Академический Документы
Профессиональный Документы
Культура Документы
a r t i c l e
i n f o
Article history:
Received 2 February 2012
Received in revised form 1 March 2012
Accepted 1 March 2012
Available online 14 March 2012
Keywords:
Viscosity
Natural gas
Equation of state
Expanded Fluid viscosity correlation
Simulation
a b s t r a c t
The Expanded Fluid (EF) viscosity correlation was used to model the viscosity of mixtures commonly
encountered in natural gas processing. This correlation provides viscosity values as a function of uid density and characterizes each pure compound with three uid-specic parameters: c2 , so and c3 , when using
experimental densities (Version 1) and two parameters, c2 and so , when using a cubic equation of state
(Version 2). In particular, the original EF correlation for hydrocarbons was adapted for non-hydrocarbon
components, including: carbon dioxide, hydrogen sulde, nitrogen, helium, water, methanol, ethylene
glycol, diethylene glycol and triethylene glycol. A temperature dependency was introduced for parameter
c2 for components with signicant hydrogen bonding such as water and methanol. For all other components, a xed c2 was found to be adequate. Both versions of the correlation were t to experimental data
for the non-hydrocarbon components with overall average absolute relative deviation (AARD) below 6%.
The viscosities of several sweet and sour natural gas mixtures were predicted with overall AARDs of 6.3%
and 5.1% for Versions 1 and 2, respectively. The viscosities of aqueous solutions both of methanol and of
glycols were also modeled. A binary interaction parameter was required to t the data for all aqueous
mixtures in Version 2 but only for mixtures of water and methanol in Version 1. The overall AARDs of the
calculated viscosities of the aqueous solutions of methanol and glycols by Versions 1 and 2 were 6% and
8.7%, respectively.
2012 Elsevier B.V. All rights reserved.
1. Introduction
The accurate design and operation of natural gas processes
requires the prediction of the phase behavior and transport properties of natural gases of different phases and compositions. Natural
gases include non-hydrocarbon components such as water, hydrogen sulde, carbon dioxide, nitrogen, and helium [1]. In addition,
gas dehydration may involve solvents such as ethylene glycol (EG),
diethylene glycol (DEG) and triethylene glycol (TEG). This study
focuses on the prediction of the viscosity of natural gases and glycols for process and reservoir simulation.
In general, viscosity is required as a function of temperature, pressure, and composition. A suitable viscosity model for
implementation in simulators must: (1) trace the full range of single phase properties in the gas, liquid, critical, and supercritical
regions; (2) be fast; (3) predict both pure component and mixture
viscosities; (4) be continuous across the critical point and consistent with the phase equilibrium model.
Although numerous viscosity models are available in the literature, no single model is applicable consistently to all natural
gas processing applications. The viscosity of the glycolwater
applicability for components like water has not yet been investigated.
Another, recently developed, semi-theoretical model is Friction
Theory (F-theory) [10,11] which relates viscosity to the friction
forces between uid layers arising from internal pressure. A modication of the one-parameter version of this model coupled with a
cubic equation-of-state predicted the viscosity of natural gas mixtures to within 10% of the experimental values [12]. In addition, the
model coupled with a cubic equation-of-state was tested for modeling the viscosity of selected pure alcohols with 7 tuned parameters
[13] for each alcohol. However, the only adaption of the model
for water was based on the WagnerPrauss reference equation of
state [14] with 18 tuned adjustable parameters [15]. To the authors
knowledge, a single consistent application of F-theory has not yet
been demonstrated for all aspects of the natural gas processing.
Another approach is to use viscosity mixing rules to calculate the
viscosity of hydrocarbon mixtures as well as aqueous solutions of
glycols. However, as Poling et al. [16] emphasized, these models are
limited to mixtures in which the components are at reduced temperatures less than 0.7. An example is the GrunbergNissan mixing
rule [17] which is modication of Arrhenius mixing rules with
generalized binary interaction parameters [18]. For glycol/alcohol
mixtures with water, the McAllister mixing rule [19,20] can be used
but has multiple adjustable parameters for ne tuning of the calculations to the experimental values. Teja and Rice [21] developed a
mixing rule model based on corresponding states theory with two
reference uids, and used the pure components as the reference
for each binary mixture. With one adjustable binary interaction
parameter, they tted the viscosity of binary mixtures, including hydrocarbon mixtures and aqueous solutions of alcohols, with
maximum average deviations of 23% [21]. The model was applied
to aqueous solutions of glycols using two adjustable binary interaction parameters with deviations within 8.3% of the measured values
[22].
Recently, Yarranton and Satyro [23] introduced the Expanded
Fluid (EF) correlation for viscosity modeling of hydrocarbons. This
correlation relates viscosity to the uid density. The method was
rst developed with measured density data [23] but was later
extended [24] for use with calculated densities from the Advanced
PengRobinson (APR) equation of state (EoS) [25]. Both versions of
the correlation were successfully applied to model the viscosity of
the pure hydrocarbons including n-alkanes, branched alkanes, aromatics and cyclic over a wide range of pressure and temperature
in gas and liquid phases. The correlation also predicted the viscosity of over 100 mixtures of hydrocarbon components with overall
average deviations less than 8%, in both gas and liquid phases [26].
Although the EF correlation tted the experimental viscosity data of
pure hydrocarbons and predicted the viscosity of hydrocarbon mixtures, its applicability for natural gas processing applications with
non-hydrocarbon constituents has not been studied. Since, the correlation was built empirically based on hydrocarbon viscosity data;
its validity for non-hydrocarbons has yet to be determined.
The objective of this study is to investigate the applicability of
the EF correlation for the mixtures commonly dealt with in the
natural gas processing applications. First, the correlation is tested
for the pure non-hydrocarbon components including carbon dioxide, hydrogen sulde, nitrogen, helium, water, methanol, ethylene
glycol, diethyelene glycol and triethylene glycol. Then, the applicability of the correlation is studied for sweet and sour natural gas
mixtures and aqueous solutions of methanol and glycols. Both Version 1 (based on measured density) and Version 2 (based on EoS
density) are evaluated. Version 2 is intended for simulator applications. However, Version 2 is not well suited for testing model
adaptations for non-hydrocarbons because errors in the EoS densities may skew the interpretation. Therefore, any adaptations to the
model are made with Version 1 and then applied to Version 2.
57
(1)
where and o are the uid and its dilute gas viscosity, c1 and c2
are tting parameters. is the correlating parameter between the
viscosity and the uid expansion as follows:
=
(2)
exp{(s /) 1} 1
Version 1:
Version 2:
n = 0.65 [dimensionless]
n = 0.4872 [dimensionless]
c1 = 0.165 [mPa s]
c1 = 0.4214 [mPa s]
To improve viscosity predictions at elevated pressures, the compressed state density is formulated as a function of pressure:
Version 1 :
s =
so
exp(c3 P)
(3)
Version 2 :
s =
so
1 c4 (1 exp(c3 P))
(4)
[kPa1 ]
(5)
c4 = 0.035 [dimensionless]
(7)
After xing and correlating the parameters, the following uidspecic parameters remain for each component:
Version 1:
Version 2:
c2 , c3 and so
c2 and so
(8)
58
o
s,mix
=
nc
nc
wi wj
i=1 j=1
wi wj
c2,mix
=
o
2
s,mix
nc
nc
i=1 j=1
c3,mix =
1
1
o + o
s,i
s,j
c2,j
c2,i
o + o
s,i
s,j
1
(1 ij )
(1 ij )
nc
1
w
i
i=1
(9)
(10)
(11)
c3,i
(12)
0.3
0.3
)fMW + 0.04fWK
ijo = (0.244 0.45fWK
(13)
where
ijo
fMW = 1
fWK = 1
MWi MWj
(14)
MWi + MWj
WKi WKj
(15)
WKi + WKj
where MWi and WKi are the molecular weight and Watson-K factor
values of component i of mixture, respectively. Either of Eqs. (12)
or (13) were shown to provide mixture viscosity predictions with
an AARD of 8% for Version 2. However, Eq. (12) performed better
for mixtures of components with signicantly different molecular weights, and Eq. (13) provided better results for mixtures of
n-alkanes with aromatic or naphthenic components [26].
The binary interaction parameter at elevated pressures is given
by [26]:
ij = ijo exp((P 101.325))
(16)
xi o,i
i
(17)
x
j j ij
where
ij =
[1 + (o,i /o,j )
0.5
(MWi /MWj )
0.5
0.25 2
(18)
Component
NP
Temperature (K)
Source
Hydrogen sulde
Carbon dioxide
Helium
Nitrogen
Water
Methanol
Ethylene glycol
Diethylene glycol
Triethylene glycol
18
520
376
1490
1795
575
95
35
24
190483
2031100
14.2918.3
631989
2561139
176.1511.1
273428
273428
273428
4400
450,000
83,000
194,000
389,000
200,000
atm.
atm.
atm.
[3036]
[3739]
[37]
[37]
[37]
[37]
[37]
[37]
[37]
where xi and o,i are the mole fraction and dilute gas viscosity of
the component i of the mixture, respectively.
3. Results and discussion
The correlation is rst tested for pure non-hydrocarbon components and a modication for hydrogen bonding uids is introduced.
Then, the validity of the correlation is tested by comparing predicted viscosities of mixtures commonly found in natural gas
processes to experimental values.
3.1. Pure components
There are two categories of pure non-hydrocarbon components encountered in natural gas processing. The rst category is
simple non-hydrocarbons which are usually constituents of produced natural gas and include nitrogen, carbon dioxide, hydrogen
sulde and helium. These components are either non-polar and
interact via London dispersion forces or polar with polarpolar
interactions. The second category is associating non-hydrocarbon
components normally used to treat natural gas streams such as
methanol and glycols. Also included in this category is water which
is a constituent of natural gas but is also an associating species.
The molecules of these components are polar and form hydrogen
bonds with neighboring molecules due to the presence of an O H
group. The application of the EF correlation to each category of
non-hydrocarbons is presented below.
In both cases, the required input properties for the pure components are the dilute gas viscosity and the density at given pressure
and temperature. Table 1 summarizes the viscosity data compiled
for each component. Density data were also compiled from the
same sources but were only available for more limited ranges of
pressure and temperature. Version 1 of the correlation is tested
rst using the datasets which include both measured density and
viscosity. Any conceptual modications to the correlation are introduced at this point. Then, Version 2 of the correlation is tested using
densities calculated with the APR EoS.
3.1.1. Simple non-hydrocarbons
Version 1 of the correlation was tted to data by adjusting the
uid-specic parameters of c2 , c3 , and so to the values given in
Table 2. Table 3 provides a summary of the average absolute deviation (AAD), the maximum absolute deviation (MAD), the average
absolute relative deviation (AARD), the maximum absolute relative
deviation (MARD), and the bias of the tted correlation. The maximum deviation was observed for helium at very low temperatures.
The overall AARD of 4.2% and MARD of 34% are both comparable
to the deviations reported for the application of the correlation
to pure hydrocarbon components. Therefore, it is concluded that
the correlation framework, based on the non-polar hydrocarbon
components data, is sufcient for these simple non-hydrocarbons.
The uid specic parameters, c2 and so , were then determined
for Version 2, Table 2. Note that the general correlations of Eqs.
59
Table 2
Fluid specic parameters to use with Versions 1 and 2 of the correlation for pure non-hydrocarbon components.
Component
Version 1
Version 2
c2
Hydrogen sulde
Carbon dioxide
Helium
Nitrogen
Water
Methanol
Ethylene glycol
Diethylene glycol
Triethylene glycol
b 103 (K1 )
0.188
0.236
0.0517
0.1147
0.1463
0.1463
0.23
0.2487
0.2977
c3 106 (kPa1 )
1194.59
1617.66
300
1012.39
1197
1045
1193.69
1195.04
1214.28
23.1
4.22
99.519
0.6301
so (kg/m3 )
c2
0.0437
0.0644
0.0149
0.0244
0.0674
0.0674
0.1274
0.1147
0.1417
0.187
0.1
0.1
0.3
0.449
0.6
Viscosity (mPa.s)
0.4
0.3
Viscosity (mPa.s)
Version 1
Version 2
260 K
280 K
308 K
0.2
0.1
b 103 (K1 )
35.090
2.672
50000
1092.76
1571.97
286.16
938.41
1368.18
1156
1201.53
1175.86
1189.40
MD-sat. curve
exp-sat. liquid
exp-gas (1 bar)
Version 1
Version 2
0.06
a
0
18.5
10.2
so (kg/m3 )
b
0.006
100
200
300
400
500
600
Temperature (K)
Pressure (kPa)
Fig. 1. EF correlation tted to experimental data of: (a) compressed carbon dioxide, data from [3739]; (b) hydrogen sulde on co-existence curve and 1 bar pressure, data
from [3137].
(5)(7) were used for the pressure dependency parameters. A summary of the deviations for Version 2 of the correlation is provided
in Table 3. Version 2 of the correlation ts the data with the same
accuracy of the Version 1 with overall AARD and MARD of 4.2% and
31%, respectively. However, the accuracy of the Version 2 at elevated pressures is slightly less than Version 1 because the pressure
dependent parameters are correlated in Version 2 but adjustable
in Version 1. The deviation at higher pressures is shown for carbon
dioxide in Fig. 1a. The t of Version 2 can be improved if the pressure dependency parameters (c3 and c4 ) were adjusted as well, but
the generality of the correlation for mixtures would be lost.
Note that both versions of the correlation were tted to limited
experimental data for hydrogen sulde. Much of these data were
calculated at the liquid-vapor coexistence curve using molecular
dynamics (MD) simulation [36]. The calculated values are reportedly in good agreement with experimental data [40]. The only
exceptions were saturated gas data which are clearly inconsistent
with the experimental dilute gas data and were excluded from the
analysis. Fig. 1b shows that the t of both versions of the correlation
to the viscosity data of hydrogen sulde on the co-existence curve.
Table 3
Summary of the deviations for Versions 1 and 2 of the correlation for pure non-hydrocarbon components.
Component
Hydrogen sulde
Carbon dioxide
Helium
Nitrogen
Water
Methanol
Ethylene glycol
Diethylene glycol
Triethylene glycol
Version 1
Version 2
MARD (%)
AARD (%)
Bias (%)
MARD (%)
AARD (%)
Bias (%)
15
9.7
34
24
48
17
12
23
7.1
3.5
3.3
4.1
5.8
5.3
2.1
3.0
5.0
4.5
1.3
2.5
3.1
4.5
2.6
0.2
0.6
0.1
0.3
13
14
31
20
25
30
54
41
41
3.3
5.6
5.0
2.8
1.9
5.4
5.8
8.6
14
2.7
3.3
0.7
0.6
0.2
0.4
0.8
0.0
0.0
60
10
10
256 K
263 K
273 K
288 K
303 K
323 K
343 K
373 K
473 K
673 K
Version 1
Version 2
exp data
Version 2
0.1
Viscosity (mPa.s)
Viscosity (mPa.s)
Version 1
0.1
a
0.01
200
0.01
300
400
500
600
700
200000
Temperature (T)
400000
600000
Pressure (kPa)
Fig. 2. EF correlation tted to experimental data [37] of water: (a) saturated liquid; (b) compressed uid.
c2 = c2 + Kc2 exp(c2 T )
(19)
where c2 and Kc2 and c2 are tting parameters. The default value
for Kc2 and c2 is zero for all components except those with signicant hydrogen bonding such as water and methanol. Therefore, by
denition for all hydrocarbons and simple non-hydrocarbons, the
value of c2 is equal to the value of the non-temperature dependent
c2 .
Version 1 of the correlation was then tted to the experimental data of water and methanol by adjusting the parameters so , Kc2
and c2 with c2 xed at 0.1463. Version 1 t the limited available data of glycols with a xed c2 (Kc2 and c2 set to zero). The
uid specic parameters for the associating non-hydrocarbons and
a summary of the deviations of the tted correlation are given in
Tables 2 and 3, respectively. The AARD are less than 5% while the
MARD of 48% occurs for water. The correlation ts the data of water
on the co-existence curve with a maximum deviation of 14%, Fig. 2a.
Fig. 2b shows that the maximum MARD occurs for compressed liquid water at temperatures below 50 C whereas the water viscosity
decreases under compression. The correlation is not able to model
this behavior.
Having established the new temperature dependent expression
for parameter c2 , Version 2 of the correlation was then tted to
the experimental data of pure associating non-hydrocarbon uids.
Tables 2 and 3 report the uid-specic parameters and model deviations, respectively, for Version 2. As with Version 1, the parameters
so , Kc2 and c2 were adjusted for methanol and water with c2 xed
at 0.0647. Only so and c2 were adjusted to t the glycol data (Kc2
and c2 were set to zero). The overall AARD and MARD for Version
2 are 7.1% and 54%, respectively.
While the highest deviations in Version 1 were observed for
water, they are observed for glycol in Version 2, Table 3. The AARDs
for the glycols are below 14% with Version 2 but below 5% with
Version 1. For instance, for diethylene glycol, the Version 2 model
viscosities deviate at higher temperatures, Fig. 3. The main contribution to the deviations for the glycols is inaccurate density
predictions by APR EoS. Using measured densities as an input to
Version 2 reduces the overall AARD for the glycols from 9.5 to 7.7%;
that is, to nearly the same error as observed for Version 1.
Note, Version 2 of the correlation t the high pressure data of
water at temperatures below 50 C more accurately than Version
1, Fig. 2b. The improvement is attributed to the combined effects
of: (1) the form of the pressure dependency expression in Version 2
which provides a better t for water; and (2) the underestimation of
compressed liquid water density by the APR EoS which accidentally
compensates for the effect of the unusual density trends in water.
3.2. Mixtures
3.2.1. Natural gas mixtures
Table 4 provides the composition of 11 natural gases mixtures
[2,4346] used to evaluation the proposed correlation. Not included
in Table 4 but used in this study are data for sour natural gas mixtures from Elsharkawy [47]. This dataset consists of compositions
and single viscosity data points for 17 different gas samples at a
given temperature and pressure. To broaden the evaluation for sour
mixtures, 8 data points from an MD simulation [46] for mixtures of
methane, carbon dioxide and hydrogen sulde were also used.
Viscosity predictions were made with both versions of the
correlation where possible but Version 1 is limited to mixtures
with experimental density data. The correlation parameters for
the hydrocarbon components of these mixtures were taken from
[23,24] for Versions 1 and 2 of the correlation, respectively. The
exp data
100
Version 1
Version 2
Viscosity (mPa.s)
compression as found with other liquids; however, at lower temperatures its viscosity increases under compression [42].
A modication to the model was required to the model the viscosity of uids with signicant hydrogen bonding, such as water
and methanol. The parameter c2 is the proportionality of the uid
viscosity to its expansion and indirectly represents the effect of the
intermolecular forces on viscosity. Hence, a temperature dependent c2 is appropriate to model the temperature dependency of
hydrogen bonding and its effect on the viscosity. The following formulation of the temperature dependent c2 was determined to t
the data:
10
1
200
300
400
500
Temperature (T)
Fig. 3. The t of the Versions 1 and 2 of the correlation to experimental viscosity
data [37] of dietheylene glycol using xed c2 .
15
15
G1
G5
G2
G6
G3
G7
G4
G8
5
0
-5
-10
40000
G3
G7
G4
G8
5
0
-5
-15
200
-15
20000
G2
G6
-10
a
0
G1
G5
10
Deviation (%)
10
Deviation (%)
61
60000
300
400
500
Temperature (K)
Pressure (kPa)
Fig. 4. Deviations of the predicted viscosity (Version 2) for natural gas mixtures versus: (a) pressure; (b) temperature.
Table 4
Composition of the natural gas mixtures used in this study.
N2
CO2
H2 S
He
n-C1
n-C2
n-C3
C4
C5
C6
C7
C8
C9
C10+
G1 [43]
G2 [44]
5.6
0.66
1.83
84.84
8.4
0.5
94.67
3.5
G3 [2]
G4 [2]
G5 [2]
G6 [2]
G7 [45]
G8 [45]
3.2
1.4
1.4
4.8
0.9
0.55
1.7
0.66
2.19
1.39
1.00
86.33
6.8
2.4
0.91
0.22
0.1
0.04
0.03
71.71
14
8.3
2.67
0.39
0.09
0.01
0.03
80.74
8.7
2.9
1.7
0.13
0.06
0.03
Viscosity (mPa.s)
0.018
0.016
0.016
263 K
294 K
323 K
354 K
241 K
Version 1
Version 2
0.012
0.01
0.008
0.006
0
a
5000
10000
Pressure (kPa)
15000
0.02
92.20
4.34
0.54
0.33
0.07
0.06
0.04
0.00
0.00
0.00
G9 [46]
G10 [46]
G11 [46]
0.6
0.4
0.63
0.27
0.4
0.6
0.1
0.018
0.014
0.01
80.01
9.31
4.96
2.00
0.51
0.18
0.12
0.04
0.01
0.01
91.46
3.1
1.4
1.17
0.28
0.26
0.08
Viscosity (mPa.s)
Component
0.014
0.012
exp data
0.01
Version 1
Version 2
0.008
0.006
0
5000
10000
15000
20000
Pressure (kPa)
Fig. 5. Comparison of the predictions from Versions 1 and 2 of the EF correlation with measured viscosity data of: (a) natural gas mixture G1 [43]; (b) methane at 323 K [37].
62
Table 5
Summary of the deviations of the predictions by Versions 1 and 2 of the correlation for natural gas mixtures.
Version 1
MARD (%)
G1
11
8.1
G2
11
G3
G4
15
G5
14
G6
15
8.6
G7
13
G8
6.3
G9
G10
5.5
G11
6.8
Sour natural gas samples [47]
Version 2
AARD (%)
Bias (%)
MARD (%)
6.5
3.5
8.3
3.2
8.8
7.5
6.3
9.7
6.5
3.5
8.3
0.3
8.8
7.6
6.3
9.7
3.5
3.5
3.9
3.1
5.6
9.1
7.5
11
4.3
4.6
5.1
4.0
14
51
AARD (%)
Bias (%)
1.8
1.8
3.1
4.1
4.4
3.5
1.7
2.1
0.7
1.8
2.8
2.1
4.3
3.1
1.4
2.1
9.6
20
9.6
9.3
0.18
0.16
Viscosity (mPa.s)
Mixture
0.14
0.12
0.1
0.08
0.06
0.04
exp data
0.02
Version 2
0.2
0.4
0.6
0.8
mixing rule (Eq. (A.14) in Appendix A). The tted parameters are
given in Table 6.
Table 7 provides a summary of the AARD, MARD, and the bias
of the predictions of the correlation for these mixtures. Note, ij
was set to zero unless otherwise stated. The AARDs for Version 1
of the correlation are all less than 10% except for mixtures of water
and methanol with an AARD of 21%. The deviations for Version 2
with ij = 0 are considerably higher with an overall AARD of 26%.
The reduced accuracy of Version 2 for mixtures is consistent with
results previously obtained for the hydrocarbon mixtures [26].
Version 2 was then tted to the data using a binary interaction
parameter, Table 7, reducing the overall AARD to 8.7%. Unfortunately, the general ij correlations developed for the hydrocarbons
(Eqs. (12) and (13)) are not applicable for these mixtures of associating species. Both equations give ij values that increase with
increasing molecular weight difference. However, the optimized
ij values for glycols mixtures with water have the opposite trend.
Fig. 7a and b demonstrates the performance of the correlation
for aqueous solutions of diethylene glycol and methanol, respectively. Both versions of the correlation are qualitatively correct.
However, although Version 1 correctly predicts a maximum in the
Table 6
Adjusted volume translation parameters of APR EoS for the aqueous solutions of glycols and methanol.
Parameter
Methanol(1) + water(2)
A12
A21
B12
B21
0.219
0.194
4.16
2.76
0.288
0.240
1.40
1.13
0.348
0.279
0.150
3.04
0.0466
0.051
1.05
1.10
63
Table 7
Summary of deviations of the predicted viscosity by the EF correlation for aqueous solutions of glycols and methanol.
Mixture
Version 1
ij
EG + water
DEG + water
TEG + water
Methanol + water
29
17
10
42
0.067 12
AARD (%)
Bias (%)
MARD (%)
AARD (%)
Bias (%)
ij
MARD (%)
AARD (%)
Bias (%)
9.8
4.6
5.6
21
4.0
6.4
0.4
3.3
21
2.3
62
62
64
52
28
26
25
28
26
25
18
29
0.069
0.066
0.055
0.179
40
24
40
19
11
6.8
13
3.9
3.4
0.7
0.7
0.6
40
2
exp data
1.8
35
Version 1
30
Version 2, ij=-0.066
1.6
Viscosity (mPa.s)
Viscosity (mPa.s)
Version 2 (ij = 0)
MARD (%)
25
20
15
10
1.4
1.2
1
0.8
0.6
exp data
Version 1, ij=0
Version 1, ij=-0.067
Version 2, ij=-0.179
0.4
0.2
0
0
0.5
0.5
Fig. 7. Comparison of the viscosty predictions by Versions 1 and 2 of the EF corrlelation with experimental data of: (a) diethylene glycol + water at 293 K; (b) methanol + water
at 293 K.
4. Conclusion
The previously developed Expanded Fluid viscosity correlation
was extended to natural gas processing applications; in particular,
to include non-hydrocarbon components such as carbon dioxide, hydrogen sulde, nitrogen, helium, water, methanol, ethylene
glycol, diethylene glycol and triethylene glycol. Both Version 1
(measured density input) and Version 2 (EoS based density input)
were evaluated. In both cases, for uids with signicant hydrogen bonding, such as water and methanol, the c2 parameter of the
correlation was modied from a constant to the following:
c2 = c2 + Kc2 exp(c2 T )
where c2 and Kc2 and c2 are tting parameters. The default value
of Kc2 and c2 is zero for all components except those with signicant hydrogen bonding.
The modied correlation was tted to experimental data for
pure non-hydrocarbons with overall average absolute relative deviations (AARD) below 6%. The maximum absolute relative deviation
(MARD) for predictions with Versions 1 and 2 were 48% and 54%,
respectively. For all of the non-hydrocarbon components evaluated
in this study, a xed c2 value (with Kc2 and c2 equal to 0) was sufcient, except for methanol and water. Version 1 of the correlation
is not recommended for compressed water (P > 10 MPa) at temperatures below 50 C due to an unrealistic trend in the predicted
viscosity values.
b
covolume in EoS, m3 /kmol
B
tting parameter in dilute gas viscosity correlation, Eq.
(8), mPa s K1
Bij
adjustable constant in aij , Eq. (A.14), K1
c1
tting parameter in EF correlation, Eq. (1), mPa s
tting parameter in EF correlation, Eq. (1),
c2
c2
tting parameter in temperature dependent c2 , Eq. (19),
a
ac
aij
64
c3
c4
C
D
fc
fMW
fWK
kij
Kc2
MW
n
nc
P
R
s
T
Tr
V
VAPR
VT
wi
WK
xi
(T)
ij
ijo
c2
o
s
so
Acknowledgments
The authors are grateful for nancial support from the sponsors
of the NSERC Industrial Research Chair in Heavy Oil Properties and
Processing, including NSERC, Schlumberger, Shell Canada Energy
Ltd., and Petrobras. We also thank Virtual Materials Group for providing VMGSim process simulation software and Mr. Carl Landra
from Virtual Material Group for his advice on tuning APR EoS for
density predictions of aqueous solutions of associating uids.
where ac is the attractive term at critical point and (T) is an empirical dimensionless scaling function of temperature. This empirical
function is correlated to acentric factor for non-polar or slightly
pure components such as hydrocarbons.
For hydrocarbons, (T) is given by:
i = 1 + fi (1 Tri )
(A.4)
(A.5)
For polar compounds the APR EoS uses the following empirical
function of reduced temperature [52] suggested by Mathias and
Copeman [53]:
(T ) = 1 + Ai (1
Tri ) + Bi (1
Tri ) + Ci (1
Tri )
(A.6)
(1 kij )
ai aj xi xj
xi bi
(A.7)
(A.8)
0.41
0.41 +
(A.9)
(A.10)
V2
RT
P
V
(A.11)
T
si xi + sExcess (T, x )
(A.12)
The APR EoS is based on the Peng and Robinson equation of state
[51]:
a
RT
P=
2
V b
V + 2bV b2
(A.3)
smix =
Appendix A. Advanced PengRobinson equation of state
(APR EoS)
(A.2)
sExcess (T, x ) =
nc
i=1
xi ln
nc
xj aij
(A.13)
j=1
(A.14)
where Aij and Bij are adjustable constants that can be used to match
liquid densities of mixtures. The default value of these constants are
zero unless otherwise set.
References
[1] A.J. Kidnay, W.R. Parrish, D.G. McCartney, Fundamentals of Natural Gas Processing, CRC Press, Boca Raton, FL, USA, 2011.
[2] A.L. Lee, M.H. Gonzalez, B.E. Eakin, The viscosity of natural gases, J. Petrol.
Technol. 18 (8) (1966) 9971000.
[3] C. Whitson, M. Brule, Phase behavior, monograph, Soc. Petrol. Eng. 20 (2000)
26.
[4] F.E. Londono, R.A. Archer, T.A. Blasingame, Simplied Correlations for Hydrocarbon Gas Viscosity and Gas DensityValidation and Correlation of Behavior
Using a Large-Scale Database, Paper No. 75721 at the SPE Gas Technol. Symp.,
Calgary, Alberta, Canada, 2002.
[5] K. Lucas, Phase Equilibria and Fluid Properties in the Chemical Industry,
Dechema, Frankfurt, 1980, p. 573.
[6] M.L. Huber, Transport properties o f uids, their correlation, prediction and estimation, in: J. Millat, J.H. Dymond, C.A. Nieto de Castro (Eds.), IUPAC, Cambridge
Univ. Press, Cambridge, 1996 (Chapter 12).
[7] H.J.M. Hanley, Prediction of the viscosity and thermal conductivity coefcients
of mixtures, Cryogenics 16 (11) (1976) 643651.
[8] J.F. Ely, H.J.M. Hanley, Transport properties. 1. Viscosity of uids and mixtures,
Ind. Eng. Chem. Fundam. 20 (4) (1981) 323332.
[9] J.F. Ely, H.J.M. Hanley, A Computer Program for the Prediction of Viscosity and
Thermal Conductivity in Hydrocarbon Mixtures (TRAPP), NBS Technical Note
1039, 1981.
65