Chemical @+neering Sc~e Vol. 39, No. 6, pp. 1011-1016, 1984 oo(lR2509/84 13.00+ .

a0
Printed in the U.S.A. Pqmoo FTmlLtd.

VISCOSITY OF CRUDE OILS

KAREN SCHOU PEDERSEN and AAGE FREDENSLUND*

Instituttet for Kemiteknik, The Tech&al University of Denmark, DK-2800 Lyagby, Demark

PETER L. CHRISTENSEN
Risa National Laboratory, DK-4000 Roskilde, Denmark

and
PER THOMASSEN
Statoil, Forus, N-4001 Stavanger, Norway

(Received 25 July 1983; accepted 22 September 1983)

m&-A series of experimental viscosity data far seven different North Sea oils is presented. A new
corresponding states method for prediction of the viscosity of both gaseous and liquid hydrocarbon fluids
has heen developed. The required input is critical constants and the molecular weight of each component.
Very accurate results wew obtained for tbc crude oils using the C,+-&aracterization procedure of
Pedersen et al.[l]. The viscosity correIation is also shown to give satisfactory results for pure hydrocarbons
and binary mixtures.

1. INTRODUCIION The cylinder has a fixed bore, and the size of the
In order to use a viscosity correlation in petroleum sphere is selected according to the viscosity range of
reservoir simulation studies and in design of oil interest. During the calibration procedure a sphere
processing equipment, the correlation must be appli- size was chosen which over the range 0.2-2.0 CP gave
cable for a wide range of compositions, temperatures a linear relationship between the viscosity and
and pressures. Safe and fast convergence is needed t.(pwham- patid), In this way one assures that the same
because the viscosity must be predicted a great num- flow regime exists throughout the given range of
ber of times. To secure consistent calculations in the viscosity. The instrument can be positioned in three
critical region it is important that the correlation can different angles, and each position is calibrated.
he used for both gaseous and liquid phases. The
accuracy of the crude oil viscosity prediction de- Procedure
pends, naturally, on the quality of the analytical data The saturated fluid is charged under constant
available, and the results are sensitive to the charac- pressure to the evacuated viscometer. Charging is
terization procedures used for the heavy hydrocarbon terminated when the pressure in the viscometer
mixtures. The data comprise viscosities ranging from equals that of the charging vessel. Then the system is
0.2 to 4.8 cP. Two viscosity correlations commonly allowed to come to equilibrium and the pressure is
used for complex hydrocarbon mixtures have been maintained constant.
tested, and a new viscosity correlation based on the The amount of time needed for the sphere to travel
corresponding states principle is presented. through the fluid is recorded for each angle. When the
same value for the viscosity ( f 0.001 cP) is obtained
7..DATA for all three angles, the pressure is reduced, and a new
The saturated fluid viscosities of samples Nos. l-7 series of measurements may he started.
of Appendix A are determined using a Ruska model
1402-818-00 rolling ball viscometer. The instrument Accuracy
must be calibrated with fluids of known viscosity. The Some of the data reported in this work have been
viscosity of the unknown fluids is related to the time measured in parallel at two independent laboratories
t needal for a sphere to fall (roll) through an inclined both using rolling ball viscometers: Statoil and Core-
cylinder filled with the fluid of interest. lab (Aberdeen). The disagreement between the data
was never more than + 1% of the reported viscosities.
Calibration The viscosities were determined at different pres-
The instrument was calibrated with pure hydro- sures at the reservoir temperature. Inside the two-
carbons of known viscosities ranging from 0.2 to phase region the measured values refer to the liquid
2.0 CP. phase. The viscosity predictions are found very de-
pendent on the characterization procedures, and be-
*Author to whom compondena ihould be addressed. low the saturation pressure the phase equilibrium

1011
1012 K. S. PBDERSBNet al.

calculation has a large intluence on the calculated A group of substances obey the corresponding
results. To give an idea of the accuracy of the phase states principle if the functional dependence of 1 on
equilibrium calculations Appendix B lists measured p, and T,:
and predicted saturation pressures. The calculated
bubble point pressures are seen to be always within
10% of the experimental values.
The viscosity data for samples Nos. 8 and 9 of is the same for all substances within the group. In
Appendix A were obtained using a oscillating that case comprehensive viscosity data are only
viscometcr [2]. needed for one of the components of the group. That
All the viscosity data are given in Appendix C. component is then used as reference substance (u),
and the viscosity of another component (x) within
the group can be calculated from the following
3. ExlsTfNG vISCOSlTY CORRELATIONS
expression:
In the correlation of Jossi ef ~I.[[31the viscosity, q,
is related to a fourth degree polynomial in the
reduced density, pr = p Jp,:

[(II - t] ‘)r + lo-4]“4 = a, + a,. p, + aj. p,’

+a,*p,3+#y,‘p; (7)
(1)

where In Ely and Hanley’s model the deviation from the

simple corresponding states principle is expressed in
(I, = 0.10230 terms of two shape factors, 0( T,, V,, w ) and Q (T,, V,,
w), which enter eqn (7) as follows:
a, = 0.023364
a, = 0.058533
a, = - 0.040758

a, = 0.0093324

q* is the low pressure gas mixture viscosity and:

0 and 8 are functions of the reduced volume, V,

* = [$,(V%)~‘6 which itself is found by the corresponding states
model. Therefore the determination of 0 and @
i ~~,.~~~~R[f~~j.~~j~~ (2) involves an iterative procedure. To use the corre-
[ j=* sponding states method on a mixture it is necessary
to represent the mixture as a hypothetical pure fluid
n is the number of components in the mixture.
with a given T,, pc and MW. Ely and Hanley have
Pure components have a well defined pe but it is not
adapted the mixing rules of MO and Gubbins[6] for
obvious which pc to use for a complex hydrocarbon
T, and pe, whereas the average molecular weight is
mixture. Lohrenz et al. [4] suggest to calculate the
formed from considerations regarding the radial dis-
critical density of a petroleum fluid as follows:
tribution function for a fluid in non-equilibrium due
to presence of a shear. A version of Ely and Hanley’s
(3) model which uses methane as reference substance is
available from Gas Processors Association as a pro-
gramme called TRAPP. In a recent publication
Baltatu[q has extended the correlation to cover
petroleum fractions.
where the critical molar volume (ft3/lbmole) of the
For liquids, Tham and Gubbins[8] have suggested
C,+-fraction is found from the expression:
to express the deviation from the simple correspond-
ing states principle as a function of a rotational
v++ = 21.573 +0_015122MWc,, - 27.656. SG,,
coupling coefficient, an;, which enters the viscosity
+ 0.070615 MWc,+ -SG,+. (4) correlation as follows:

Ely and Hanley[5j have presented a corresponding

states model for prediction of viscosities. The starting
point is a reduced viscosity, rl,, deflned as:
 Viiity of crude oils 1013

In this expression P and Tare wed as independent where 6, is a fitted constant. Instead of using a mixing
variables instead of p and T. The coeihcient am is rule like equation (11) to 6nd an; of a mixture this
treated as an adjustable parameter, and for a number parameter is determined directly from the mixture
of pure fluids it is determined by a fit to experimental molecular weight:
liquid viscosities.
antin = 1 + b,MW&. (1%
4. DEVELOPMENT OF A NEW
VlSCOSlTY CORRELATION Then the final expression for a of the mixture be-
Christensen and Fredenslund[9] have extended the comes:
work of Tham and Gubbins to a model for prediction
of thermal conductivities and viscosities of both &=I+ F ‘(a,,,-_& (16)
gaseous and liquid mixtures. They use the mixing 0 6

rules for T,, P,, V, and MW suggested in [6]. To apply which reduces to:
the model to both gases and liquids a is found from
the following expression: % = 1 + bti,‘M W:i, where b2 = b&-b (17)
2.4349
P,
cr=1+
( >
-
2.660s
(am - 1) (10) The constants bl - ba in eqns (14) and (17) were
determined by a least squares fit to the experimental
data given in Appendix C. The resulting values were:
where aTy;is the a-value reported in [8]. The following
mixing rule is used for a:
6, = 0.291

b2= 7.747 x lo+

b, = 4.265
The latter model was in this work tried for predic- bd= 0.8579
tion of the viscosities of hydrocarbon mixtures. Good
results were obtained for mixtures of components of
approximately equal size, whereas the predicted vis- 5. SUMMARY OF THE CALCULATION PROCEDURE
cosities were much too low in those cases where the The jiscosity q&(P, T) of a hydrocarbon mixture
molecular weights of the components diRered much at a given pressure P and temperature Tis determined
as in crude oils. The difficulties of representing poly- from the relation:
disperse mixtures seem to be associated with the
computation of the average molecular weight. For
mixtures of molecules of different size the mixing rule
in [6] and [5’lboth lead to molecular weights below the
number average molecular weight, IMW,, expressed in
eqn (12), and the deviation increases with the degree
of polydispersity:
where
MW,=zx&fW,.
I

As opposed to these findings, our calculations

show that the larger molecules make a relatively
greater contribution to the viscosity than do the
smaller ones. If the contribution of each molecule had
been proportional to its weight, a weight average
molecular weight should be used:

zxiMW:
(13)
MWw = flx,MW; - -
MW,,,,. = MW. + 0.291(MW, - MW,,);

That average molecular weight appeared, however, MW,,= cxJfWi (21)

to be too large and it was finally decided to find the
mixture molecular weight, MW& from the following
expression: a& = 1 + 7.747 * 10-s pp.265A!fWmfZr) (22)

M W,, = M W, + b,(MW, - iii%.) (14) a, = 1 + 8.374 - 1O-4 p,4.265 (23)

1014 IL S. PEDEXS~N
er al.

In eqns (22) and (23) Method 2 is identical to method 1 with the excep-
tion that the constants a, - a, of eqn (1) are found by
T-T, P.P,, a least squares fit to the experimental viscosities of
’ Tcpix ‘c Appendix C. The resulting values are:
P,= (24)
PC0
al = 0.12040
The reference substance is methane and the methane
a, = -0.13808
correlation given in [IO] is used.
as = 0.31723
6. RESULTS
a, = - 0.15878
Four different procedures for viscosity prediction
were tried. In all cases the characterization pro- a, = 0.025992.
cedures reported in [I] were used for the C,+-fraction.
Method 3 is the TRAPP-programme developed by
Method 1 uses eqn (I) where the critical density p, Ely and Hanley[S].
is found from eqns (3) and (4) and SRK-equation is V, of a C,+-fraction is found. by the method of
used to find the density p. The dilute gas mixture Riedel[l3]:
viscosity tl* is determined as follows [111:
v, = 7 [3.72 + 0.26& - 7.0)]-’ (2%
c

where

1
(TAT,) ln PC
a, = 0.9076 1.o f (30)
For the dilute gas viscosity ~7 of the individual 1.0 - T,,/T,
components the following expressions are used[ 121:
Method 4 is the new corresponding states pro-
r7: = 34 x IO-$-l- ciW for T,, < 1.5 cedure outlined in Section 5.
(26)
ti The results are presented in Appendix C and
summarized in Table 1. In Tables 2 and 3 are given
11: = 17.78 x LO--‘1(4.58T,, - 1.67)5” for T,, > 1.5 the results obtained for a number of pure fluids and
t binary mixtures. Convergence problems were en-
(27) countered with method 3 in the iterative calculation
of the shape factors of the heaviest components.
where No experimental gas viscosity data were included
in the development of the model presented in this
work. Appendix D gives predicted gas phase vis-
cosities for the gas phase in equilibrium with the
crude oil liquid phases below the saturation pressure.
In eqn (28) T, is in K and PC,in atm. The viscosities Methods 14 and method 5, the correlation of Carr
of eqns (25)(27) are expressed in cP. et al [14] give results which agree to within 10%.

Table 1. Summary of the results for crude oils (detailed results in Appendix c)

_
N: Number of experimental resqlts
AAD: Average absolute per cent deviation

BIAS: Per cent bias

 Viscosity of crude oils 1015

Table 2. Summary of resuh for pure fluids

I
' Method 1 Method 2 Method 3 Method 4
Substance N Reference
AADIBIAS 1 AAD [BIAS i RAD BIAS 1 AAD IBIAS I

I co, 1 10 1 32.3) -32.31 46.51 46.51 4.31- 2.8 I 15.11-15.~ 15 I

I C, 1 7 1 12.7 1 -12.71 78.91 78.91 6.5 1 6.2 1 5.71- 0.91 16 I

1 C, 1 8 127.51 -25.51 39.51 39.51 5.91- 4.5 1 12.51-10.61 17 I
I nc, 1 2 1 53.51 -53.51 17.41-17.41 2.6 I- 0.9 ] 4.51- 1.91 18 I
% 5 49.0 -49.0 10.9 8.7 2.9 - 1.7 6.9 - 6.9 19

Cyclohexane 4 64.2 -64.2 42.9 -42.9 43.1 -43.1 29.4 -29.4 20

I I I I I I I I I I

BenZene 1 4 1 42.41 -42.41 18.21 14.21 7.11- 6.7 1 4.81- 4.81 19 1

"C7 4 59.1 -59.1 29.6 -29.6 3.1 2.9 4.0 2.2 18

Toluene 4 45.6 -45.6 10.0 2.9 5.4 5.4 12.7 12.7 19

“=a 4 67.8 -67.8 44.0 -44.8 1.5 1.5 4.6 4.1 18

"C9 4 67.2 -67.2 40.1 -40.1 3.0 2.4 7.2 2.7 18

nc.. 4 73.2 -73.2 46.0 -46.0 2.7 2.7 8.1 5.1 18

"Ci2 3 89.7 -89.7 81.3 -81.3 20.0 -17.9 10.5 0.3 21

"Cl5 2 94.7 -94.7 89.9 -89.9 * - - 28.2 4.5 21

"'18 2 97.4 -97.4 94.5 -94.5 + - - 29.8 12.5 21

N, AAD and BIAS are defined in table 1

l No results of method 3 are given for Ci.5 and Cl8 due to convergence problems

Table 3. Summary of results for binary mixtures

Method 2 Method 3 Method 4

Reference
AAD IBIAS AAD BIAS AAD BIAS
I I I I t I

73.1 73.1 3.5 0.0 6.4 - 3.1 22

I-C
:,3
.3 Dimetfivlbutane 2 53.2 53.2
Dimfjzhylbutane
0 2 60.7 60.7
nC
2.5 Dimet (i
vlhexane 1 61.8 -61.8

nC6 - "Cl6 4 83.7 -83.7

"C7 - "Cl2 3 78.1 -78.1

2.4 DiIn$~hylpentane
7 2 55.3 -55.3
2.4 Dimetbylpentane-
2.2.5 Trimethvlhexane 2 65.4 -65.4
2,5 Dimethylhexana-
2,2.4 Trimethvlr.xntane 2 61.4 -61.4
llC - nc 2 74.6 -74.6 56.81-56.8 1 3.91 3.9 9.9 1 9.9 25
8 I

N, AAD and BIAS are difined in table 1.

1016 K. S. Pea- et al.
7. DISCUSSION Greek symbols
The best results for crude oil viscosities are ob- parameter defined in eqn (17)
tained with method 4, i.e. using the corresponding parameter defined in eqn (9)
states principle with correction factor u. The results parameter defined in eqn (30)
for pure fluids and binary mixtures show that the viscosity
method is not limited to crude oils. In most cases the dilute gas viscosity
predictions are in excellent agreement with experi- density
mental values. However, for binary mixtures of com- parameter &fined in eqn (2)
ponents of very different size like methanedecane the parameter defined in eqn (28)
deviations between experimental and predicted vis- shape factors
cosities is somewhat large. The density dependence of
Subscripts
the parameter tl expressed in eqn (22) secures a
c critical property
continuous change towards the simple corresponding
i,j component indices
states principle as the fluid becomes less dense and
mix mixture
good results are therefore also to be expected for
o reference substance
gases.
r reduced property
Since a is determined directly from the mixture
x substance x
properties the computation is fast and without con-
vergence problems. REFERENCES
The method of Jossi et al. [3] does not seem flexible [l] Pedersen K. S., Thomassen P. and Fredenslund Aa.,
eneough to be able to represent the viscosity vari- Thermodynamics of Petroleum Mixtures Containing
ations which take place during, e.g. a pressure de- HeavyHydrocarbonsInd. Engng Chem. Proc. Des. Dev.,
pletion process. Neither method 1 with the original in Press.
[Z] 0ye H. A. and Knapstad B., Viscosity of petroleum
a1 - 6 coefficients nor method 2 where the
reservoir@Xs. Report No. LAB-KSA/RNH 18.11.81,
coefficients have been determined by a fit to the University of Trondheim- Norwegian Institute of Tech-
experimental data give satisfactory results. nology 1981.
The results obtained with the TRAPP-programme [3] Jossi J. A., Stiel L. I. and Thodos G., A.1.Ch.E. J. 1962
8 59.
are quite satisfactory for low viscosities. However,
[4] Lohrcnz J., Bray B. G. and Clark C. R., J. Pet. Technol.
the programme seems unable to represent highly Oct. 1964 1171.
viscous oils. [5] Ely I. F. and Hanley H. J. M., Znd. Engng Chem. Funa%
1981 XI 323.
8. CONCLUSION [6] MO K. C. and Gubbins K. E., Mol. Phys. 1976 31 825
171
_ _ Baltatu M. E.. Znd.Ewnz -_ Chem. Proc. Des. Dev. 1982
A new viscosity correlation based on the corre- 21 192.
sponding states principle has been developed. It is [8] Tham M. J. and Gubbins K. E., Ind. Engng Chem.
applicable to naturally occuring hydrocarbon mix- Fun& 1970 9 63.
[9] Christensen P. L. and Fredenslund Aa., Chem. Engng
tures ranging from dilute gases to heavy oils. The
Sci. 1980 35 871.
only needed input is critical constants and the molec- [lo] Hanley H. J. M., McCarty R. D. and Haynes W. M.,
ular weight of each component. Low computation Cryogenics 1975 15 413.
time and safe convergence make the model specially [ 1l] Heming F. and Zipperer L., Gas u. ~uwerfach 1936 79
suited in reservoir simulation studies and other two 69.
[12] Stiel L. I. and Thodos G., A.I.Ch.E.1. 1961 7 611.
phase flow calculations. [13] Riedel L., Chem. 1ng. Technick 1954 26 83.
[14] Cart N. L., Kobayashi R. and Burrows D. B., J. Pef.
Although the parameters of the correlation were Technol. 1954 6 47.
fitted to viscosity data for crude oils with very many [ 15] Herrcman W., Grevcndonk W. and De Bock A., J.
Chem. Phys. 1970 53 185.
components, the correlation is found to yield good
[16] Swift G. W., L.ohrenz J. and Kurata F., A.1.Ch.E. J.
results also for pure and binary hydrocarbon systems. 1960 6 415.
[ 17] Carmichael L. T., Berry V. M. and Sage B. H., J. Chem.
SUPPLEMENT Engng Data 1964 9 411.
[18] Doolittle A. K. and Petersen R. H. J. Am. Chem. Sot.
Appendices A-D are contained in a supplement 1951 73 2145.
which may be obtained from Aage Fredenslund. [ 191 Medani M. S. and Hasan M. A., J. Chem. &gng Data
1978 23 34.
1201
_ _ Jonas J.. Hasha D. and Huana S. G., J. Phys. Chem.
NOTATION 1980 84-109.
constants detied in eqn (1) [2 l] Hogenboom D. L., Webb W. and Dixon J. A., J. Chem.
UI - a5
constants defined in eqns (14)-(17) Phys. 1967 46 2586.
b, -bs [22] Huang E. T. S., Swift G. W. and Kurata F., A.I.Ch.E.
MW molecular weight J. 1967 13 846.
n number of components in the mixture [23] Lee A. L., Gonzalez M. H. and Fakin B. E., J. Chem.
P pressure gravity Engng Data 1966 11 281.
T temperature _ _ Heric E. L. and Brewer J. G., J. Chem. Engrrg Data
1241
1967 12 574.
TB normal boiling point [25] Dixon J. A., J. Chem. Engng Data 1959 1 289.
mole fraction [26] Trevoy D. J. and Drickamer H. G., J. Chem. Phys.
z compressibility factor 1949 17 582.