Getting A Handle On Advanced Cubic Equations of State: Measurement & Control
Getting A Handle On Advanced Cubic Equations of State: Measurement & Control
Getting A Handle On Advanced Cubic Equations of State: Measurement & Control
ngineers must often perform complex phaseequilibria calculations to model systems typically found in the refining and chemical industries. Cubic equations of state (CEOS) are currently the equations of state considered most applicable
for such calculations. This article focuses on the enhancements made to the CEOS that are considered industry-wide standards and points out the strengths and
limitations of these CEOS and their mixing rules.
The Soave modification of the Redlich-Kwong (SRK)
equation (1) has been very successful in extending the applicability of CEOS to systems that contain non-polar and
slightly polar components. However, Soave-type CEOS that
use the van der Waals mixing rules have difficulty handling
the phase behavior of polar systems. Numerous efforts have
been made to develop mixing rules and EOS models for
associating fluids. The modern approach of combining
CEOS with excess Helmholtz or Gibbs energy models is
one example. With so many EOS and mixing rules available, it is no wonder that chemical engineers have difficulty
selecting an appropriate thermodynamic method for the
simulation of their flowsheets. This article addresses questions frequently asked at the inception of flowsheet simulation, namely, Which equation of state do I use? and,
What mixing rules should be selected for my system?
Twu-Sim-Tassone CEOS
A general two-parameter CEOS can be represented by the
equation:
RT
a
P =
(1)
vb
(v + ub)(v + wb)
There are two parameters, a and b. Generally, there are two
ways to evaluate them. One is to fit the parameters to experimental data, usually the vapor pressure and liquid density.
The other is to derive the parameters from the critical con-
58
www.cepmagazine.org
November 2002
CEP
straints by setting the first and second derivatives of the pressure, with respect to volume, equal to zero at the critical point.
Since cubic equations of state do not represent PVT behavior
well, particularly near the critical region, the use of a CEOS
to predict liquid densities results in large errors. This is one of
the inherent limitations of any CEOS. Using regression to
force-fit the EOS to liquid density by regression not only fails
to satisfy the critical constraints, but also sacrifices the ability
of the EOS to predict K-values, and also leads to an over estimation of critical temperature and critical pressure.
However, the critical constraints induce a constant value
of the critical compressibility factor, Zc, for all components.
The Zc for real fluids is known from experiment to be generally smaller than that predicted from the CEOS. Thus, predicted liquid densities differ considerably from their experimental values. SRK, Peng-Robinson (PR), and TST equations of state (2, 3) predict a constant Zc, equal to 0.3333,
0.3074 and 0.2962 for all substances, respectively. The value
of Zc obtained from TST is closest to the real value of Zc for
most substances.
The constants u and w are EOS-dependent. Twu et. al. (2,
3) propose a methodology to find a CEOS that predicts liquid
densities for heavy hydrocarbons and polar components better than the SRK or PR equations. For the TST equation, u =
0.5 and w = 3.0. For the SRK equation, u = 0, w = 1. For
the PR equation, u = 0.4142 and w = 2.4141. SRK is the
equation of choice for predicting the liquid density for
methane and PR is best for n-pentane up to n-heptane; and
TST is superior for n-octane and higher carbon numbers, as
well as for polar components. Another special feature of the
CEOS is that while it typically under-predicts liquid density,
this shortcoming may be corrected by volume translation
without affecting the K-value calculation (4).
Although the ability to accurately predict liquid-density
varies from one CEOS to the next, no single equation has an
advantage over another in terms of its modeling accuracy as
Nomenclature
long as the same alpha function (for pure-component vapor
pressure correlation) and the same mixing rules (for mixtureproperty correlation) are used (5). Since the ability of any
CEOS to predict phase equilibria of mixtures depends on the
alpha function and the mixing rule, these two elements will be
the focus of the remainder of this article.
RT
a T 0.5
vb
v( v + b )
(2)
(3)
(4)
(5)
The m parameter is obtained by forcing the equation to reproduce vapor pressures for light hydrocarbons corresponding to
C1 through C10 at Tr = 0.7, and is correlated as a function of :
m = 0.480 + 1.57 0.1762
(6)
a, b
= CEOS parameters
a*, b* = reduced parameters of a and b
A
= Helmholtz energy
c1, c2, c3 = constants in (T) function (Eq. 14, 15)
C1
= constant at infinite pressure
Cr
= constant at zero pressure
Cv0
= function at zero pressure
G
= Gibbs energy
k ,k
= binary interaction parameter
ij ij
l
= binary interaction parameter
ij
m
= parameter in Soaves equation for (T), Eq. 5
n
= parameter in Soaves expanded equation for (T), Eq. 7
L, M, N = parameters in the Twu function
P
= pressure
R
= ideal gas constant
r
= constant reduced liquid volume at zero-pressure
T
= temperature
u, w
= CEOS constants
v
= molar volume
v0*
= reduced liquid volume at zero pressure
V
= total volume
xi
= mole fraction of component i in liquid phase
Z
= compressibility facto
Greek letters
=
c
=
i, j ,k
=
ij, ji
=
ik, ki
=
r
=
vdw
=
zero pressure
infinite pressure
critical property
property of component i, j, k
interaction property between components i and j
interaction property between components i and k
reduced property
van der Waals
Superscripts
*
= reduced property
E
= excess property
(0) and (1) = denotes value of when = 0 and = 1
Any
(7)
where m and n are obtained by correlating vapor-pressure values generated by the Lee-Kesler EOS:
m = 0.484 + 1.515 0.442
n = 2.756m 0.700
CEP
November 2002
(8)
(9)
www.cepmagazine.org
59
Tr 1
Parameter
Unfortunately, the accuracy of predicting vapor pressure from Soaves new (T)
was no better than the original one, and
consequently, was never accepted for application. Numerous investigators have
tried to improve Soaves alpha function
by either altering the parameters or
adding extra terms. But as long as the
same or a similar form of the alpha function is used, and the same approach as
Soave is applied (i.e., plotting 0.5 vs.
Tr0.5), the functions inherent weaknesswill not be overcome.
(0)
= Tr
(1) = Tr
(0)
(1)
(M
(M
(0)
(1)
1)
1)
(10)
(0)
(1)
(1 T
r
(1 T
r
(0)
N
(1)
L
M
N
L
M
N
November 2002
0.0319134
1.28756
8.000000
Tr > 1
(1)
0.625701
0.792014
2.46022
(0)
(1)
0.373949
4.73020
0.200000
0.0239035
1.24615
8.000000
The L, M and N databank of the generalized alpha function for Eqs. 11 and 12 with the
Peng-Robinson EOS for subcritical and supercritical conditions.
Table 3.
Tr 1
Parameter
L
M
N
(0)
0.196545
0.906437
1.26251
Tr > 1
(1)
0.704001
0.790407
2.13086
(0)
(1)
0.358826
4.23478
0.200000
0.0206444
1.22942
8.000000
The L, M and N databank of the generalized alpha function for Eqs. 11 and 12 with the
Twu-Sim-Tassone EOS for subcritical and supercritical conditions.
(12)
(13)
CEP
(0)
0.272838
0.924779
1.19764
www.cepmagazine.org
0.379919
5.67342
0.200000
Tr 1
Parameter
The superscripts (0) and (1) in Eqs. 10, 11 and 12 are consistent with the definition of the acentric factor at = 0 and
=1, respectively. In other words, these two alpha functions are
forced to pass through the saturated vapor pressure at Tr = 0.7
for = 0 and = 1, respectively.
Twu et. al. used the latest data from the DIPPR (12) databank to generate the values of (0) and (1) for SRK, PR and
TST cubic equations of state. Tables 13 show the L, M, and N
values used with Eqs. 11 and 12 for SRK, PR and TST CEOS,
respectively. The advantage of having a linear acentric factor is
that one can reliably extrapolate its value for heavy hydrocarbons. The new generalized (T) function for SRK, PR and
TST, in Eqs. 1012, allows the accurate prediction of vaporpressure data from the triple point to the critical point for light
or heavy hydrocarbons using any of these EOS.
Although the Twu alpha function mentioned above works
very well for non-polar components, it is generally not suitable for polar components. In an attenpt to resolve this difficulty, Soave (13) presents a two-parameter alpha function:
60
(1)
Table 2.
(1)
0.544306
0.802404
3.10835
(0)
The L, M and N databank of the generalized alpha function for Eqs. 11 and 12 with the
Soave-Redlich-Kwong EOS for subcritical and supercritical conditions.
(11)
Tr > 1
(1)
0.544000
1.01309
0.935995
(0)
(0)
(14)
(16)
L ( 1 Tr
NM
(T ) = Tr
e
(17)
Eq. 17 has three parameters, L, M and N, which, unlike those
in Tables 13, are unique for each component and are determined from the regression of pure-component vapor-pressure
data. The authors have set up databanks for L, M and N for a
)
wide range of components, which can be obtained by contacting the authors directly. Twus (T) function has been found
to be more accurate than other (T) functions found in literature for predicting vapor pressures for polar and non-polar
components at temperatures ranging from the triple point to
the critical point. Using the component-dependent parameters, the accuracy of reproducing the vapor pressure from Eq.
17 is within experimental error given in DIPPR.
Twus (T) function has some unique features. Fluids such
as hydrogen exhibit a maximum alpha at low temperatures
(about 19 K). Twus (T) function accurately describes this
behavior and is the only function so far that shows this maximum at low temperature. In addition, it is continuous at the
critical point for all components and extrapolates very well to
supercritical conditions without exhibiting abnormal characteristics, such as increasing with increasing temperature.
Due to the accuracy and wide temperature range of application of the advanced Twu (T) functions (Eqs. 10, 11, 12 and
17) for predicting vapor pressures when used in the SRK, PR
and TST equations of state, these equations have been selected
for use in the SRK, PR and TST models as follows. For
petroleum fractions and components that do not have vaporpressure data available, Eqs. 10, 11 and 12 should be used. For
all components that do have vapor-pressure data available, Eq.
17 can be used.
a = xi x j ai a j 1 kij
i
(18)
1
(19)
b = xi x j bi + b j
2
i j
where kij is the binary interaction parameter that is obtained
from the regression of vapor-liquid equilibrium data. The van
der Waals mixing rules are capable of accurately representing
vapor-liquid equilibria using only one binary-interaction parameter for non-polar or slightly polar systems. However,
many mixtures of interest in the chemical industry contain
strongly polar or associating components. One way to modify
the van der Waals mixing rules so that they can better describe
polar systems is to use the asymmetric kij for the a parameter,
as developed by Twu et. al. (17):
a = xi x j ai a j
i
xi x j ai a j
i
j
( ) (
16
(1 k ) +
ij
k ji kij
1/ 3
(20)
a*
1 G
a* = b * xi i* +
(21)
C1 RT
i bi
The parameters a* and b* in Eq. 21 are defined as:
a* = Pa/R2T2 and b* = Pb/RT
(21a, b)
(22)
1
1+ w
ln
(23)
1+ u
( w u)
where u and w are EOS-dependent constants used to represent
a particular two-parameter CEOS. For example, u = 0.5 and
w = 3.0 for the TST CEOS.
Although Huron and Vidal presented their model more than
two decades ago, their mixing rule is not widely used due to lack
of parameters available in literature for the excess Gibbs energy
at infinite pressure and the inability to accurately describe nonpolar hydrocarbon mixtures and to satisfy the quadratic composition dependence of the second virial coefficient.
C1 =
CEP
61
a*
1 A
a * = b * xi i* +
(24)
C1 RT
i bi
n n
a
xi x j bij ij
RT
j
b= i
(25)
*
a
1 *
b
The Wong-Sandler model introduced a binary interaction parameter, kij, to correct the assumption that the excess
Helmholtz free energy at infinite pressure can be approximated by the excess Gibbs free energy at low pressure:
a
aj
a
1
bij ij = bi i + b j
(26)
1 kij
RT 2
RT
RT
Wong and Sandler demonstrated that parameters in the activity coefficient models correlated at low temperatures can be
used to extrapolate to higher temperatures. Michelsen and
Heidemann (20) suggest that this success is fortuitous.
In an effort to reduce the Wong-Sandler mixing rules to the
conventional van der Waals mixing rules, Orbey and Sandler
(21) modify the GE model of NRTL differently. Consequently,
the mixing rules used in their model cannot employ the NRTL
parameters reported in the DECHEMA Chemistry Data
Series.
(27)
*
RT
bvdw C1 RT
*
*
bvdw
avdw
b* =
(28)
E
E
*
avdw
1 A
A vdw
1 * +
RT
bvdw C1 RT
C1 in Eqs. 27 and 28 is defined in Eq. 23. The asymmetric van
der Waals mixing rule is applied to the avdw parameter, and the
linear mixing rule is used for the bvdw parameter:
avdw = xi x j ai a j 1 kij +
i
xi x j ai a j
i
j
( ) (k
1/ 6
ji
kij
1/ 3
(29)
1
bvdw = xi x j b + b j
i j
2
(30)
Subscript vdw in AE vdw in Eqs. 27 and 28 denotes that the properties are evaluated from the CEOS using the van der Waals mixing
62
www.cepmagazine.org
November 2002
CEP
x j ji G ji
G
j
= xi n
RT
i
x k Gki
n
(31)
ij = Aij/T
(32)
Gij = exp[ijij]
(33)
ij = (ijbi)/2
(34)
Gji = bj/bi
where:
(35)
aj
ai a j
ai
+ 2 kij
(36)
bi
bj
bi b j
Eqs. 34 and 35 are expressed in terms of the CEOS parameters ai and bi and the binary interaction parameter kij. Substituting Eqs. 34 and 35 into Eq. 31, the excess Gibbs free
energy of the van der Waals fluid is obtained. Subsequently, Eqs. 27 and 28 reduce to the classical van der Waals
one-fluid mixing rules (Eqs. 18 and 19). The discussion
above demonstrates that Eq. 31 is more generic than
NRTL. Both the NRTL and van der Waals fluid models are
special cases of the excess Gibbs free energy function expressed in Eq. 31.
The TST infinite-pressure mixing rules are represented by Eqs.
2736 and can be used to model van der Waals fluids and highly
non-ideal mixtures. One can choose the appropriate expressions
for ij and Gij from Eqs. 34 and 35 for binary pairs that are best described by the van der Waals one-fluid mixing rules, while Eqs. 32
and 33 can be used for pairs that behave non-ideally.
C
ij = 1
RT
*
RT
bvdw Cv RT
0
*
*
bvdw
avdw
b* =
(37)
(38)
E
E
*
avdw
bvdw
A vdw
1 A
1 * +
ln
b
RT
bvdw Cv RT
Eqs. 37 and 38 are labeled as Twu(v) mixing rules. The parenthetic v parameter signifies that the mixing rule is volume-de-
bvdw
= xi x j 1 (bi
i
+ b j
( 40)
v * +w
1
ln 0 *
( w u) v0 +u vdw
( 41)
1 a*
u w
2 b *
2
2
a*
a*
u + w * 4 uw + *
( 42)
b
b
Eq. 42 has a root as long as the following condition is met:
a*
(2 + u + w ) + 2 (u + 1)( w + 1)
( 43)
b*
In Eqs. 37 and 38, AE0 and AE0 vdw are the excess Helmholtz
energies at zero pressure. As stated, subscript vdw in AE0 vdw
denotes that the properties are evaluated from the CEOS using
the van der Waals mixing rule (Eq. 3940) for its a and b parameters.
There are numerous useful features present in the Twu(v)
mixing rules, such as their ability to reduce to the van der
Waals mixing rule when AE0 = AE0 vdw. The mixing rules also
satisfy the quadratic composition dependence of the second
virial coefficient boundary condition. But most importantly,
the Twu(v) mixing rules are density dependent in an explicit
form, which allows them to reproduce, almost exactly, the
incorporated GE model (Eq. 31). The mixing rules provide
excellent agreement between experimental data and VLE
predictions over a wide range of temperatures and pressures
using the information contained only in this GE model.
The mixing rule for parameter b defined in Eq. 38 satisfies the quadratic composition dependence of the second
virial coefficient. Alternatively, the conventional linear mixing rule (Eq. 19) could be chosen to solve for b, in which
case the quadratic composition dependence of the second
virial coefficient would be ignored. The mixing rule either with or without the second virial coefficient constraint
has little effect on the phase equilibrium prediction (28).
CEP
63
(44)
r+w
1
ln
( w u) r + u
( 45)
a * = b * vdw
+
ln vdw
RT
b
bvdw Cr RT
b* =
*
*
bvdw
avdw
E
E
a*
b
1 A
A vdw
1 vdw
+
ln vdw
b
C
RT
RT
b
r
vdw
( 46)
( 47)
Eqs. 46 and 47 are called the volume-independent Twu mixing rules, expressed as Twu(r) to differentiate them from the
volume-dependent mixing rules, Twu(v). The parameter (r)
signifies that the Twu(r) model is not a function of liquid density v0*. Rather, it depends on the selected value of r.
Literature Cited
1. Soave, G., Equilibrium Constants from a Modified Redlich-Kwong
Equation of State, Chem. Eng. Sci., 27, pp. 11971203 (1972).
2. Twu, C. H., et. al., An Extension of CEOS/AE Zero-Pressure Mixing
Rule for An Optimum Two-parameter Cubic Equation of State, Ind.
Eng. Chem. Res., 41, pp. 931937 (2002).
3. Twu, C. H., et. al., A Versatile Liquid Activity Model for SRK, PR and
A New Cubic Equation of State TST, Fluid Phase Equilibria, 194197,
pp. 385399 (2002).
4. Peneloux, A., et. al., A Consistent Correction for Redlich-Kwong-Soave
Volumes, Fluid Phase Equilibria, 8, pp. 723 (1982).
5. Twu, C. H., et. al., Comparison of the Peng-Robinson and SoaveRedlich-Kwong Equations of State Using a New Zero-Pressure-Based
Mixing Rule for the Prediction of High Pressure and High Temperature
Phase Equilibria, Ind. Eng. Chem. Res., 37, pp. 15801585 (1998).
6. Redlich, O., and N. S. Kwong, On the Thermodynamics of Solutions.
V: An Equation of State. Fugacities of Gaseous Solutions, Chem. Rev.,
44, pp. 233244 (1949).
7. Wilson, G. M., Vapor-Liquid Equilibria Correlated by Means of a Modified Redlich-Kwong Equation of State, Adv. Cryog. Eng., 9, pp.
168176 (1964).
8. Wilson, G. M., Calculation of Enthalpy Data from a Modified RedlichKwong Equation of State, Adv. Cryog. Eng., 11, pp. 392400 (1966).
9. Soave, G., Improving the Treatment of Heavy Hydrocarbons by the
SRK EOS, Fluid Phase Equilibria, 84, pp. 339342 (1993).
10. Twu, C. H., et. al., A New Generalized Alpha Function for a Cubic
Equation of State: Part 1. Peng-Robinson Equation, Fluid Phase Equilibria, 105, pp. 4959 (1995).
11. Twu, C. H., et. al., A New Generalized Alpha Function for a Cubic
Equation of State: Part 2. Redlich-Kwong Equation, Fluid Phase Equilibria, 105, pp. 6169 (1995).
12.Daubert, T. E., et. al., DIPPR Data Compilation of Pure Compound
Properties, Project 801 Sponsor Release, Design Institute for Physical
Property Data, AIChE, New York (Jan. 2001).
13. Soave, G., Application of a Cubic Equation of State to Vapor-liquid
Equilibria of Systems Containing Polar Compounds, Inst. Chem. Eng.
Symposium Series, No. 56, 1.2/1 (1979).
14. Mathias, P. M., A Versatile Phase Equilibrium Equation of State, Ind.
Eng. Chem., Process Des. Dev., 22, pp. 385391 (1983).
15. Mathias, P. M., and T. W. Copeman, Extension of the Peng-Robinson
Equation of State to Polar Fluids and Fluid Mixtures, Fluid Phase Equilibria, 13, pp. 91108 (1983).
64
www.cepmagazine.org
November 2002
CEP
( 48)
Since v0*vdw depends on the composition and mixture temperature, the assumption of constant v0*vdw sacrifices, to some extent, the quality of the match between the EOS and the GE
model. Nevertheless, this kind of simplification has major
benefits. As previously mentioned, the rigorous zero-pressure
mixing rules require the value of zero-pressure liquid volume
at system temperature. However, high temperatures can result
in low values of a*/b*, thereby violating Eq. 43 and making it
impossible to find a liquid-volume root of the EOS at zero
pressure. The simplification of the zero-pressure mixing rules
obviates the need to calculate liquid volume at zero pressure
and is one way to extend the range of the method.
The simplified Twu(r) mixing rules given by Eqs. 46 and 47
are no longer a function of liquid density v0*. Rather, they depend only on the selected value of r. Therefore, the selection of
an appropriate value of r in the mixing rule is the key factor for
improving the accuracy of VLE calculations. A universal value
of r = 1.18, determined from the information of the incorporated GE model for SRK, has been recommended by Twu et. al.
(29) for use in the phase equilibrium prediction for all systems.
( 49)
avdw = xi x j ai a j 1 kij
i
(50)
Alternatively, the binary interaction parameter, lij, can be introduced to the van der Waals mixing rule for its bvdw:
)(
1
bvdw = xi x j bi + b j 1 lij
(51)
i j
2
CEP
65