð1DS =2Þ
parameter, f is the porosity, t the tortuosity defined as the A½<a ¼ b 1 a ð8Þ
2 2
square of the ratio between the effective channel length
Le due to the tortuous path and the length L of the porous where b is a positive constant. Assuming that pore area may
body (t  (Le/L)2). Comparison of (3) with Darcy’s law be roughly approximated by the square of its linear
(1) gives the following permeability expression: dimension l, equation (8) becomes:
R2 f  
k¼c ð4Þ DS DS ð2DS Þ
8 t A½<a ¼ b 1 l ð9Þ
2 2
f
The concept of hydraulic radius Rh = av ð1fÞ,
defined as
the ratio of the pore volume to the solid-fluid interfacial Now we define lm and lM as the minimum and maximum
area, must be introduced to replace R, because even in a scale length respectively, which bound the validity of a
natural homogeneous porous medium it is impossible to fractal description, and Alm and AlM as the corresponding
define a representative radius (av is the specific internal areas. Relationship (9) gives us:
surface area, i.e. the ratio of the exposed surface to the  2DS
solid volume). Accounting for this definition, equation (4) AlM ¼ Alm lM =lm ð10Þ
can be written as:
On the other hand, for the pore cross-sectional area AlM we
f3 consider equation (2), corrected for effective volume and
k ¼ Ckc ð5Þ
ð1  fÞ2 tortuosity, to be valid:
 2DS
where Ckc = c/(8a2v t). The ratio F = t/f is known as AlM f 1 dp Al lM f 1 dp
U ¼ ¼ m ð11Þ
formation factor and it is related to the porosity by the a t h dx a lm t h dx
empirically based Archie relationship:
In the fractal frame, porosity can be expressed in terms of lm,
F ¼ b=fm ð6Þ lM and of the dimensions of the fractal embedded in the 3D
space DV: f = 1  (lm/lM)3DV [e.g., Perrier et al., 1996].
where b is an empirical factor which varies in the range Therefore we obtain (lm/lM) = (1  f)1/(3DV), and
0.6 < b < 2 and m is the empirical cementation or
tortuosity factor belonging to the range 1 < m < 4. The  2DS 2D
 S
empirically derived Archie law can be theoretically lM =lm ¼ ð1  fÞ 3DV ð12Þ
obtained assuming that tortuosity behaves as a fractal
[e.g., Nigmatullin et al., 1992]. Considering the relation- Defining q = (2  DS)/(3  DV), we can write:
ship (6) in (5) we have [Bayles et al., 1989]:
AlM f 1 dp Al 1 f 1 dp
f2þm U ¼ ¼ m ð13Þ
k¼C ð7Þ a t h dx a ð1  fÞq t h dx
2
ð1  fÞ

In the next section we propose a generalised model of the Comparing relationship (13) with equation (1), we obtain
KC equation based on the hypothesis, supported by the equation for the permeability-porosity relationship:
observations [Bayles et al., 1989] that pore cross-sectional Alm f 1
areas are fractals. The model does not need to use the k¼ ð14Þ
a t ð1  fÞq
unclear concept of hydraulic radius and appears capable
to reproduce observed permeability-porosity data. Now we consider exclusively the fractal class with DS ffi
DV  1 (additive law [e.g. Sreenivasan, 1991]). This
3. Reexamination of the KC Based on Fractal implies q = (2  DS)/(3  DV) ffi 1 and means that we
Pore-Space Geometry consider random fractals with no axes of symmetry
[Mandelbrot, 1983; Sreenivasan, 1991]. Finally, inserting
[6] Evidence suggests that, in a given length scale range,
in (14) the fractal relationship for tortuosity in the form of
natural porous media, can be described as a fractal of
the Archie equation (6), we have:
dimension D [e.g., Katz and Thompson, 1985; Turcotte,
1986; Bayles et al., 1989]. There is even stronger support
that surfaces of rock grain and of whole rock samples are fm
k ¼ Cc ð15Þ
fractal [Orford and Whalley, 1983; Avnir et al., 1984]. Here ð1  fÞ

2 of 5
L02318 COSTA: PERMEABILITY-POROSITY RELATIONSHIP L02318

Rodriguez et al. [2004]. This was able to describe the well


defined experimental k  f relationships for glass and
natural fiber mats:

fnþ1
k ¼ C* ð17Þ
ð1  fÞn

where n and C* are empirical parameters [Rodriguez et al.,


Figure 1. Fiber mat permeability as a function of porosity 2004]. (We note that (17) can be recovered in the frame of
for (left) sisal and jute fiber and (right) random fiberglass. our derivation assuming m = q + 1). Relationship (17) was
Solid circles represent values measured by Rodriguez et al. compared with (5), and (15) directly on the little scattered
[2004], dotted line values predicted by equation (5), solid fiber mat data by Rodriguez et al. [2004] (we excluded
line by equation (15) and dashed line by equation (17). Best equation (7) because for all the examined cases it showed
parameters are Ckc = 1.09  107 m2 (s2kc = 7.4  1014 a very poor performance and even gave unrealistic values
m4), Cc = 5.32  107 m2, m = 3.48 (sc2 = 3.1  1014 m4), for m). Figure 1 shows fitting results for both natural (sisal
C* = 2.09  107 m2, n = 1.45 (s*2 = 2.1  1014 m4), for and jute) fiber mats and glass fiber mats. Equation (15)
natural fiber, and Ckc = 2.20  108 m2 (s2kc = 1.6  1013 successfully reproduced experimental data, and, for the
m4), Cc = 1.02  107 m2, m = 1.46 (s2c = 1.09  1014 glass fiber case, even gave lower s2 and higher correlation
m4), C* = 1.45  107 m2, n = 0.84 (s*2 = 1.15  1014 than the fully empirical equation (17).
m4), for glass fiber. As weighting factors, wj / 1/Yj(exp) [9] Unlike fiber mat data reported above, many natural
were chosen in both cases. systems commonly show large variability in permeability.
For instance, permeability-porosity relationships of vesicu-
lar volcanic materials present a defined bulk trend but
where Cc = Alm/(b a) is a factor like the Kozeny also show a large scatter [Klug and Cashman, 1996;
coefficient and m the Archie exponent. Saar and Manga, 1999; Melnik and Sparks, 2002; Rust
and Cashman, 2004; Mueller et al., 2005]. In granular
4. Application to Porous Materials materials fluids flow around spheroidal grains so that pores
[8] In this section we compare the performance of dif- have the shape of the space around impermeable spheres. In
ferent parameterisations to fit measured permeability-poros- contrast, in vesicular materials fluids flow through approx-
ity values of two different kinds of non-granular porous imately spherical to ellipsoidal vesicles where the apertures
media: volcanic materials and fiber mats. These exhibit a between vesicles largely determine k [Saar and Manga,
scattered and non-scattered permeability-porosity relation- 1999]. As a consequence, for a given porosity it is possible
ship respectively. Model parameters, such as Cc and m in to get a range of permeabilities depending for instance on
equation (15), were found by minimising the Standard the aperture widths as well as on degassing-related dis-
Deviation of Regression: tortions of bubbles from ellipsoidal shapes [Saar and
Manga, 1999]. Now we consider permeability-porosity data
sets of volcanic material permeabilities reported by Rust
1 X N  2
and Cashman [2004], Melnik and Sparks [2002], Klug and
s2 ¼ wj Yj ðexpÞ  Yj ðcalcÞ ð16Þ
N  P j¼1 Cashman [1996], and Saar and Manga [1999].
[10] Figures 2, 3, 4 and 5, show that even for vesicular
where wj are weighting factors, P is the number of volcanic materials which are not satisfactory described by
parameters, N is the number of data points and Yj(exp) classical models as explained by Saar and Manga [1999],
and Yj(calc) represent the measured and calculated values
respectively. Weighting factors wj are dependent on the
distribution of random errors in the dependent variable.
Permeability variance can change because of the natural
variability of porosity, pore size and sorting [e.g., Panda
and Lake, 1994]. When experimental scatter is observed to
be the same in all parts of the curve, wj = 1 should be used.
When experimental scatter is observed to vary along the
curve, weighting by wj / 1/Yj(exp) (statistical weight) or
wj / 1/Yj2(exp) (proportional weight) should be used.
Generally, the best weighting factor is that which yields a
random residual plot, i.e., [Yj(exp)  Yj(calc)]/s2, with no Figure 2. Permeability of natural eruptive material as a
functional dependencies. Since the classical KC model is function of porosity from data by Rust and Cashman
often unable to describe the measured permeabilities of [2004]. Solid circles represent measured values, solid line
several porous media [see, e.g., Lukasiewicz and Reed, values predicted by equation (15), dashed line by equation
1988; Saar and Manga, 1999; Rodriguez et al., 2004], (17) and dotted line by equation (5). Best parameters are
different semi-empirical parameterisations were presented Ckc = 1.75  1013 m2 (s2kc = 5.50  1014 m4), Cc =
in literature in order to improve prediction. For instance 6.15  1013 m2, m = 2.89 (s2c = 5.42  1022 m4), C* =
the two-parameters equation (7) [e.g., Bayles et al., 1989] 3.14  1013 m2, n = 1.37 (s*2 = 5.42  1022 m4). As
or the fully empirical KC-like equation proposed by weighting factors, wj / 1/Yj(exp) were chosen.

3 of 5
L02318 COSTA: PERMEABILITY-POROSITY RELATIONSHIP L02318

Figure 3. Permeability of natural eruptive material as a Figure 5. Permeability of vesicular basalts as a function of
function of porosity from data by Melnik and Sparks [2002]. porosity from Saar and Manga [1999]. Solid circles
Solid circles represent measured values, solid line values represent measured values, solid line values predicted by
predicted by equation (15), dashed line by equation (17) and equation (15), dashed line by equation (17) and dotted line
dotted line by equation (5). Best parameters are Ckc = 1.0  by equation (5). Best parameters are Ckc = 1.77  1012 m2
1012m2 (s2kc = 2.7  1023 m4), Cc = 2.80  1012 m2, (s2kc = 3.9  1022 m4), Cc = 3.87  1012 m2, m = 1.99
m = 2.76 (s2c = 2.49  1023 m4), C* = 1.54  1012 m2, (s2c = 3.47  1022 m4), C* = 3.68  1012 m2, n = 1.07
n = 1.39 (s*2 = 2.49  1023 m4). As weighting factors, (s*2 = 3.47  1022 m4). As weighting factors, wj = 1 were
wj / 1/Yj(exp) were chosen. chosen.

equation (15) appears to be typically most suitable in all the space geometry assumption and on the Archie law. The
examined data, giving the lowest s2. In particular, percola- equation contains only two parameters and presents a
tion theory predicts that below a certain critical porosity, fcr, capability to describe measured permeability values of
the permeability must be zero, although several experimen- different porous media better than other models. In particu-
tal results show high k-values at f < fcr, in contrast with the lar, the equation was successfully used to predict permeabil-
theory prediction. Saar and Manga [1999] and Mueller et al. ity of different non-granular systems such as fiber mats and
[2005] suggested that high k at f < fcr can be due to fluid vesicular rocks.
flow through pathways and microcracks generated, before
degassing-related bubble collapse, when f was greater than
[12] Acknowledgments. This work was supported by NERC research
fcr. If so, we can understand why a capillary-tube model like grant reference NE/C509958/1. I thank T. M. Gernon for his corrections
(15), is more appropriate to approximate permeability of and K. Cashman and M. O. Saar who furnished the original data used in
vesicular materials. However for these complex systems, it Figures 2 – 4 and 5. The paper highly benefited from the comments of M. O.
Saar and an anonymous reviewer.
represents an attempt to describe permeability in terms of a
k  f relationship only. In principle, if Cc and m (or C* and
n) are different, equation (15), or (17), can reproduce References
different k for the same f but further investigations and Avnir, D., D. Farin, and P. Pfeifer (1984), Molecular fractal surfaces, Nat-
ure, 308, 261 – 263.
more detailed data are needed. Bayles, G., G. Klinzing, and S. Chiang (1989), Fractal mathematics applied
to flow in porous systems, Part. Part. Syst. Charact., 6, 168 – 175.
Carman, P. (1937), Fluid flow through a granular bed, Trans. Inst. Chem.
5. Conclusion Eng., 15, 150 – 167.
Dullien, F. (1979), Porous Media, Fluid Transport and Pore Structure,
[11] Starting from the classical Kozeny-Carman approach, Elsevier, New York.
a new permeability-porosity relationship was theoretically Katz, A., and A. Thompson (1985), Fractal sandstone pores: Implications
derived. The proposed equation is based on a fractal pore- for conductivity and pore formation, Phys. Rev. Lett., 54, 1325 – 1328.
Klug, C., and K. Cashman (1996), Permeability development in vesiculating
magmas: Implications for fragmentation, Bull. Volcanol., 58, 87 – 100.
Kozeny, J. (1927), Uber kapillare Leitung der Wasser in Boden, Sitzungs-
ber. Akad. Wiss. Wien, 136, 271 – 306.
Lukasiewicz, S., and J. Reed (1988), Specific permeability of porous com-
pacts as described by a capillary model, J. Am. Ceram. Soc., 71, 1008 –
10014.
Mandelbrot, B. (1983), The Fractal Geometry of Nature, 3rd ed., W. H.
Freeman, New York.
Melnik, O., and R. Sparks (2002), Dynamics of magma ascent and
lava extrusion at Soufriére Hills Volcano, Montserrat, in The Eruption
of Soufriére Hills Volcano, Montserrat, From 1995 to 1999, edited
by T. Druitt and B. Kokelaar, pp. 153 – 171, Geol. Soc. of London,
London.
Mortensen, N., F. Okkels, and H. Bruus (2005), Reexamination of Hagen-
Figure 4. Permeability of silicic pumices as a function of Poiseuille flow: Shape dependence of the hydraulic resistance in micro-
porosity from Klug and Cashman [1996]. Solid circles channels, Phys. Rev. E, 71, doi:10.1103/PhysRevE.71.057301.
represent measured values, solid line values predicted by Mueller, S., O. Melnik, O. Spieler, B. Scheu, and D. Dingwell (2005),
Permeability and degassing of dome lavas undergoing rapid decompres-
equation (15), dashed line by equation (17) and dotted line sion: An experimental determination, Bull. Volcanol., 67, 526 – 538,
by equation (5). Best parameters are Ckc = 1.66  1013 m2 doi:10.1007/s00445-004-0392-4.
(s2kc = 1.25  1021 m4), Cc = 1.46  1013 m2, m = Nigmatullin, R., L. Dissado, and N. Soutougin (1992), A fractal pore model
for Archie’s law in sedimentary rocks, J. Phys. D Appl. Phys., 25, 32 – 37.
1.47 (s2c = 1.20  1021 m4), C* = 2.42  1013 m2, n = Orford, J., and W. Whalley (1983), The use of the fractal dimension to
0.76 (s*2 = 1.20  1021 m4). As weighting factors, quantify the morphology of irregular-shaped particles, Sedimentology, 30,
wj / 1/Yj(exp) were chosen. 655 – 668.

4 of 5
L02318 COSTA: PERMEABILITY-POROSITY RELATIONSHIP L02318

Panda, M., and W. Lake (1994), Estimation of single-phase permeability Saar, M., and M. Manga (1999), Permeability-porosity relationship in ve-
from parameters of particle-size distribution, AAPG Bull., 78, 1028 – sicular basalts, Geophys. Res. Lett., 26, 111 – 114.
1039. Sreenivasan, K. (1991), Fractals and multifractals in fluid turbulence, Annu.
Paterson, M. (1983), The equivalent channel model for permeability and Rev. Fluid Mech., 23, 539 – 600.
resistivity in fluid-saturated rock—A re-appraisal, Mech. Mater., 2, 345 – Turcotte, D. (1986), Fractals and fragmentation, J. Geophys. Res., 91,
352. 1921 – 1926.
Perrier, E., M. Rieu, G. Sposito, and G. de Marsily (1996), Models of the
water retention curve for soils with a fractal pore size distribution, Water
Resour. Res., 32, 3025 – 3031.
Rodriguez, E., F. Giacomelli, and A. Vazquez (2004), Permeability-porosity
relationship in RTM for different fiberglass and natural reinforcements, 
J. Compos. Mater., 38, 259 – 268. A. Costa, Centre for Environmental and Geophysical Flows, Department
Rust, A., and K. Cashman (2004), Permeability of vesicular silicic magma: of Earth Sciences, University of Bristol, Wills Memorial Building, Queen’s
Inertial and hysteresys effects, Earth Planet. Sci. Lett., 228, 93 – 107. Road, Bristol BS8 1RJ, UK. (a.costa@bris.ac.uk)

5 of 5

Você também pode gostar