Available online at www.sciencedirect.

com

International Journal of Heat and Mass Transfer 51 (2008) 33873398

www.elsevier.com/locate/ijhmt

Mathematical formulation and numerical modeling of wax

deposition in pipelines from enthalpyporosity approach
and irreversible thermodynamics
R. Banki a,1, H. Hoteit b,2, A. Firoozabadi b,c,*
a
Imperial College, London, United Kingdom
b
Reservoir Engineering Research Institute (RERI), Palo Alto, CA, USA
c
Yale University, New Haven, CT, USA

Received 6 July 2007; received in revised form 1 November 2007

Available online 7 February 2008

Abstract

In the last 10 years, there have been a number of studies in modeling of the deposition processes in owlines. Most of these models: (1)
assume empirical or semi-empirical correlations to predict the pressure drop and temperature prole, (2) ignore the radial convection
ow in the layer composed of the two-phase wax and oil (that is the gel layer), and (3) use Ficks law to describe the diusion ux of
species towards the wall by using the chain rule to relate concentration gradient to temperature gradient. In this work, a rigorous math-
ematical model for the prediction of wax deposition in pipelines is presented for laminar ow. The transient deposition of each compo-
nent is calculated from the solution of the coupled momentum, energy and, species balance equations, and a thermodynamic wax
precipitation model at the local level. An enthalpy formulation based on a xed-grid approach is used to approximate the convection
ow in the gel layer. We do not use the chain rule to relate composition gradient to temperature gradient in Ficks law to avoid violating
the laws of irreversible thermodynamics. Our diusion ux expression includes molecular diusion (concentration gradient is driving
force) and thermal diusion (temperature gradient is driving force) with appropriate diusion coecients. This work also includes
the description of the numerical solution of the governing equations. Numerical results and features of wax deposition as well as model
verication with experimental data are presented in a separate paper.
2007 Published by Elsevier Ltd.

Keywords: Wax deposition; Multicomponent diusion; Wax hardening and aging

1. Introduction and wax and scale formation in owlines in petroleum

Solid deposition from liquids in both natural and indus-
trial processes can be undesirable and may result in harm.
Examples of solid precipitation include frost formation on
cold surfaces, crystallization fouling in heat exchangers,
*
Corresponding author. Address: 385 Sherman Avenue, Suite 5, Palo
Alto, CA 94306, USA. Tel.: +1 650 326 9172.
E-mail address: abbas.roozabadi@yale.edu (A. Firoozabadi).
1
Now with Halliburton, 2107 CityWest Blvd., Building 2, Houston, TX
77042, United States.
2
Now with Conocophillips, 600 North Dairy Ashford (77079-1175),
Houston, TX 77252, United States.
production.
As the search for oil and gas moves towards deeper
waters such as the Gulf of Mexico and in the North Sea,
the deposition of wax crystals in oil and gas pipelines
becomes a major concern. The deposition in production
tubings and pipelines is undesirable because of decrease
in the ow rate and other operational complexities. To pre-
vent blockage of pipelines, wax deposits should be removed
periodically. Dierent mechanical, thermal, and chemical
techniques can be used for wax removal [15].
Petroleum uids are composed of a diverse group of spe-
cies. The paranic groups with carbon numbers of say 15

0017-9310/$ - see front matter 2007 Published by Elsevier Ltd.

doi:10.1016/j.ijheatmasstransfer.2007.11.012
3388 R. Banki et al. / International Journal of Heat and Mass Transfer 51 (2008) 33873398

Nomenclature

Greek symbols Dhfi molar enthalpy of fusion of component I,

dik = dki interaction parameter between component i J/mol
and k (Jj,i)r radial diusion of component i, mol/m2 s
e porosity Jj,i diusion ux of component i in phase j, mol/
l viscosity, Pa s m2 s
q density, kg/m3 Jj diusion ux vector in phase j
s viscous stress J total diusion ux vector
ki molar latent heat of solidication of component K permeability, m2
k thermal conductivity, W/m K
Alphabet symbols L pipe length, m
A momentum equation sourceP P term coecient n total number of components
a PR-EOS parameter ni1 nk1 xi xk aik ns number of solid phases
1=2 1=2
aik PR-EOS parameter 1  dik ai ak p pressure, Pa
2 2 Pci critical pressure of component i, Pa
ai PR-EOS parameter P 0:45724R T ci =P ci
b PR-EOS parameter ni1 xi bi P fi melting point pressure of component i, Pa
bi PR-EOS parameter 0:07780RT ci =P ci p0 outlet pressure, Pa
c total molar density, mol/m3 Q inlet volumetric ow rate, m3/h
cj molar density of phase j, mol/m3 q energy ux relative to the massaverage veloc-
cji molar density of component i in phase j, mol/m3 ity, W/m2
Cpji heat capacity per unit mole of component i in q(x) Dufour eect, W/m2
phasej, J/mol K R radial distance from pipe center, m
Cpj heat capacity of per unit mole of phase j, J/mol K R gas constant, J/mol K
DCpi dierence in heat capacity of component i in R0 pipe radius, m
liquid and solid state, J/mol K Re Reynolds number
DM j;i;k molecular diusion coecient in phase j, m2/s S source term
DM molecular diusion coecient matrix Sj saturation of phase j
DM molecular diusion coecient matrix of phase j T time, s
j
T temperature, K
DjiT
thermal diusion coecient in phase j, m2/K s Ta ambient temperature, K
T
D thermal diusion coecient vector Tci critical temperature of component i, K
DTj thermal diusion coecient vector of phase j Tin inlet temperature, K
e multicomponent energy ux, J/m2 s v velocity, m/s
F feed mole number, mol vji velocity of component i in phase j, m/s
fi fugacity of component i, Pa vr radial velocity, m/s
Hj molar enthalpy of component i in phase j, J/mol vz axial velocity, m/s
H* ideal gas molar enthalpy of mixture at zero pres- xji molar composition of component i in phase j
sure, J/mol Z compressibility factor
H i ideal gas molar enthalpy of component I at zero zi overall composition of component i
pressure, J/mol

to some 60 have high crystallization temperatures and may
crystallize even at low concentration in the petroleum uid
mixture. The crystallites from petroleum uids are often
referred to as waxes; the temperature at which crystalliza-
tion occurs is referred to as wax-appearance temperature
(WAT).
An understanding of the mechanisms of the wax deposi-
tion and modeling of the process is likely to be a key step
towards an optimum design and prevention of the
problem.
There has been an extensive eort in the last decade to
develop theoretical models for wax deposition calculations.
In recent years, there has been success in the prediction of
wax precipitation from crude oils and gas condensate uids
for local equilibrium calculations [6]. However, despite
much progress, the more complicated problem of wax
deposition in non-isothermal owlines is still in an early
stage of development.
There are two main processes that aect the deposition
of wax in owlines: (1) heat transfer, and (2) species ow.
The heat transfer is mainly from conduction and convec-
tion. The ux of species by convection and diusion is
expected to strongly inuence the deposition. To the best
of our knowledge, the radial convective velocity in the
region containing the deposition for the laminar ow has
been neglected in the past [716]. The radial velocity is
 R. Banki et al. / International Journal of Heat and Mass Transfer 51 (2008) 33873398
expected to be small in comparison to axial velocity, but its
contribution to species ux in the radial direction may be
comparable to radial diusion ux (due to non-uniform
wax deposition, radial convection develops). One objective
of this work is to include the radial convective ux in the
formulation.
Wax deposition is a moving boundary problem and the
numerical solution of the balance equations can be a chal-
lenge. A large number of numerical techniques are avail-
able for the solution of a moving boundary problem.
There are two main approaches. In the transformed-grid
approach, the governing dierential equations and their
boundary conditions are cast into a generalized curvilinear
coordinate system. The equations can then be solved on a
uniform-rectangular grid, which remains xed in space
and time. In essence, the moving boundary is immobilized.
Extra terms in balance equations may be required for
conservation of mass, momentum, and energy [17]. In the
xed-grid approach, which is also known as the
enthalpyporosity approach [18], a xed grid is applied
directly in real space (that is, the problem domain) and
the interface conditions are accounted for by the denition
of suitable source terms in the governing equations [1820].
The enthalpyporosity approach has been successfully
implemented in many engineering problems that involve
a liquid to solid phase change such as freezing of a pure
liquid in a thermal cavity and melting of a pure metal
[18,21,22]. In this work, we adopt this approach for the rst
time in the formulation of wax deposition by treating the
layer with deposition as a pseudo-porous medium. Various
authors have shown that the crystallization of parans in
oil leads to the formation of gel with a complex morphol-
ogy. The gelation is due to ocuulation of orthorhombic
wax crystallites that appear in the solution. Observation
with cross-polarized microscopy has revealed that the
crystallites have structures of platelets that overlap and
interlock and a network structure of wax crystals are
formed [15,23,24]. Therefore, the gel behaves as porous
media. In this work the Darcy-type source term is added
to the momentum equation to describe the deceleration
of ow in the gel layer. The energy equation for the multi-
component, two-phase ow is written in terms of the
enthalpy. The relationship between the enthalpy and the
temperature can be described by an equation of state.
To the best of our knowledge, in all the models in the
literature the driving force for radial diusion is rst
written in terms of concentration gradient and then
through the chain rule, the temperature gradient is intro-
duced [711,1316,25,26]. In other words, various authors
use the molecular diusion coecients with the driving
force represented by temperature gradient. Thermodynam-
ics of irreversible processes oers a dierent representation
of diusion with temperature gradient. In our formulation,
we represent the molecular diusion (driving force is
concentration gradient), and thermal diusion (driving
force is temperature gradient) from thermodynamics of
irreversible processes. Proper representation of diusion
3389
uxes and the study of diusion eects are major goals of
this work.
This paper is organized as follows. First, a detailed
description of the mathematical model is presented for
the coupled momentum, energy, and mass balances. We
then provide the thermodynamics of wax-precipitation
model followed by presenting the numerical formulation
of the governing equations and the algorithm. The work
is ended with a general summary and concluding remarks
emphasizing that concentration gradient and temperature
gradient may not be related through a chain rule.
2. Domain denition and basic assumptions
We consider a horizontal pipe of length L and inner
radius R0 (Fig. 1). A multicomponent uid is injected at
one end with a volumetric ow rate Q at temperature,
Tin. The pipe wall temperature is set at constant tempera-
ture, Ta where Ta < Tin. The outlet pressure is held
constant at p0. We assume single-phase and gel subdomains
as shown in Fig. 1. The formation of solid wax crystals
creates a gel layer, which consists of a liquid phase and a
nonmoving solid phase [15]. Studies of distillate fuels
and model petroleum uids reveal that as little as 2%
of the crystal precipitate is required to gel the uid
[15,23,24,27]. When the amount of crystallites increases,
the gel hardens. The process is called aging of the gel layer.
The following assumptions are made in our work:
(1) Shear dispersion and Brownian diusion are
neglected.
(2) Gravity is neglected.
(3) Diusions (both thermal and molecular) are
neglected in the axial direction.
(4) Diusion is neglected in the solid phase.
(5) Solid heat capacity and uid thermal conductivity of
each component are assumed temperature indepen-
dent.
(6) Laminar ow is assumed.
(7) Multisolid wax precipitation is assumed.
Ta
Gel layer
R0
Liquid region
r
z
Fig. 1. Computational domain, the geometry of problem is a 2D circular
coordinate system.
3390 R. Banki et al. / International Journal of Heat and Mass Transfer 51 (2008) 33873398

We use vz and vr to denote the velocity components in the

liquid phase in the z- and r-directions, respectively. In the
two equations above, the stress terms are given by
   
ovz 1 ovr 1
szz 2l  r:v ; srr 2l  rv ;
oz 3 or 3
   
vr 1 ovz ovr
shh 2l  r  v ; srz szr l :
r 3 or oz

The divergence of velocity vector v in 2D-cylindrical coor-

Fig. 2. Wax deposition in a subsea pipeline, a schematic view on how
severely wax deposits can obstruct the pipeline.
By neglecting the gravity eect, the 3D computational
problem is reduced to 2D (see Fig. 1). There is ample evi-
dence, based on the examination of the deposition in pipe-
lines, that this is a good assumption. Fig. 2 shows one
example of wax deposition in subsea pipeline where the
whole line had to be abandoned due to a hardened nature
of wax deposits.
3. Mathematical model
Below the wax appearance temperature (WAT), part of
heavy components may precipitate and form a gel layer
next to the pipe wall. The gel layer consists of liquid and
immobile solid wax. The hardening of the gel layer is
mainly by convection and diusion and the subsequent
increase in the fraction of deposited wax. The temperature,
pressure, velocity eld, and composition are related
through:
(1) Momentum balance.
(2) Energy balance.
(3) Species balance.
(4) Local solid/liquid equilibrium.
dinates is given by
ovz 1 o
rv rvr : 3
oz r or
The system of Eqs. (1) and (2) is subject to the following
boundary conditions:
vz r R0 ; z; t 0; vz r; z 0; t v0 ;
vr r 0; z; t 0; vr r R0 ; z; t 0;
pr; z L; t p0 ; 4

ovz 
0:
or  r0;z;t
In the above equations, v0 is the uid axial-velocity at the
inlet and p0 is the pressure at the outlet. As mentioned be-
fore, the formation of the solid wax produces interlocked
platelets with liquid in the solid pore space resulting in
gel formation. Due to crystallization, a moving solidliquid
boundary forms. The major challenge is how to account for
the deceleration of the uid ow in the gel region with a
xed grid. Voller et al. [17,18,21] have introduced an en-
thalpyporosity approach to simulate freezing of liquids
in a thermal cavity. The enthalpyporosity approach treats
the phase change as a pseudo-porous medium with poros-
ity, e, decreasing from 1 to 0 as the solid fraction increases
from 0 to 1. A Darcy-type source term is added to the
momentum equations in order to describe the velocity
deceleration in the mushy region.
In this work, we adopt the enthalpyporosity approach
to model the bulk ow in the gel layer and add the follow-
ing source terms:

3.1. Momentum balance S z Avz ; 5

S r Avr 6
In cylindrical coordinates, the momentum equation is
given in the z- and r-directions for a compressible single- to the right-hand side of Eqs. (1) and (2), respectively. In
phase liquid ow by the following expressions [28,29]: the liquid region, the source terms are set to zero. In the
  gel layer (where liquid is trapped in the wax platelets),
oqvz 1 orqvr vz oqvz vz the parameter A is large enough such that the source terms

ot r or oz dominate the transient, convective, and diusive terms and
 
op 1 o oszz thus the momentum equation approximates Darcys law.
 rsrz ; 1
oz r or oz As the solid saturation increases to one, the source terms
  dominate all the other terms in the momentum equation,
oqvr 1 orqvr vr oqvz vr
and forces the liquid velocities close to zero. The source
ot r or oz
  term is thus a function of the porosity of the gel medium
op 1 o shh oszr (i.e., the wax saturation). One way to dene a suitable form
 rsrr  : 2
or r or r oz for A is to use the CarmanKoseny equation [21]. The Car-
 R. Banki et al. / International Journal of Heat and Mass Transfer 51 (2008) 33873398 3391
!
manKoseny equation relates the eective permeability K o X n X
n

to the porosity e through the equation [30]: ci H i r ci H i v r  krT : 13

ot i1 i1
1 e3
K ; 7 The energy equation in the gel layer is written in terms of
C 1  e2 the enthalpies of liquid and solid phases:
! !
where C is the morphology coecient. Darcys law can o X X
then be approximated by S j cj H j r: S j c j H j vj
ot jo;s jo;s
2
l 1  e r  k eff rT ; 14
rp  v  Cl v: 8
K e3
where the index j = o, s refers to the oil phase and wax
From Eq. (8) we obtain phase (solid phase), respectively; Hj and Sj are, respec-
tively, the molar enthalpy and volume fraction (saturation)
1  e2
A Cl : 9 of phase j, and k and ke are the liquid thermal conductivity
e3 q in the single phase liquid region and the eective thermal
The value of C depends on the morphology of the porous conductivity in the gel layer, respectively.
media. As an example for a packed bed of spheres C = 150/ The energy equation is subject to the following bound-
Dp2, where Dp is the sphere diameter; with this expression ary conditions:
and Eq. (7) one can estimates an accurate value for perme- T R0 ; z; t T a ; T r; 0; t T in ;
ability [31]. For metal formation with grains of Dp = 1 mm  
oT  oT  15
a value of C = 103 m2 is obtained; this value was used for 0; 0:
or  oz 
the Gallium solidication study [21]. For wax morphology r0;z;t r;zL;t

we nd C = 106 m2 for describing ow in the gel layer as The last expression in Eq. (15), which ignores the axial
we will comment in Hoteit et al. [32]. thermal conduction at the outlet, is based on a pseudo-stea-
dy state assumption. The eective thermal conductivity ke
3.2. Energy balance in Eq. (14) can be calculated from the Maxwell correlation
[33]:
The general form of the energy equation for a single 2k o k s  2S s k o  k s 
phase, multicomponent mixture can be written as [28]: k eff ko: 16
2k o k s S s k o  k s 
!
o X n
DP One may use other models in the calculation of the eective
ci H i r  e T : rv; 10
ot i1 Dt thermal conductivity of a two-phase mixture [34]. The mo-
lar enthalpy of the oil phase can be calculated from the PR-
where H i is the partial molar enthalpy of component i, and EOS [35]:
DP/Dt is the energetic contribution of the pressure due to
H o T ; p; xo H  T ; 0; xo RT Z  1
expansion or compression. The term T is the stress tensor  
and T:$v is the energetic contribution of viscous dissipa- T daT aT Z 2:414B
p  p ln ;
tion. The term e is the multicomponent energy ux relative 2 2b dT 2 2b Z  2:414B
to xed coordinates; it is dened as [28]: 17
X
n
where Z = pvo/RT is the compressibility factor of the oil
ev ci H i q: 11
i1
phase and H*(T, 0, xo) is the ideal gas enthalpy of liquid
P mixture at zero pressure, which can be calculated from:
The term v ni1 ci H i is the energy ux due to convection. X
n
The term q is given by [28]: H  T ; 0; xo xo;i H i T ; 0: 18
i1
X
n
q krT J i H i qx ; 12 In Eq. (18), H i T ; 0 is the ideal gas enthalpy of component
i1
i. Passut and Danner [36] provide correlations for ideal gas
whereP k$T is the energy ux due to conduction and the enthalpy of some 90 substances.
n
term i1 J i H i represents energy transfer due to diusion We use the following expression to calculate the
of species. The term q(x) represents the Dufour or diu- enthalpy of the solid wax:
sion-thermo eect, which is usually negligible. Neglecting X
n

the energetic contributions of the pressure DP/Dt, viscous H s T ; p; xs xs;i H s;i T ; p: 19

i1
dissipation
Pn T:$v, and the inter-diusion of enthalpy
i1 J i H i , the energy equation for the single-phase liquid Assuming that pressure has negligible eect on the molar
ow can be written as enthalpy of the solid, and neglecting the eect of tempera-
3392 R. Banki et al. / International Journal of Heat and Mass Transfer 51 (2008) 33873398

ture variation on heat capacity of component i in the solid The total molar density c is dened by
state, we can write [35]: X
c S j cj : 29
H s;i T ; p  H s;i T ; pfi  H s;i T fi ; pfi Cps;i T  T fi : 20 jo;s

From the denition of molar latent heat of component i The overall composition of component i is related to the
(heat of fusion): phase composition of component i by
X
ki H o;i T fi ; pfi  H s;i T fi ; pfi : 21 czi S j cj xj;i ; i 1; . . . ; n: 30
jo;s
Eqs. (19)(21) are combined to obtain
Using Eq. (30) one can write Eq. (28) as
X
n
H s T ; P ; xs xs;i H o;i T fi ; pfi  ki Cps;i T  T fi : o X
i1 czi r  S j cj xj;i vj S j J j;i 0; i 1; . . . ; n:
ot jo;s
22
31
In this work, the terms ki ; H o;i T fi ; pfi ; and Cps,i are as-
sumed constant; ki can be evaluated as the average value The total molar balance from Eq. (31) can be written as
at T and T fi . oc X
r S j cj xj;i vj 0; i 1; . . . ; n: 32
ot jo;s
3.3. Species balance
Eqs. (31) and (32) are subject to the following boundary
The gel layer has structure of platelets that overlap and conditions:
interlock and may contain up to 98% liquid in early time of
deposition [23,24,37]. The gel layer may grow with time. zi zi in at z 0;
Let us consider ow in two-phase in the domain and write J j;i r 0 at r 0; R0 ; 33
the mass balance of species i: ozi
0 at r 0:
o X X or
S j cj;i r: S j cj;i vj;i 0; i 1; . . . ; n; 23
ot jo;s jo;s
3.3.1. Diusion ux
where cj,i and vj,i are the molar density of component i in The molar diusion ux in a vector form can be written
phase j, and the velocity vector for component i in phase as [38]:
j, respectively. The relation between the bulk velocity, vj
of phase j, molar diusive ux, Jj,i, and velocity of compo- J cDM rx DT rT DP rp: 34
nent i in phase j, vj,i is given by M T P
where J, D , D and D are the total diusion ux vector,
S j cj;i vj;i S j cj;i vj S j J j;i ; i 1; . . . ; n: 24 molecular diusion coecient matrix, thermal diusion
coecient vector, and pressure diusion coecient vector,
Substitution of Eq. (24) into Eq. (23) yields
respectively. On the right side, the rst, second, and third
o X X terms represent molecular diusion, thermal diusion,
S j cj xj;i r  S j cj xj;i vj S j J j;i 0; i 1; . . . ; n;
ot jo;s jo;w and pressure diusion, respectively. Neglecting the pressure
diusion, the diusion ux of component i, i = 1, . . . , n  1,
25
in phase j is given by
where cj and xj,i are the molar density of phase j and mole
J j cj DM T
j rxj Dj rT : 35
fraction of component i in phase j. The molar density of
phase j and molar density of component i in phase j are re- where Jj = [Jj,i] and $xj = [$xj,i]; DM M
j Dj;i;k ; k 1; . . . ;
lated by T T
n  1 and Dj Dj;i  are the molecular diusion coe-
cj;i xj;i cj : 26 cients and the thermal diusion coecients in phase j,
respectively. Firoozabadi et al. [38,39] describe the method-
Note that in each phase ology for calculating molecular and thermal diusion
X
n coecients. Eq. (35) consists of two terms. The rst term
J j;i 0; j o; s: 27 on the right side is known as the Ficks law and the second
i1
term is often referred to as the Soret eect. In a multicom-
By summing Eq. (25) for i = 1, . . . , n and using
P ponent mixture such as a petroleum uid with varying
n
i1 xj;i 1; j o; s, and Eq. (27), the total molar balance molecular sizes and molecular shapes, in addition to diag-
expression is given by onal molecular diusion coecients, cross coecients may
o X X become important [40].
S j cj r  S j cj vj 0: 28 To the best of our knowledge, in all of the existing
ot jo;s jo;s models for wax deposition, the diusion ux in a (pseudo)
 R. Banki et al. / International Journal of Heat and Mass Transfer 51 (2008) 33873398 3393
" #
binary mixture in the radial direction is written in the fol- X
n
N s;k
lowing form: zi  xo;i 1  0; i 1; . . . ; n  ns ; 40
knns 1
F
oz1 oz1 oT
J o;1 cDM
o;1;1 cDM
o;1;1 : 36 where fo,i(p,T,xo) is the fugacity of component i in the oil
or oT or
phase with composition xo,i, Ns,i is the number of moles
The concentration gradient in the Ficks law of diusion is
of component i in the wax phase, F is the total number
subject to constant temperature and pressure and thus the
of moles of the wax components.
expression above is not valid. The appropriate form of the
The fugacities of a pure component i in solid and liquid
diusion ux without pressure diusion is given by Eq.
states are related by [33]:
(35).
fs;ipure p; T fo;i
pure
p; T exp Dli ; 41
3.4. Solidliquid equilibria 
where Dli lpure pure f f f
o;i  ls;i =RT Dhi =RT i  DCp i =RT i
There are two types of models for wax precipitation cal- T fi =T  1  DCpi =RT fi ln T fi =T :
culations in petroleum uids; (1) solid solution, and (2) The symbols are dened in the Nomenclature. The
multisolid-phase. The multisolid model can describe the fugacity of pure component i in the liquid state can be
WAT and the amount of precipitation more accurately calculated from the PengRobinson equation of state.
than the solid solution model [41,42]. In this work, we per-
form the solidliquid phase calculations by using the mul- 4. Numerical model
tisolid-phase model by Lira-Galeana et al. [6], where each
precipitated component forms a solid layer, which does The momentum, energy, and species mass balance equa-
not mix with the other solid layers. The model uses a tions and the local wax solidliquid model are coupled and
two-step procedure for the phase-split calculation: solved with initial and boundary conditions to predict the
pressure, temperature, velocity and composition in the gel
(1) Stability analysis. and liquid regions as well as the wax fraction (i.e., volume
(2) Phase-split calculation. fraction or weight fraction). The 2D computational
domain is discretized into a structured grid of rectangular
3.4.1. Stability analysis elements. A nite-volume based method is used for the spa-
This is a preprocessing step to identify which of the com- tial approximations. Dierent temporal, spatial and linear-
ponents are precipitating and which are not. A component ization schemes are implemented to solve the governing
i precipitates at a given temperature and pressure if the fol- equations. The scalar unknowns like the pressure, temper-
lowing condition is satised [33]: ature, enthalpies, densities, mole fractions and saturation
are approximated over the same control volumes while
fi p; T ; z  fs;ipure p; T P 0; 37
the velocity is approximated over a dierent control vol-
where fi(p,T,z) and fs;ipure p; T
are the fugacity of compo- ume. The numerical algorithm uses two levels of iterations
nent i with overall composition z, and fugacity of pure so- at each time step. In the inner level, we solve the coupled
lid-component i at pressure p and temperature T, momentum and total mass balance equations to approxi-
respectively. mate the pressure and the velocity eld. In the outer level,
we solve the energy equation and the species balance equa-
3.4.2. Phase-split calculation tions to calculate the temperature and the overall composi-
Suppose the stability-analysis provides the condition tion. Once the overall composition, temperature, and
that ns components of the n-component mixture (i.e., the pressure are known, we apply the phase-equilibria model
components (n  ns + 1), . . . ,n) precipitate. These compo- to calculate the oil and wax compositions and the wax sat-
nents, therefore, fulll Eq. (37). uration in the gel layer.
The governing equations for the wax precipitation are Here, we provide the basic numerical methods used to
given by the equilibrium and the material balance equa- solve our system of equations.
tions. For every precipitating component i, the equilibrium
and material balance equations are, respectively, given by 4.1. Solution of the momentum equation
fo;i p; T ; xo fs;ipure p; T ; i n  ns 1; . . . ; n; 38
The momentum equation is solved by using the SIM-
and " # PLER method of Patankar [43]. In this method, a stag-
X
n
N s;k N s;i gered or a displaced grid for the velocity components is
zi  xo;i 1  0; i n  ns 1; . . . ; n: used. The velocity components are calculated for the points
knns 1
F F
that lie on the faces of the control volumes and the pressure
39 at the main grid points (see Fig. 3). For presentational con-
The material balance equations for the non-precipitating venience, the general discretization of axial and radial
components are momentum equations (combined Eqs. (1) and (5), and
3394 R. Banki et al. / International Journal of Heat and Mass Transfer 51 (2008) 33873398

A rst-order upwind scheme is used to dene the coe-

N cients anb nb
z and br at the control volume boundaries. These
n coecients with the right-hand term in Eq. (45) are evalu-
ated from the information at previous time steps. The
W w P E sought unknowns from Eqs. (42)(44) are the velocities vz
P and vr, and the pressure. A full description of the algorithm
can be found in Refs. [43,44].

4.2. Solution of the energy equation

Pressure
Axial velocity S Assuming that the solid wax phase is immobile, the
energy balance equation, Eq. (14) simplies to
Radial velocity !
o X
Fig. 3. Computational cells, discretization and staggered model based on S j cj H j r:S o co H o vo r  k eff rT : 46
ot jo;s
SIMPLER method.

The nite-volume method is used for the spatial discretiza-

Eqs. (2) and (6)) at the west- and north-cell faces of a scalar tion of Eq. (46). The convection term, which is the second
control volume P are given by: term in the left-hand side of Eq. (46) is solved by using a
X rst-order upwind scheme. The left-hand side of Eq. (46)
w
Awz S wz vwz anb nb P w w
z vz p  p az bz ; 42
nb
representing heat conduction is approximated by a central
nite dierence scheme. The time operator is approximated
and by a semi-implicit time scheme. The enthalpies Ho and Hs,
X and the temperature T are implicit in time. The other vari-
n
Anr S nr vnr anb nb P N n
r vr p  p ar br : 43
nb
ables are known from the previous time step iteration. The
spatial and temporal discretization of Eq. (46) over a con-
The spatial, temporal and uid dependant terms are incor- trol volume P can be written in the form (see the
porated into the coecients A, a and b in Eqs. (42) and Appendix):
(43). These coecients are calculated by using the power X
law scheme of Patankar [41]. The subscript nb refers to GP T  ao;P H o;P as;P H s;P  F nb H nb
the neighboring nodes of the control volume P. The dis- nb
X
cretized Eqs. (42) and (43) are similar to the general discret-  Dnb T nb  bP 0: 47
ized equations dened by Patankar except in two nb
additional terms S wz and S nr [43]. These terms, which appear
The coecients ao,P,aw,P,bP,Fnb and Dnb are calculated
on the right-hand side of Eqs. (42) and (43), are a conse-
from information at the previous iterations (see the Appen-
quence of the added Darcy-type source term to the momen-
dix). The enthalpies in Eq. (47) are functions of tempera-
tum equations. Note that these terms end up at the
ture and composition; they are dened in Eqs. (17) and
diagonal of the linear system whose solution is the velocity
(19). The linearization of Eq. (47) by the NewtonRaphson
eld. The values of the axial and radial velocities are thus
(NR) method yields
inversely proportional to S wz and S nr , respectively.
The discretization of the momentum equations leads to ao;P Cpo;P as;P Cps;P DT P
X
two equations and three unknowns; p, vz, and vr. To close  F nb Cpnb;P Dnb DT nb;P GP T 1 ; 48
the system, Eqs. (42) and (43) are coupled with the total nb
molar conservation equation. Assuming that the solid
phase does not ow, the total molar conservation equation where refers to the iteration counter and Cpo,nb = oHo,nb/
becomes oT, Cps,nb = oHs,nb/oT. The main steps of the algorithm
are:
oc
r  S o co vo 0: 44
ot 1. For given velocities, densities, and saturation compute
the coecients ao,P,as,P,Fnb and Dnb.
A nite-volume method is used to discretize Eq. (44). The
2. For given pressure, temperature and composition apply
scalar variables So, co and c are approximated over the
the PR-EOS and calculate the derivatives oHo/oT and
same control volumes as the pressure. The spatial and tem-
oHs/oT.
poral discretization of Eq. (44) can be written in the general
3. Solve the linear system given in Eq. (47) and correct the
form as
temperature from T = T1 + DT.
X X cn  cn1 4. Repeat steps 13 until the convergence criterion
anb nb 
bnb nb
 z vz r vr : 45
kDTk < TOL is satised.
nb nb
Dt
 R. Banki et al. / International Journal of Heat and Mass Transfer 51 (2008) 33873398 3395

4.3. Solution of the species balance equations 7. Normalize the oil mole fraction xo,i.
8. Check if the convergence criterion kxo  xold
o k < TOL is
Neglecting the velocity and diusion in the solid phase satised. If not, set xold
o xo and repeat steps 28.
and using Eq. (31), the species balance can be written as
o The computational speed of each SSI iteration is quite
czi r  cvo zi r  S o J o;i r  S s cs xs;i vo ; fast but generally it requires more iterations than the NR
ot
method. In the NR algorithm, we dene the residual func-
i 1; . . . ; n: 49
tion gi from (see Eq. (38)),
The nite-volume scheme is also used for the spatial dis- gi xo  fo;i p; T ; xo  fs;i p; T 0; i n  ns 1; . . . ; n:
cretization of the species balance equations. The convec-
tion term in the left-hand side of Eq. (49) is treated 50
implicitly in time with a rst-order upwinding scheme. The composition of the precipitating components is then
The terms on the right-hand side are calculated explicitly. computed by solving:
JDxo G; 51
4.4. Solution of the phase-equilibria model
where
The stability analysis can be readily performed by using  
ogi
the PR-EOS and the expression of the fugacity of the wax J ; Dxo xi inns 1;...;n ; and G
oxo;j i;jnns 1;...;n
phase given in Eq. (41). The phase-split calculation can be
performed eciently by using the successive-substitution- gi inns 1;...;n :
iteration (SSI) or the NR methods. Here, we provide the
algorithms for the phase-split calculation by both methods. Note that the size of the linear system is equal to the num-
The governing system has (n + ns) equations given by ber of precipitating components. The mole fractions of the
Eqs. (38)(40) and (n + ns) unknowns, which are the mole oil phase for the precipitating components are then up-
fraction of the oil phase (xo,i; i = 1, . . . , n) and the number dated from
of mole of the precipitated wax components (Ns,i;
xo;i xold
o;i Dxo;i ; i n  ns 1; . . . ; n: 52
j = (n  ns + 1), . . . , n).
The main steps of the SSI algorithm are as follows: The NR algorithm is similar to SSI algorithm described
above and is implemented by replacing the update of the
1. Guess an initial oil mole fraction xold o . If no information composition of the precipitating components in Eq. (52)
is available, take the oil mole fraction to be the same as by step 3 in the SSI algorithm. The other steps are exactly
the feed mole fraction (i.e., xold
o z). the same.
2. Calculate the fugacity fo;i p; T ; xold
o from the PR-EOS. In all our calculations, even very close to phase
3. Update the oil mole fraction of the precipitating compo- boundaries, the SSI method performed very well. Unlike
nents from (see Eq. (38)): vapor-liquid equilibria calculations, where close to phase
fs;i p; T boundaries the SSI algorithm becomes inecient, there
xo;i xold
o;i ; i n  ns 1; . . . ; n: was no convergence problem with the SSI algorithm in
fo;i p; T ; xold
o
the entire saturation range in wax precipitation
4. Dene Ns from calculations.
X
n
N s;j
Ns :
jnns 1
F 5. General algorithm

Ns can be calculated by summing Eq. (39) for The main steps of the algorithm for predicting the wax
i = (n  ns + 1), . . . ,n, that is, deposition in pipelines are the following:
Zs  X o
Ns ; Initialize the uid and pipe parameters.
1  Xo
P P Iterate over the time step (outer loop).
where X o ninns 1 xo;i and Z s ninns 1 zi . 1. Update the diusion coecients and the uid
5. Use Eq. (39) to calculate the number of moles of the pre- viscosity.
cipitating components 2. Iterate until convergence (inner loop).
N s;i zi  xo;i 1  N s ; i n  ns 1; . . . ; n:
2.1 Solve the momentum and the total mass balance
6. Use Eq. (40) to calculate the oil mole fractions of the equations to approximate the pressure and
non-precipitating components velocity eld.
zi 2.2 Solve the energy equation to calculate the
xo;i ; i 1; . . . ; n  ns :
1  Ns temperature.
3396 R. Banki et al. / International Journal of Heat and Mass Transfer 51 (2008) 33873398

2.3 Solve the species balance equations to calculate gel layer, the source term increases gradually with a
the overall composition. CarmanKoseny-type function as the solid saturation
2.4 Set cold = c, where c is the overall density. increases. The velocity eld is inversely proportional
2.5 Apply the phase-equilibria model to calculate to source function. The increase in the source term
phase compositions, saturation, and oil, wax leads to a decrease in the velocity eld. The energy
and overall densities. equation is written in terms of the liquid and solid
2.6 Check the convergence criterion enthalpies.
kcold  ck < TOL. If no convergence, repeat
steps 2.12.6. A detailed numerical algorithm to solve the mathemati-
Repeat until the assigned time is reached. cal model is also provided. The momentum equation cou-
pled with the total mass balance equation is solved by the
The uid viscosity is a function of temperature, pressure SIMPLER method to approximate the pressure and veloc-
and composition. It is calculated from the correlation of ity eld. The energy equation is discretized by a nite vol-
Lohrentz et al. [45]. The diusion coecients and the vis- ume scheme and linearized by the NR method. The liquid
cosity are updated outside the inner loop for their weak enthalpy is related to temperature using the PR-EOS. A
dependency on the temperature and pressure variation dur- robust algorithm for solidliquid equilibria is presented.
ing the convergence process of the inner loop. The algorithm uses the SSI method or the NR method.
With the NR method, the size of the linear system is the
6. Summary and concluding remarks number of precipitating components which makes the algo-
rithm very ecient.
A mathematical model for wax deposition in pipelines
for a multicomponent uid is presented. The model couples
the momentum, energy and species balance equations and Acknowledgements
a multisolid-wax precipitation model. There are two main
features that characterize our model: The funding for this work was provided by the member
companies of the Reservoir Engineering Research Institute
1. We include molecular diusion and thermal diusion (RERI). We thank Daniel Rosner and Alana Leahy-Dios
in our formulation. Both aect the ux of species of Yale University for reading the paper and their
towards the pipe wall by diusion. The driving force comments.
for molecular diusion is the concentration gradient.
The driving force for thermal diusion is temperature Appendix A
gradient. These two gradients may not be proportional
in the gel layer. The temperature gradient is expected The energy balance equation (Eq. (46)), in cylindrical
to be high close to the pipe wall. The concentration coordinates, is written as
gradient, because of phase change, is expected to be !
high close to the interface of the single liquid phase o X o 1 o
S j cj H j J z J r 0; A:1
region and the gel layer. Furthermore the temperature ot jo;s oz r or
gradient close to the wall may reach a pseudo steady
state quickly and stay large close to the wall. This where
may not be true for the concentration gradients. All  
these aspects may lead to incorrect results by the com- oT
J z S o co H o vo;z  k eff and
monly accepted use of the chain rule to replace con- oz
centration gradient by temperature gradient in the  
oT
Ficks law. In Hoteit et al. [32], the numerical results J r r S o co H o vo;r  k eff :
or
will demonstrate the importance of the use of the
proper diusion ux expression and the need for con- Let P be a control volume with edges e, w, n and s, and
sistency in diusion ux expression from irreversible neighboring control volumes E, W, N, and S. Fig. A1 illus-
thermodynamics. With consistent formulation the trates the ve point grid cluster at P. For the sake of sim-
model is predictive. plicity, we suppose that the domain is uniformly discretized
2. An enthalpyporosity approach is used to account for with space steps Dz and Dr along the z- and r-directions,
the ow deceleration in the gel layer where it is mod- respectively (Fig. A1).
eled as a pseudo-porous medium. A Darcy-type Integrating Eq. (A.1) over the control volume P, we get
source term is added to the momentum equations. the semi-discretized equation:
This source term is a function of the porosity (oil sat-
!
uration) in the gel layer. In the single-phase liquid o X
region, this function is set to zero and the velocity S j cj H j DzDr J e  J w J n  J s 0: A:2
ot jo;s
is calculated from the momentum equation. In the P
 R. Banki et al. / International Journal of Heat and Mass Transfer 51 (2008) 33873398 3397

N (A.3) at the current inner-loop step k + 1 and the current

time step n + 1 is written as
n
n1;k n1;k1 n1;k n1;k1
ao;P H o;P as;P H s;P  bnP
W w P e E X n1;k n1;k1 X n1;k n1;k1
r
F nb H o;nb Dnb T nb ; A:4
nb nb
s
where, aj;P DzDr=DtS j cj P ; j o; s, and bnP DzDr=Dt
S

P n
jo;s S j cj H j . The unknowns in Eq. (A.4) are the tem-
P
r z perature and enthalpy variables. The coecients F nb n1;k
,
z n1;k n1;k n1;k n
Dnb , ao;P , as;P , and bP are calculated from previous
Fig. A1. Grid point cluster at location P. Staggered discretization method iterations. For the sake of clarity, we drop the indicators
to implement SIMPLER algorithm. Scalar variables are presented at the k and n and dene the function GP in terms of the temper-
center of the control volume. Vector variables are located at the surface of ature and enthalpy variables:
control volume.
X
GP T  ao;P H o;P as;P H s;P  F nb H nb
nb
The uxes Je, Jw, Jn, and Js in Eq. (A.2) are dened by X
 Dnb T nb  bP 0: A:5
J e Fe e H o;e  De T E  T P ; J w Fe w H o;w  D w T P  T W ; nb

J n Fe n H o;n  Dn T N  T P ; J s Fe s H o;s  Ds T P  T S ;
References
where
( [1] D. Shock, J.D. Sudburg, J.J. Crockett, Studies of the mechanism of
Fe nb S o co vo;z nb Dr paran deposition and its control, J. Pet. Tech. 7 (9) (1955) 2330.
for nb w; e; and [2] R.M. Jorda, Paran deposition and prevention in oil wells, J. Pet.
Dnb k eff nb Dr
Dz Tech. 237 (1966) 16051612.
( [3] C. Narvaez, A.A. Ferrer, S.A. Corpoven, Prevention of paran well
Fe nb rrnbP S o co vo;z nb Dz plugging by plunger-lift use, SPE 21640 presented at the SPE
for nb n; s:
Dnb rrnbP k eff nb Dr
Dz production operation symposium, Oklahoma City, Oklahoma, 79
April, 1991.
Dierent schemes, such as upstream, hybrid, and power law [4] J. Svetgo, Paran problems can be resolved with chemicals, Oil Gas
can be used to dene the variables at the interfaces between J. 82 (9) (1984) 7982.
[5] B.J. Eastund, K.J. Schmitt, D.L. Meek, D.C. Anderson, G. Grisham,
the blocks [43]. By using the upstream scheme, the semi- New system stops paran buildup, Pet. Eng. Int. 61 (1) (1989) 4651.
discretized Eq. (A.2) is written in the general form: [6] C. Lira-Galeana, A. Firoozabadi, J.M. Prausnitz, Thermodynamics
! of wax precipitation in petroleum mixtures, AIChE J. 42 (1) (1996)
o X X X
239248.
S j cj H j DzDr F nb H o;nb Dnb T nb ;
ot jo;s nb nb
[7] P.A. Bern, V.R. Withers, J.R. Cairns, Wax deposition in crude oil
P pipelines, Proc. Euro. Oshore Petro. Conf. Exhib. Paper EUR 206.
A:3 October 2124, London, England, 1980.
[8] E.D. Burger, T.K. Perkins, J.H. Striegler, Studies of wax deposition
The counter nb in the above equation refers to the center in the Trans-Alaska Pipeline, J. Pet. Tech. 33 (6) (1981) 10751086.
and edge points P, e, w, n and s. The coecient Fnb and [9] A. Majeed, B. Bringedal, S. Overa, Model calculates wax deposition
Dnb are dened as for N. Sea oils, Oil Gas J. 88 (25) (1990) 6369.
[10] T.S. Brown, V.G. Niesen, D.D. Erickson, Measurement and predic-
F e max Fe e ; 0; F w max Fe w ; 0; F n max Fe n ; 0; tion of the kinetics of paran deposition, J. Pet. Tech. 47 (4) (1995)
X 328329.
F s max Fe s ; 0; FP  F nb and [11] J.A. Svendsen, Mathematical modeling of wax deposition in oil
nbn;w;e;s pipeline systems, AIChE J. 39 (8) (1993) 13771388.
X
DP  Dnb : [12] A.A. Hamouda, J.M. Ravnoy, Prediction of wax deposition in
pipelines and eld experience on the inuence of wax on drag-reducer
nbn;w;e;s
performance, in: Oshore Tech. Conf., OTC 7060 24th Annual OTC
in Houston, Texas, May 47, 1992.
As presented in the main text, there are two iterative pro- [13] F.S. Ribero, P.R.S. Mends, S.L. Braga, Obstruction of pipelines due
cedures in the algorithm. In the inner loop, we solve the to paran deposition during the ow of the crude oil, Int. J. Heat
nonlinear system by coupling the momentum, energy and Mass Transfer 40 (18) (1997) 43194328.
species balance equations. In the outer loop, we iterate [14] J.L. Creek, B. Matzain, M. Apte, M. Volk, E. Delle Case, H. Lund,
Mechanisms for wax deposition, AIChE Spring National Meeting,
on the time step. Let k and n denote the inner and outer
Houston, TX, March 1999.
loops, respectively. By using the backward-Euler scheme [15] P. Singh, H.S. Fogler, R. Venkatsean, N. Nagarajan, Formation and
to discretized the time operator in semi-discretized Eq. aging of incipient thin lm waxoil gels, AIChE J. 46 (5) (2000) 1059
(A.3), the spatial and temporal discretized form of Eq. 1074.
3398 R. Banki et al. / International Journal of Heat and Mass Transfer 51 (2008) 33873398
[16] E. Ramirez-Jaramillo, C. Lira-Galeana, O. Manero, Modeling wax
deposition in pipelines, Petroleum Sci. Technol. 22 (78) (2004) 821
861.
[17] M. Lacroix, V.R. Voller, Finite dierence solutions of solidication
phase change problems: transformed versus xed grids, Numer. Heat
Mass Transfer Part B Fundamentals 17 (1) (1990) 2541.
[18] A.D. Brent, V.R. Voller, K.J. Reid, Enthalpy porosity technique for
modeling convectiondiusion phase-change: application to the
melting of pure metal, Numer. Heat Transfer 13 (3) (1988) 297318.
[19] K. Fukui, K. Meada, Numerical simulation of dynamic layer
solidication for a eutectic binary system, J. Chem. Eng. Jpn. 31 (3)
(1998) 445450.
[20] M. Mbaye, E. Bilgen, Phase change process by natural convection
diusion in rectangular enclosures, Heat Mass Transfer 37 (1) (2001)
3542.
[21] V.R. Voller, C. Prakash, A xed grid numerical modeling method-
ology for convectiondiusion mushy region phase-change problems,
J. Heat Mass Transfer 30 (18) (1987) 17091719.
[22] C.H.S. Lira, Solidication in square section, Theoria 10 (2001) 4756.
[23] G.A. Holder, J. Winkler, Wax crystallization from distillate fuels: I.
cloud and pour phenomena exhibited by solutions of binary n-
paran mixtures, J. Inst. Petrol. 51 (499) (1965) 228235.
[24] G.A. Holder, J. Winkler, Wax crystallization from distillate fuels.
Part II: mechanisms of pour depression, J. Inst. Petrol. 51 (499) (1965)
235243.
[25] A.A. Hamouda, S. Davidsen, An approach for simulation of paran
deposition in pipelines as a function of ow characteristics with
reference to Tesside oil pipeline, in: SPE 28966-MS, Proc. SPE Int.
Symp. on Oileld Chemistry, San Antonio, TX, February 1417, 1995.
[26] G.M. Elphingston, K.L. Greenhill, J.J.C. Hsu, Modeling of multi-
phase wax deposition, J. Energy Res. Tech. 121 (2) (1999) 8185.
[27] P. Singh, H.S. Fogler, R. Venkatesan, Morphological evolution of
thick wax deposits during aging, AIChE J. 47 (1) (2001) 618.
[28] W.M. Deen, Analysis of Transport Phenomena, Oxford University
Press, New York, 1998.
[29] D.E. Rosner, Transport Processes in Chemically Reacting Flow
Systems, Courier Dover Publications, 2000.
[30] P.C. Carman, Fluid ow through granular beds, Trans. Inst. Chem.
Eng. 15a (1937) 150166.
[31] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena,
second ed., John Wiley, 2002.
[32] H. Hoteit, R. Banki, A. Firoozabadi, Deposition and aging in ow
lines from irreversible thermodynamics, submitted for publication.
[33] H.S. Carslaw, J.C. Jaeger, Conduction of Heat in Solids, second ed.,
Oxford University Press, New York, 1959.
[34] S. Whitaker, The Method of Volume Averaging, Kluwer Academic,
Springer, Boston, 1999.
[35] A. Firoozabadi, Thermodynamics of Hydrocarbon Reservoirs,
McGraw-Hill, New York, 1999.
[36] C.A. Passut, R.P. Danner, Heat capacity and entropy, Ind. Eng.
Chem. Process Des. Develop. 11 (4) (1972) 543.
[37] P. Singh, H.S. Fogler, N. Nagarajan, Prediction of the wax content of
the incipient waxoil gel in a owloop: An application of the
controlled-stress rheometer, J. Rheol. 43 (6) (1999) 14371459.
[38] K. Ghorayeb, A. Firoozabadi, Molecular pressure and thermal
diusion in nonideal multicomponent mixtures, AIChE J. 46 (5)
(2000) 883891.
[39] A. Firoozabadi, K. Ghorayeb, K. Shukla, Theoretical model of
thermal diusion factors in multicomponent mixtures, AIChE J. 46
(5) (2000) 892900.
[40] A. Leahy-Dios, M.M. Bou-Ali, J.K. Platten, A. Firoozabadi,
Measurements of molecular and thermal diusion coecients in
ternary mixtures, J. Chem. Phys. 122 (23) (2005) 112.
[41] D. Nichita, L. Goual, A. Firoozabadi, Wax precipitation in gas
condensate systems, SPE Prod. Facilit. 16 (4) (2001) 250259.
[42] R.A. Heidemann, J. Madsen, E.H. Stenby, S.I. Andersen, Wax
precipitation modeled with many mixed solid phases, AIChE J. 51 (1)
(2005) 298308.
[43] S.V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere
Publishing Corp., Washington, 1980.
[44] H.K. Versteeg, W. Malalasekera, An Introduction to Computational
Fluid Dynamics, Pearson Prentice Hall, Essex, England, 1995.
[45] J. Lohrentz, B. Bray, R. Clark, Calculating viscosities of reservoir
uids from their compositions, J. Pet. Tech. 16 (1964) 11711176.