See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/254508572

Validation of a Modified Carman-Kozeny Equation To Model Two-Phase Relative

Permeabilities

Article · October 1999

DOI: 10.2118/56479-MS

CITATIONS READS
31 435

3 authors, including:

Faruk Omer Alpak Larry Lake

Shell International Exploration and Production Inc. University of Texas at Austin
128 PUBLICATIONS   816 CITATIONS    365 PUBLICATIONS   8,194 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Uncertainty quantification View project

Pore-scale Direct Numerical Simulation of Multi-phase Multi-component Flow Via Thermodynamically-Based Computational Fluid Dynamics Techniques View project

All content following this page was uploaded by Larry Lake on 11 April 2014.

The user has requested enhancement of the downloaded file.

SPE 56479

Validation of a Modified Carman-Kozeny Equation

To Model Two-Phase Relative Permeabilities
Faruk O. Alpak, SPE, and Larry W. Lake, SPE, The University of Texas at Austin, and Sonia M. Embid, SPE, Intevep S.A.

Copyright 1999, Society of Petroleum Engineers Inc.

Empirical correlations have been successfully applied in
This paper was prepared for presentation at the 1999 SPE Annual Technical Conference and describing kr’s.1 However, empirical models do not contain
Exhibition held in Houston, Texas, 3–6 October 1999.
explanations of the physical phenomena occurring during
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
multiphase flow2 and most of the empirical models in the
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to literature are for water-wet media even though many
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at hydrocarbon reservoirs are intermediate- or oil-wet.
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
Embid3 modified the original single-phase Carman-Kozeny
for commercial purposes without the written consent of the Society of Petroleum Engineers is equation to account for the two-phase kr of porous media with
prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300
words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment arbitrary wettability. The modified Carman-Kozeny (MCK)
of where and by whom the paper was presented. Write Librarian, SPE, P.O. Box 833836, model should also account for hysteresis in the kr. This work is
Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
a validation of the MCK model developed by Embid3 and a
comparison to a common empirical kr model.
Abstract
This paper presents a validation of an internally consistent, Literature Review
physically based model for relative permeability based on an kr depends on phase saturation, pore structure, wettability, and
extension of the Carman-Kozeny (CK) equation. The modified saturation history.4 A theoretical approach to predict kr should
CK (MCK) expression is a function of surface areas of fluid- account for these relationships. The structural variety of
fluid and fluid-rock interfaces, as well as fluid saturations and permeable media and of multiphase flow, combined with
tortuosity. The model uses interfacial and surface areas varying wetting characteristics of the media, have made
determined from capillary pressure measurements and, by this theoretical modeling difficult. Consequently, neither the effect
means, can incorporate variable wettability and hysteresis as of state variables nor their relationship to fundamental physical
well as assuring consistency of petrophysical properties. properties, such as wettability, interfacial tension, viscosity,
To validate the MCK approach, the model is fit to and rock properties, are quantitatively known.3
experiments where both capillary pressure and relative Panda and Lake5 imposed modifications to the basic
permeability are measured simultaneously during flow. The Carman-Kozeny (CK) equation for single-phase permeability
MCK model is further fit to literature-reported water-oil to investigate the quantitative connection between textural
experimental data. Besides the MCK model, each data set is fit properties and single-phase permeability of consolidated
with a modification of the commonly used Brooks-Corey media. Rapoport and Leas6 proposed a model based on an
(MBC) model to compare the performances of the two. extension of the CK theory to multiphase flow. Their results
The surface areas derived from capillary pressure are for limiting values of the kr to a wetting phase. Blair and
relationship used in the MCK model provide a good Berryman7 used the CK model and Archie’s laws to determine
description of the experimental relative permeabilities a finite percolation threshold to calculate kr’s.
measured under the same conditions. The investigated MCK Embid3 modified the CK equation by assuming each phase
model fits experimental data almost as well as the MBC model. behaves as if it were alone in a pseudorock, represented by the
Furthermore, the MCK model is physically based and appears other phases and the immobile rock. The final equations for
to agree with the wetting characteristics of the investigated kr’s are functions of surface areas, phase saturations, and the
porous media when these are known. residual saturations. The wetting character and hysteresis are
accounted for by rock-phase and phase-phase surface areas.
Introduction
Relative permeability, kr, is a measure of the ease of flow for Mathematical Formulation
two or more immiscible phases flowing through a permeable Modified Carman-Kozeny Model. Equations 1a and 2a
medium. present the MCK model for the kr for water, phase 1. Similarly,
Eqs. 1b and 2b are for the kr for oil, phase 2. Equation 1 is in
2 F.O. ALPAK, L.W.LAKE, S.M. EMBID SPE 56479

terms of surface areas. Equation 2 shows the kr as being  σ   dA12 dA 2−rock 

parameterized with other equations presented in the Appendix. Pc ≡ P2 − P1 =  12   + cosθ  …..…....(4)

 Vbφ   dS 2 dS 2 
k r1 =
3τ (A1−rock + A 2−rock ) 2 ………………………..(1a) θ is the contact angle that forms between the solid surface
τ 1 (A12 + A 1−rock ) 2
S1
and the fluid-fluid interface measured through the most dense
phase (water-phase).

kr 2 = S2
3τ (A1−rock + A 2−rock ) 2 ...……………………..(1b) Equation 4 assumes that the oil and water are completely
immiscible. The equation, further, is valid for any wetting
τ 2 (A 21 + A 2−rock ) 2 state, any starting saturation, and any pore geometry. These
factors set our work apart from previous attempts to calculate
In the above equations, S1 and S2 are the saturations of phases 1 kr from capillary pressure, Pc.6,9,10 When parameterized, Pc is a
and 2, Aj-rock is the contact area between the rock and phase j, function of the phase saturation and six other parameters:
A12 = A21 is the interfacial area between phases 1 and 2, τ is the
single-phase tortuosity of the rock, and τ 1 and τ 2 are the Pc = Pc (S1 ; A, B,α , β , cosθ , K ) ……………………….….(5)
effective tortuosities for phases 1 and 2. Pc, being a static quantity, is independent of the tortuosity
The final equation for the kr of phase 1 is a function of parameters, γ 1 and γ 2, since tortuosity is a flow characteristic.
phase saturation and five other parameters: The contact angle, θ, appears in Eq. 4 because of the force
k r1 = k r1 (S1 ; A / B,α , β , K , γ 1 ) ………………………….(2a) balance at the fluid-fluid-rock contact line.3 The kr’s are
independent of θ. Because the parameters in the kr and Pc
where the notation means that the quantities to the right of the models do not entirely coincide, it is not possible to calculate kr
semicolon are parameters. Similarly, for phase 2 the final entirely from a Pc measurement. However, knowing Pc reduces
equation for the kr is a function of the phase saturation and five the fitting to a single-parameter regression, and the combined
other parameters: fit ensures consistency between the two petrophysical
k r 2 = k r 2 (S 2 ; A / B,α , β , K , γ 2 ) .………….…………….(2b) properties.

The additional parameters in Eqs. 2a and 2b are as follows: Brooks-Corey Model. We compare the performance of the
1. α and β are constants that parameterize the interfacial MCK model to a modification of the Brooks-Corey (MBC)
surface area between phases 1 and 2. See Eq. A-4. model11 for all data sets. The MBC model used is
2. A/B represents the ratio of the maximum area that the n
interface can attain to the total contacted rock surface area. See o S1 − S1r 1
k r1 = k r1   ...……………………………(6a)
Eqs. A-5a and A-5b.  1 − S1r − S 2 r 
3. K parameterizes the rock surface in contact with the
phase. Large values of K correspond to media with strong n
wetting to phase 1, and small values of K correspond to strong o  1 − S1 − S 2 r  2
kr 2 = k r 2   ……………..…………...…(6b)
wetting to phase 2. An intermediate-wet medium is represented  1 − S1r − S 2 r 
by values of K close to 1. See Fig.1 and Eq. A-1. We show in
the Appendix that the parameter K can be regarded as a In the above equations, kor1 and kor2 are end-point kr’s for
reaction equilibrium constant. phases 1 and 2, S1r and S2r are residual saturations for phases 1
We calculate phase tortuosities from a simple percolation and 2, and n1 and n2 are exponents to be fit to the experimental
model.8 data.
γ1
Unlike the MCK model, the MBC model explicitly
τ 1 = τ S1 .…………………………….……………….(3a) contains residual phase saturations and end-points.

γ2 Systematic Methods for Validation

τ 2 = τ S 2 .………………..…..……….…………...….(3b)
We use two approaches to regress the parameters of the MCK
Theoretically, as the saturation of a phase increases, the model. The first approach sequentially fits both kr and Pc data.
tortuosity of that phase should decrease. Equation 3 is This approach is used when both kr and Pc data are available.
consistent with this behavior if γ 1, γ 2 < 0. Figure 2 presents The second approach fits only the kr measurements. Since only
typical behavior of τ 1/τ and τ 2/τ for negative values of γ 1 and a few data sets have simultaneous Pc and kr measurements, the
γ 2 from Eqs. 3a and 3b. number of data sets in the first approach is much smaller than
The MCK model for kr’s does not contain residual phase for the second. The parameters for the two models are
saturations or end-point kr’s. determined by constrained nonlinear regression using a
Ideally, the surface areas used in the MCK model come spreadsheet application.12 The constraints are γ 1 < 0, γ 2 < 0,
from equilibrium thermodynamics as follows: A12 > 0, and Aj-rock > 0.
SPE 56479 VALIDATION OF A MODIFIED CARMAN-KOZENY EQUATION TO MODEL TWO-PHASE RELATIVE PERMEABILITIES 3

Sequential-Fit Method. We used a two-step method for fitting permeabilities with the single-phase permeability of the porous
experimental data sets with both Pc and kr data present. In the medium). The parameter C is basically the ratio of the
first step, the experimental data are fit with the Pc relationship normalizing single-phase permeabilities. For such cases, the
of Eq. 5 to determine the parameters A, B, α, β, Κ and cosθ. number of fit parameters increases from five to six.
k r1 = k r1 (S1 ; C , A / B,α , β , K , γ 1 ) …………………...…..(9a)
The second step fits the experimental kr, using the previously
calculated parameters; A, B, α, β, and K, by regressing on the
tortuosity parameters γ 1 and γ 2 in Eq. 3.
k r 2 = k r 2 (S 2 ; C , A / B,α , β , K , γ 2 ) .………………….….(9b)
Single-Step Fit Methods. We used a single-step fit method The corrections in Eq. 9 are not needed on the MBC model
when only kr measurements are available. The regression fits.
parameters in the MCK model now are A/B, α, β, K, γ 1,and γ 2.
To reduce the number of fit parameters and, in this manner, Results
to minimize the possibility of nonunique fit, for some cases, Experimental Data. Knowing that there would be few data
we related the cosθ to the K in the following manner: sets that contained both kr and Pc data on the same sample, we
cos θ = tanh [log( K )] ……………………………….……(7)
conducted some kr measurements ourselves. These were
steady-state measurements in which the oil phase was decane
Both the contact angle and the parameter K, of course, relate to and the water a 1% NaCl brine. The medium was a sandpack
the wetting state of the medium, so a relationship between that was chemically altered to produce a variable wettability.
them is plausible.12 Equation 7 passes through the following Besides conducting experiments in media of different wetting
limits: state, we both increased and decreased the water saturation to
1. cos θ = +1 when K is large. investigate the effects of hysteresis. See Embid3 for other
2. cosθ = 0 when K = 1. details.
3. cos θ = −1 when K is small. A major novelty of this work is the construction of an
Using Eq. 7 makes it possible to apply the MCK approach experimental apparatus to measure kr and Pc simultaneously.
in cases of neutral wetting. Note from Eq. 4 that the Pc This was accomplished by imbedding four porous rings in the
sandpack holder, equally spaced from inlet to outlet. The rings
relationship does not contain Aj-rock when cosθ = 0.
Besides requiring convergence of the regression algorithm, were chemically treated so that the first and third were always
water-wet and the second and fourth oil-wet. When brought
we verify the goodness of the fits by the following:
into contact with the flowing fluids, therefore, pressure
1. Visual inspection of the agreement between the
transducers connected to rings one and three measure a water-
regressed models and the data.
phase pressure and those connected to rings two and four an
2. Analysis of the regression residuals to insure that they
oil-phase pressure. At steady state, the pressures in both phases
are independent and have a Gaussian distribution.
3. Inspection of the surface area relations to see that they at rings two and three could be determined by linear
interpolation between the others. Since the water saturation in
are behaving as we suppose they should.
the pack is known from material balance, an entire Pc curve
4. Inspection of a j-function type plot as calculated below.
can be constructed by varying the flowing water fractions, as is
Vbφ 1   dA dA2−rock  required for the kr measurement. See Embid3 for other details.
j = Pc =    12 + cosθ
  dS
 ………..(8) Figures 4-7 show data from this apparatus.
σ 12 AT  AT  2 dS 2 
When fitting the experimental kr data with the MBC model, Selected Fits. The results of the sequential-fit method applied
the residual saturation and end-point kr values required by Eqs. to the experimental measurements for an intermediate-wet
6a and 6b are not regressed since these can be estimated sandpack are in Fig. 4 for the Pc data and in Figs. 5 to 7 for the
directly from the first and last data points in the data set. Only kr data.
the parameters n1 and n2 are determined by single-step We note several things from Figs. 4-7:
regression. 1. The Pc model, Eq. 5, does an excellent job of fitting the
experimental data on all hysteresis cycles. This is true for all
Occasional Modification. Several data sets are normalized by wetting states.3,12
the oil-phase permeability at residual water saturation. The 2. The MBC model fits the kr data quite well. This is to be
others are normalized by the air permeability of the porous expected since the MBC model is forced to fit the end-point
medium. When both of the normalizing permeabilities are data.
known, it is easy to adjust all curves to a common normalizing 3. The MCK model fits the kr data acceptably but less well
factor. When the two normalizing permeabilities are not than the MBC model. The region of poorest agreement is in the
known, we introduced an additional fit parameter C to convert vicinity of end-point kr.
experimental kr’s (determined by normalization of phase 4. Even though there are no explicit residual phase
permeabilities with any reference permeability) to modified- saturations in the MCK model, it seems to predict an apparent
experimental kr’s (determined by normalization of phase
4 F.O. ALPAK, L.W.LAKE, S.M. EMBID SPE 56479

residual saturation--one in which the phase permeability is We checked the linear independence between the regressed
very small--fairly well. variables. All of them were either independent of the others or
Figures 8-11 demonstrate the results obtained from single- weakly dependent,12 except for the cosθ –K relationship as
step fit method on a kr data for a selected water-wet medium discussed further below.
from Ref. 13. These figures show some of the plots used for a
quality check on the regressions. Figure 8 shows an agreement Relating to Wettability Index. One useful application of the
between regressed models and data, Fig. 9 of the predicted MCK model would be to relate a measured kr curve to a
interfacial areas, Fig. 10 of the predicted contact areas, and specific wetting state, the state being quantified by a standard
Fig. 11 of the j-function. measure of wettability.
This fit is interesting because the MCK model actually Figure 14 shows the behavior of the Pc relationship, Eq. 4,
seems to fit the small kr range better than the MBC model, for various wetting conditions as determined by the Amott
though again the MBC model does better near the end-point. index (IA).16 IA = 1 indicates a strongly water-wet medium
The inferred interfacial area, Fig. 11, is highly skewed to the whereas IA = –1 indicates a strongly oil-wet medium. Neutral
left indicating, possibly, the presence of a stable water film wettability is represented by IA = 0.
down to small water saturations. The Amott indices are calculated using the following
relationships:
Aggregate Results. We tested the significance of the
variability explained by the regressions of both models at a 5% I A = WI1 − WI 2 .……………...……………….………..(10)
probability level for the appropriate degrees of freedom.14 The
residuals associated with regression parameters are generally where
uncorrelated and Gaussian. Table 1 shows that none of the S 2i
MBC fits failed the 5% significance level while some of the WI1 = .…………...……………….……..…(11a)
S 2i + S 2 d
MCK fits did. Generally, both models fit the data well in the
sense that their regressed values explain most of the variability
in the data, though the MBC did slightly better. S1i
WI 2 = .……………….……..………..……...(11b)
Figure 12 shows a Box plot presenting the range of values S1i + S1d
for the MCK model parameters: A/B, α, β, K, cosθ, γ 1, and γ 2.
Similarly, Fig. 13 is a Box plot for the MBC model The parameters used in the above equations are: WI 1 and WI 2
parameters: S1r, S2r, kor1, kor2, n1, and n2. are wettability indices for water and oil, S1i and S2i represent
Box plots are constructed with a Kaleidagraph15 graphing the volume of water displaced by oil imbibition and the
function. For each parameter the plot shows a vertical volume of oil displaced by water imbibition, and S1d and S2d
rectangle that contains 75% of the values. The horizontal line represent the volume of water displaced by forced
through the middle of each rectangle is the median value and displacement of oil and the volume of oil displaced by forced
the lines extending above and below the rectangle indicate the displacement of water.
range of the fitted data, exclusive of outliers. The occasional The absence of residual phase saturations from the MCK
open circles indicate outliers. model results in the following:
The impression from Fig. 12 is of the enormous range--
nearly 6 factors of 10--encountered by the A/B ratio. This is to S1i + S1d = S 2i + S 2 d = 1.0
be expected from the definitions of these parameters, Eqs. A- and
5a and A-5b, since the B in the ratio A/B contains the total
contacted area, a measure of the single-phase permeability, S1i = 1 − S 2i .
which is known to exhibit enormous natural variability. Other
ranges are much more benign, especially those corresponding Then the Amott index can be easily calculated as follows:
to the tortuosity parameters. I A = 2 S 2i − 1 ……...……………….……………….....(12)
The main variability in the MBC parameters, Fig.13, lies in
the end-point kr values. Again, this is expected since the We calculated Amott indices for three wetting conditions
information on the medium’s wetting state, which is highly using S2i values determined from the Pc curves. Figure 15 is a
variable, is contained in these values for this model. sketch of a generic Pc curve that shows the saturations
Interestingly, the variability in residual water saturation is corresponding to S1i, S1d, S2i, and S2d. S2i and S1d correspond to
small, though that corresponding to the residual oil saturation the saturation range where Pc has positive values whereas S2d
is somewhat larger. The median value of n1 and n2, and to and S1i to the range where Pc has negative values.
some extent the narrowness of the range exhibited, is Figures 16-18 show the behavior of MCK model kr
somewhat of a surprise. n1 = n2 = 2, which is near the median relationships for water-wet, oil-wet, and intermediate-wet
for both oil and water, is what would occur for segregated flow cases. The curves presented are constructed using typical
in a tube. values of the MCK model parameters. The value of the
parameter K is varied to cause the variation in wettability. For
SPE 56479 VALIDATION OF A MODIFIED CARMAN-KOZENY EQUATION TO MODEL TWO-PHASE RELATIVE PERMEABILITIES 5

the parameter cosθ, the correlation between cosθ and K, Eq. 7, the j-function, plotted to ensure consistency of the fits, is
is used. Another note is that if one of the parameters α and β is generally consistent with typical capillary pressure curves.
nonzero, then the other is usually zero. This is the typical result For both categories of data, the trends obtained for the
we obtained from our fits. dimensionless interfacial areas appear to be consistent with
The curves in Fig. 16 exhibit that a water-wet behavior those reported by Reeves and Celia.19 Most importantly, the
consistent with the rationale of the K parameter (see Appendix) parameters derived from the MCK fits are usually consistent
is obtained when β is equal to zero and α has a typical value. with the wetting characteristics of the investigated media,
Similarly, to obtain a strongly oil-wet behavior α should be when those characteristics are known. That is, we show K > 1
equal to zero (Fig 17). If we use typical values for both α and for water-wet media and K < 1 for oil-wet media.
β, we obtain an intermediate-wet behavior for kr’s (Fig. 18). The absence of residual phase saturations from the MCK
These results indicate that there is some sort of a relationship model is one of its more useful features for it suggests a means
of predicting these highly significant quantities. This is the
between α and β and wetting state even though this was not
subject of ongoing work; but one thing is clear now, the
detected when the parameters were checked for linear
equivalency between the residual saturations and the saturation
dependence.
at which Pc = 0 is obviously not true, as a comparison of Figs.
The shifts in the Figs. 16-18 curves are consistent with
4 and 6 reveals. While S2r = 1 – S1|Pc =0 may be true for strongly
what is known about the effects of wettability on kr data: cross-
wetted media, the MCK model suggests that residual phase
over water saturations shift to smaller values as oil wetting
saturations are associated with the Pc curve asymptotes. This
increases and a phase’s end-point kr lowers as that phase
further implies that the well-studied mechanism of snap-off
becomes more wetting. The results suggest that the MCK
may be relevant only for strongly wetted media.11
model may be useful in shifting a measured kr curve to other
We note, finally, that the Pc relation, Eq. 4, is based on
wetting conditions. This type of shift is possible because the
minimizing thermodynamic free energies in a closed system,
MCK has a physical base.
which media supporting flow definitely are not. This
inconsistency means that the MCK predictions derived from Pc
Relating to Contact Angle. Figure 19 presents the cosθ–K
measurements are more likely to be accurate in the limit of
relationship. The cosθ values are regressed for the data sets very slow or high capillary number flow.
that have both Pc and kr data. However, for the other data sets A detailed documentation of the results for both categories
this parameter is calculated using Eq. 7. The regressed data of data is presented in Ref. 12.
seem to be consistent with the predictions of Eq. 7 for the mid-
range saturations. Summary and Conclusions
In an oil-water-rock system, θ = 0° corresponds to strong This work presents a validation of the modified Carman-
wetting for water whereas θ = 180° to strong wetting for oil. Kozeny kr model and thereby determines best-fit results for
Neutral wettability is represented by θ = 90°. parameters of this model. Both stepwise and direct-fit methods
The contact angle, θ, appears in the MCK model Pc are used to fit experimental data.
relationship, Eq. 4, through the parameter cosθ, which is The following conclusions come from this study:
obtained either directly by means of fitting or indirectly by 1. The surface areas derived from the thermodynamics of
calculating from Eq. 7 (which uses a fit parameter, K). Pc, used in the modified Carman-Kozeny Pc model, provide a
We compared our results for θ with the results reported for realistic description of the experimental kr’s measured under
some data sets in the literature in which θ was separately the same conditions.
measured. Our results are usually satisfactory. For example, 2. The parameters derived from the fits are consistent with
Ref. 17 reports a θ of 180° for a data set for which we obtained the wetting state of the investigated porous media.
θ =159°. In another case, from the same reference, a θ of 47° is 3. The modified Carman-Kozeny model fits experimental
reported and we calculated 58°. Similarly, a θ of 87° is reported data almost as well as the more commonly-used Brooks-Corey
in Ref. 18 for a preserved core and our result is 62.5°. These model.
results are encouraging as they indicate that the MCK model
seems to be successful in consistently representing the wetting Nomenclature
state of the investigated media. A, B = MCK model parameters (interfacial and contact
area calculations)
2
Discussion A = area, L
2
For the data sets that have both Pc and kr measurements, the AI = normalizing factor for interfacial area, L
2
MCK model is generally in good agreement with the AT = total surface area, L
experimental data. Both models show roughly the same degree C = conversion coefficient (MCK model parameter)
of agreement with this data. C = concentration
On the other hand, for the data sets that have only kr CK = Carman-Kozeny
measurements, for most of the cases, the MCK model fits the IA = Amott index
experimental data better than the MBC model. The behavior of K = equilibrium constant for reactions on the phase-
6 F.O. ALPAK, L.W.LAKE, S.M. EMBID SPE 56479

rock surfaces (MCK model parameter, contact 4. Van Genuchten, M.T.: “A Closed Equation for Predicting the
area calculation) Hydraulic Conductivity of Unsaturated Soils,” Soil Sci. Soc. Am.
kr = relative permeability, fraction J. (1980) 44, 892-898.
o 5. Panda, M.N. and Lake, L.W.: “A Physical Model of Cementation
kr = end-point relative permeability, fraction and Its Effects on Single-Phase Permeability,” AAPG Bulletin
MBC = Modified Brooks-Corey model (1995) 79, 431-443.
MCK = Modified Carman-Kozeny model 6. Rapoport, L.A. and Leas, W.J.: “Relative Permeability to Liquid
n = MBC model parameter (curvature) in a Liquid-Gas System,” Trans., AIME (1951) 192, 83-98.
2
P = phase pressure, M/(L-t ) 7. Blair, S.C. and Berryman, J.G.: “Estimates of Permeability and
Pc =
2
capillary pressure, M/(L-t ) Relative Permeability for Sandstone Using Image Analysis of
Cross Sections,” Rock Mechanics as a Multidisciplinary Science,
S = saturation, fraction
3 A.A. Balkema, Rotterdam (1991) 365.
Vb = bulk volume, L 8. Bedrikovetsky, P.G. and Bruining, J.: “A Percolation Based
WI = wettability index Upscaling Technique for Viscous Force Dominated Water
Flooding in Uncorrelated Heterogeneous Reservoirs” paper
Greek Symbols presented at the 1995 European Symposium on Improved Oil
Recovery, Vienna, May 15-17.
α, β = interfacial area model parameters 9. Fatt, T. and Dykstra, H.: “Relative Permeability Studies,” Trans.,
γ = tortuosity parameter (relative permeability AIME (1951) 192, 249-256.
equations) 10. Burdine, N.T.: “Relative Permeability Calculations from Pore
θ = contact angle (capillary pressure relationship) Size Distribution Data,” Trans., AIME (1953) 198, 71-78.
σ
2
= interfacial tension, M/t 11. Lake, L.W.: Enhanced Oil Recovery, Prentice Hall, Englewood
Cliffs (1989) 61-62.
τ = tortuosity
12. Alpak, F.O.: “Validation of a Thermodynamically-Based
τ 1, τ 2 = effective tortuosity for phase 1 and 2 Relative Permeability Model,” MS thesis, The U. of Texas,
φ = porosity, fraction Austin, TX, in progress (1999).
13. Morgan, T.J. and Gordon, D.T.: “Influence of Pore Geometry on
Subscripts Water-Oil Relative Permeability,” JPT (October 1970) 1199,
calc. = calculated 470.
I = interfacial 14. Jensen, J., Lake, L.W., Corbett, P.W.M., and Goggin, D.J.:
exp. = experimental Statistics for Petroleum Engineers and Geoscientists, Prentice
Hall, Upper Saddle River (1997) 226-229.
j = phase
15. KaleidaGraph, Abelbeck Software. Reference Guide, 3rd Ed.
j = j-function Synergy Software (1991).
ref = reference 16. Amott, E.: “Observations Relating to Wettability of Porous
rock = rock Rock,” Trans., AIME (1959) 216, 156-162.
1, 2 = water, oil 17. Owens, W.W. and Archer, D.L.: “The Effect of Rock Wettability
1d, 1i = water displaced, water imbibed on Oil-Water Relative Permeability Relationships,” JPT (July
1971) 251, 873-878..
1s, 2s = water-surface, oil-surface
18. Mungan, N.: “Enhanced Oil Recovery Using Water as a Driving
2d, 2i = oil displaced, oil imbibed Fluid; Part 2-Interfacial Phenomena and Oil Recovery:
Wettability,” World Oil (March 1981) 77-83
Acknowledgments 19. Reeves, P.C. and Celia, M.A.: “A Functional Relationship
This work was supported by a grant from the American Between Capillary Pressure, Saturation, and Interfacial Area as
Chemical Society and by the Reservoir Engineering Research Revealed by a Pore-Scale Network Model,” Water Resour. Res.
Program of the Center for Petroleum and Geosystems (1996) 32, No. 8, 2345-2358.
Engineering at The University of Texas at Austin. 20. Morrow, N.R.: “Wettability and Its Effect on Oil Recovery,” JPT
(December 1990) 42, 1476-1484.
Larry W. Lake holds the W. A. Moncrief (Monty) Centennial
Endowed Chair at The University of Texas at Austin. Janaka
Appendix—Surface Areas Used in the MCK Model
Paulis performed experimental work reported on the strongly
water-wet media. We can parameterize the phase-rock surface (contact) areas as
follows:
References
Sj
1. Honarpour, M., Koederitz, L.F., and Harvey, A.H.: “Empirical A j − rock = AT ……………..………..…(A-1)
Equations for Estimating Two-Phase Relative Permeability in S j + K (1 − S j )
Consolidated Rock,” JPT (December 1982) 34, 2905-2908.
2. Honarpour, M., Koederitz, L.F., and Harvey, A.H.: Relative If we take phase 2 as the reference phase, the contact area
Permeability of Petroleum Reservoirs, CRC Press, Inc., Boca equation for phase 1-rock is specifically:
Raton (1986).
3. Embid-Droz, S.M.: “Modeling Capillary Pressure and Relative K (1 − S 2 )
Permeability for Systems with Heterogeneous Wettability,” PhD A1− rock = AT ……..………..……….…(A-2)
K (1 − S 2 ) + S 2
dissertation, The U. of Texas, Austin, TX (1997).
SPE 56479 VALIDATION OF A MODIFIED CARMAN-KOZENY EQUATION TO MODEL TWO-PHASE RELATIVE PERMEABILITIES 7

The constant K is a parameter related to the wettability of TABLE 1— SIGNIFICANCE TEST RESULTS AT 5 PERCENT
the rock surface to the reference phase. AT is the total surface PROBABILITY LEVEL
area and can be represented as follows: TOTAL NUMBER OF
NUMBER DATA SETS
AT = A1− rock + A 2 − rock …...………….……..………...(A-3) DATA SET OF FAILING THE
DATA SIGNIFICANCE
Small values of K represent media with strong wetting for SETS TEST
phase 2 (the rock surface is completely wetted by the reference Capillary pressure (MCK) 14 3
phase for very small S2 values), and large values of K Water relative permeability (MCK) 40 2
correspond to media with weak wettability to phase 2 (a large Oil relative permeability (MCK) 40 4
S2 is required to completely wet the rock surface). See Fig. 1. K Water relative permeability (MBC) 40 0
values that are close to 1 correspond to intermediate-wet Oil relative permeability (MBC) 40 0
media.
Interestingly, Eq. A-2 and the parameter K contained
therein may have an origin in the thermodynamics of reactions.
It has long been established that adsorption of surface-active
chemical from one of the phases plays a role in determining a
medium’s wetting.20 Suppose a chemical reaction of the
following type:
Oil − surface + Water = Water − surface + Oil
where oil-surface indicates an oil-soluble compound that sorbs
on the medium’s surface, and water is a water-soluble surface
agent. (The other compounds are similarly defined.) The
equilibrium constant for this reaction is
[C1s ][C2 ]
K=
[C2 s ][C1 ]
which becomes Eq. A-2 if we replace the water and oil
concentrations by saturations, take the surface compound
concentrations to be equal to surface areas, and use Eq. A-3.
The surface area of the fluid-fluid interface is computed
with Eq. A-4 and its behavior with varying fluid saturation
plotted in Fig. 3.
A12 α β
= S 2 (1 − S 2 ) ...………………..……………..….(A-4)
AI

The idea behind this equation is that the interfacial area

between phases must be always positive and must be equal to
zero when the phase 2 saturation is equal to zero or one (when
only one phase is present). Therefore, the interfacial area
function must pass through at least one maximum at an
intermediate saturation.
The normalization by AI is to allow adjustment in the level
of the curve. The behavior of A12 with varying fluid saturation
is determined by the exponents α and β (see Fig. 3).
The ratio of AI to the total rock surface area, AT, is
represented by the single parameter A/B. The terms A and B are
as follows:
σ 12
A= A ………………………….…...…………..(A-5a)
Vbφ I

σ 12
B= A …………………………...…………..…(A-5b)
Vbφ T
8 F.O. ALPAK, L.W.LAKE, S.M. EMBID SPE 56479

1.0 1.50
0.01
Contact Area / Total Area

0.8 1.00

Capillary Pressure, psi

0.1
0.50 1st Desaturation
0.6
K=1
0.00
0.4 2nd Desaturation
-0.50 Resaturation
10
0.2
-1.00
100
0.0 -1.50
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Sref Water Saturation
Fig. 1—Wetted area between a reference phase and the rock Fig. 4—Experimental and calculated Pc’s for intermediate-wet
as a function of the wettability of a reference phase (from medium (data set from Embid3). Points are experimental; lines are
Embid3). fitted with Eq. 5.

5 1.0
Reference phase
4 relative tortuosity 0.8
Relative Permeability
Relative Tortuosity

Non-reference phase
3 relative tortuosity
0.6 kr1 (exp.)
kr2 (exp.)
2 0.4 kr1 (MBC)
kr2 (MBC)
1 0.2 kr1 (MCK)
kr2 (MCK)
0 0.0
0 0.2 0.4 0.6 0.8 1 0.0 0.2 0.4 0.6 0.8 1.0
Sref Water Saturation
Fig. 2—The behavior of phase tortuosities. Fig. 5—Experimental and calculated kr’s for primarily increasing
water saturation (1 = water, 2 = oil) (intermediate-wet medium,
data set from Embid3).

0.50 1.0
α β
Aij = S j (1− S j )
Interfacial Area / Total Area

0.8
Relative Permeability

0.6 kr1 (exp.)

α =1, β =1
0.25 kr2 (exp.)
0.4 kr1 (MBC)
kr2 (MBC)

α = 5 , β =1 kr1 (MCK)
α =1, β = 5 0.2
kr2 (MCK)

0.00 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Sref Water Saturation

Fig. 3—The behavior of dimensionless interfacial area in a Fig. 6—Experimental and calculated kr’s for decreasing water
two-phase system. saturation (1 = water, 2 = oil), (intermediate-wet medium, data
set from Embid3).
SPE 56479 VALIDATION OF A MODIFIED CARMAN-KOZENY EQUATION TO MODEL TWO-PHASE RELATIVE PERMEABILITIES 9

1.0 1

Contact Area / Total Area

0.8 0.8
Relative Permeability

0.6 kr1 (exp.) 0.6

kr2 (exp.)
0.4 kr1 (MBC) 0.4 Contact area (oil) / Total area
kr2 (MBC) Contact area (water) / Total area
0.2 kr1 (MCK) 0.2
kr2 (MCK)
0.0 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Water Saturation Water Saturation
Fig. 7—Experimental and calculated kr’s for secondarily Fig. 10—Dimensionless contact areas (water-wet medium, data
increasing water saturation (1 = water, 2 = oil), (intermediate-wet set from Morgan and Gordon12). Dotted lines indicate
medium, data set from Embid3). extrapolation. Parameter values in Fig. 9.

1 0.25

Dimensionless Interfacial Area

kr1 (exp.)
kr2 (exp.)
Relative Permeability

0.2
kr1 (MCK)
0.1
kr2 (MCK) 0.15
kr1 (MBC)
kr2 (MBC) 0.1
0.01
0.05

0.001 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Water Saturation Water Saturation

Fig. 8—Experimental and calculated kr’s (1 = water, 2 = oil), Fig. 11—Dimensionless interfacial area (water-wet medium, data
(water-wet medium, data set from Morgan and Gordon12). set from Morgan and Gordon12). Dotted lines indicate
extrapolation. Parameter values in Fig. 9.
5.0
A / B = 0.2213
4.0 α = 2.7361
β =0
3.0 K = 20.1473
cos (θ ) = 0.8628
j

2.0

1.0

0.0
0.0 0.2 0.4 0.6 0.8 1.0
S1
Fig. 9—Calculated j-function behavior (water-wet medium, data
set from Morgan and Gordon12). Dotted lines indicate
extrapolation.
 10 F.O. ALPAK, L.W.LAKE, S.M. EMBID SPE 56479

40 75

30 50 Intermediate-wet
Amott Index (IA) = 0.12
20 25 Strongly water-wet
Amott Index (IA) = 1
Value

10 0

0 -25 Strongly oil-wet

Amott Index (IA) = -1
-10 -50

-20 -75
log(A/B) α β log(K) cos(θ) γ1 γ2
0 0.2 0.4 0.6 0.8 1
Water Saturation
Fig. 12— Box plot for MCK model parameters (1 = water, 2 = oil). The
Fig. 14—Calculated Pc behavior for three wetting conditions
values for log(A/B), log(K), and cos(θ) are multiplied by 10 to be able
(MCK model).
to show all parameters on the same scale.

1 100

0.8 75
Capillary Pressure, psi

50
0.6
Value

25
S 2d = S 1i
0.4
0 zero line
S 1d = S 2i
0.2 -25

0 -50
S1r S2r kr1o kr2o n1 n2 0 0.2 0.4 0.6 0.8 1
S1

Fig. 13—Box plot for MBC model parameters (1 = water, Fig. 15—A typical MCK model Pc curve. Superimposed are
2 = oil). The values for n1 and n2 are divided by 10 to be able used saturations while calculating the Amott index (IA)
to show all parameters on the same scale. (1 = water, 2 =oil).
SPE 56479 VALIDATION OF A MODIFIED CARMAN-KOZENY EQUATION TO MODEL TWO-PHASE RELATIVE PERMEABILITIES 11

1 1
IA = 1 IA = 0.12

0.8 0.8
A / B = 1.5241 A / B = 1.5241
α = 1.945 α = 1.945
0.6 β = 0.0 0.6 β = 2.63
kr2
K = 100 K = 1.5
γ 1 = -1.175 γ 1 = -1.175
γ 2 = -3.6 γ 2 = -3.6
0.4 0.4
kr2
kr1

0.2 0.2

kr1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

Water Saturation Water Saturation

Fig. 16—Calculated kr for water-wet medium (MCK model), Fig. 18—Calculated kr for intermediate-wet medium (MCK
(1 = water, 2 = oil). model), (1 = water, 2 = oil).

1
IA = -1
10000
0.8 Independent
A / B = 1.5241 Correlation (Eq.7)
α = 0.0 100
0.6 β = 2.63
K = 0.001
γ 1 = -1.175
K

1
γ 2 = -3.6
0.4
kr1

0.01
0.2

kr2
0.0001
0
-1.0 -0.5 0.0 0.5 1.0
0 0.2 0.4 0.6 0.8 1 cos θ
Water Saturation
Fig. 17—Calculated kr for oil-wet medium (MCK model), Fig. 19—Relationship between K and cosθ (MCK model).
(1 = water, 2 = oil).

View publication stats