Sie sind auf Seite 1von 20

Article

Cite This: Ind. Eng. Chem. Res. 2019, 58, 21797−21816 pubs.acs.org/IECR

Fundamental Study of Wax Deposition in Crude Oil Flows in a


Pipeline via Interface-Resolved Numerical Simulations
Mirco Magnini*,† and Omar K. Matar‡

Department of Mechanical, Materials and Manufacturing Engineering, University of Nottingham, Nottingham NG7 2RD, U.K.

Department of Chemical Engineering, Imperial College London, London SW7 2AZ, U.K.
*
S Supporting Information

ABSTRACT: This work presents a fundamental analysis of the


See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

mechanisms governing wax deposition and removal in crude oil


transportation pipelines. We utilize a numerical framework where oil
and deposit are treated as two immiscible phases, and the volume-of-
fluid (VOF) method is adopted to resolve the unsteady dynamics of the
Downloaded via UNIV PARIS-SUD on July 18, 2020 at 04:48:00 (UTC).

free interface. Deposition is modeled locally at the oil−deposit interface


via a chemical equilibria model, here adapted to the VOF method.
Deposit ageing is included via a thixotropic rheological model. The
results emphasize that the deposit pattern may appear as a uniform axisymmetric film covering the pipe wall or be completely
stratified. Although different mechanisms of deposit mobilization may occur, the removal rates correlate well with the Reynolds
number of the bulk flow and the viscosity of the deposit layer. The simulation data are used to benchmark closure laws for the
velocity and temperature within the film, and a prediction method for the steady-state deposit thickness is proposed.

1. INTRODUCTION millions of dollars,11 and therefore it is imperative to predict


Wax deposition represents a significant flow assurance problem the deposition behavior accurately in order to efficiently
for the oil and gas industry because it can lead to restriction schedule the pipeline maintenance.
and then blockage of oil reservoir formations, pipelines, and A number of mechanisms have been suggested to be
process equipment.1−3 Crude oils from extraction wells are responsible for the deposition of solids from waxy crude oil
complex mixtures of hydrocarbons consisting of paraffins, mixtures, including molecular diffusion, Soret diffusion,
aromatics, naphthenes, resins, and asphaltenes. Waxes are Brownian motion, shear dispersion, and gravity settling; the
identified as long-chain high-molecular weight paraffins with interested reader can refer to the comprehensive reviews of
carbon numbers ranging from 18 to 65.4 At the typical Burger et al.,12 Azevedo and Teixeira,5 and Aiyejina et al.2
reservoir temperatures (T ≥ 70 °C) and pressures Based on field data from the Trans Alaska Pipeline System,
(p ≥ 50 MPa), their solubility is sufficiently high to keep Burger et al.12 concluded that molecular diffusion dominates at
them fully dissolved in the mixture.1 However, as the crude oil the higher temperature (T ≈ 300 K) and heat flux conditions,
leaves the reservoir and flows through a transportation and the majority of researchers adopted it as the only
pipeline, its temperature decreases because of the cooler deposition mechanism.13−16 The molecular diffusion approach
environment. The highest temperature at which the first wax is based on the premise that a radial temperature gradient
crystals begin appearing when the oil is cooled is defined as the exists at the wall of the cooled pipe. If the pipe wall is below
wax appearance temperature (WAT) of the mixture. There- the WAT, the oil will contain precipitated solid particles and
fore, if the temperature of the pipeline falls below the WAT, the liquid phase will be in equilibrium with the solid phase.
wax precipitates and deposits along the inner walls of the Because wax solubility decreases with temperature, there will
pipeline. This becomes more severe in subsea pipelines, where be also a wall concentration gradient as a result of the
the water temperature can reach values as low as 5 °C.5 temperature gradient, and the cooler wall will have the lowest
The deposit layer grows and hardens with time as a wax−oil concentration of wax in the liquid phase. This concentration
gel containing a significant amount of oil trapped in a matrix of gradient drives the molecular diffusion of wax from the bulk
wax crystals. This leads to a series of significant operational fluid to the pipe wall and the rate of deposition is given by the
problems, from pipeline restriction, increased pumping power Fick’s law, j = Γ ∂c/∂n, where Γ is the diffusivity of wax
or reduced production, to pipeline blockage, replacement or molecules in the oil and ∂c/∂n is the wax concentration
permanent shutdown. Several remediation strategies have been gradient at the wall. Burger et al.12 suggested calculating ∂c/∂n
developed during the years; these include mechanical removal
of the deposit via pigging operation, inductive heating of the Received: September 20, 2019
pipeline,6 and use of chemical wax inhibitors7,8 and paraffin- Revised: October 27, 2019
degrading bacteria.9,10 Deferred production and implementa- Accepted: October 30, 2019
tion of adequate remediation strategies can cost tens of Published: October 30, 2019

© 2019 American Chemical Society 21797 DOI: 10.1021/acs.iecr.9b05250


Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

by assuming thermodynamic equilibrium between precipitating for example, as a constant deposit sink,28 as a term
wax crystals and liquid, such that ∂c/∂n = (dc/dT)(∂T/∂n). proportional to wall shear and deposit thickness,29−31 and as
Hence, the molecular diffusion approach relates the rate of wax a decreasing function of time to account for the deposit
deposition to the wall temperature gradient, which depends on ageing.32 A different hydrodynamic method based on an
the velocity and temperature fields within the pipeline. enthalpy-porosity formulation was devised by Banki et al.33
Laboratory flow loop experiments represent an ideal tool for and recently used by Albagli et al.34 to study the influence of
the fundamental investigation of wax deposition under the Reynolds number on wax deposition. In this method, the
different flow and heat transfer conditions.17 Singh et al.1 deposit layer at the pipe wall was treated as a porous medium
employed a model wax−oil system (an oil solvent and Mobil and the momentum equation, solved throughout the flow
M140 food grade wax with 0.67% mass fraction) in a flow loop domain, included source terms to approximate Darcy’s law in
where the test section was cooled by a heat exchange jacket. the growing gel layer. The alternative heat transfer
This experiment showed that 2−3 days were necessary to approach19,35 is based on the assumption that the liquid−
achieve a constant deposit thickness, while the wax content of deposit interface remains at the WAT throughout the
the deposit was still increasing after 5 days. The steady deposit deposition process, and the deposit thickness is evaluated by
thickness decreased with an increase of the mixture flow rate solving a steady-state heat transfer problem. Deposit removal
(laminar flow) or an increase of the wall temperature. Analyses does not need to be included in this model approach, but the
of wax−oil gel samples showed that the mass of heavy effect of shear on the ageing of the deposit has been introduced
hydrocarbons (carbon number above 29) increased with time, via a cubical-cage model of the wax crystal structure,36,37 which
while lighter components diffused out of the gel, as a impacts the deposit composition but not its thickness.
counterdiffusion phenomenon giving rise to ageing (i.e., In summary, while the various modeling approaches
hardening) of the deposit. Fong and Mehrotra,18 Bidmus substantially agree on the mechanisms responsible for wax
and Mehrotra,19 and Parthasarathi and Mehrotra20 tested a deposition, there is still a large degree of uncertainty and
different prepared mixture (Norpar13-Parowax) with much empiricism on the mechanisms and models for deposit
larger wax concentrations (from 3 to 20%), which led to removal. Deposit removal strongly depends on the hydro-
significantly faster deposition processes, with thermal steady- dynamic and thermal interactions between the bulk oil flow
state reached after about 30 min. The hot flow cases18,20 and the gel-like layer adhering to the pipe wall, and on the
(mixture temperature Tc above TWAT) indicated that the dynamics of the interface separating the two phases. The
deposit mass was lower for higher wall and mixture objective of the present work is to perform a fundamental
temperatures and lower wax concentrations, while turbulent analysis of wax deposition and removal in a crude oil flow
flow conditions yielded considerably less deposition than within a pipe by means of the numerical solution of the
laminar flow. Significantly different trends have been reported governing equations of mass, momentum, energy, and species
for cold flow conditions (Tc < TWAT) by Bidmus and conservation, coupled with an interface capturing technique38
Mehrotra,19 with the deposit rate increasing as the mixture that allows the time-dependent dynamics of the oil−deposit
temperature approached WAT and a negligible impact of the interface to be resolved. Although the use of interface
mixture flow rate. Wax deposition from South Pelto crude oil capturing methods to model liquid−vapor flows under
and Garden Banks condensate was studied in a flow loop evaporation/condensation conditions is well established,39
apparatus by Panacharoensawad and Sarica21 and Singh et al.22 the present work represents a first-of-a-kind attempt to adapt
These studies revealed much slower deposition processes such methods to the flow of complex fluid mixtures with phase
compared to the data from Mehrotra and co-workers, with the separation and deposition.
deposit thickness still increasing even after 48 h because of We employ a fluid model where liquid-like crude oil and
considerably smaller wax content (<1%) in the crude oils. The solid-like deposit are treated as two immiscible phases
deposit thickness versus time curves for different oil flow rates separated by an interface, and each phase is considered to be
were observed to intersect each other at a certain time, as the a bi-component mixture of wax and non-wax species. The rate
deposit thickness increased with the flow rate at early of deposition is evaluated through a chemical-equilibria
deposition times and vice versa at later stages. A number of model,13 here adapted for interface capturing algorithms, and
other flow loop deposition studies have been conducted,23−26 it depends on the local temperature and mass fractions at the
and the impact of flow rate, operating temperatures, and oil interface. Temperature, phase distribution, components mass
composition on the deposition trends substantially agrees with fractions as well as pressure and velocity of the fluid are
those outlined above. obtained via solution of the governing equations in a three-
Many different models to predict the build-up of the deposit dimensional model of the pipe geometry. A model for
layer on the cold wall of a pipeline have been proposed; these thixotropic fluids is utilized to account for the ageing of the
can be grouped into two main families, that is, hydrodynamic deposit with time. We carry out a parametric study to elucidate
and heat transfer models. Hydrodynamic models assume the effect of the oil flow rate, pipe diameter, surface tension,
molecular diffusion as the dominant deposition mechanism deposit rheology, and temperature on the deposit formation
and extract the radial temperature gradient which drives and removal processes. The simulations show that the deposit
deposition from hydrodynamic calculations. Early examples are may appear as a uniform axisymmetric film covering the pipe
́
the models of Singh et al.,1 Svendsen,13 Ramirez-Jaramillo et wall or be completely stratified, depending on the relative
27
al., where the space-time dependent wall temperature importance of the gravitational forces. Deposit removal occurs
gradient is obtained via analytical or numerical solution of mainly via interfacial shear, although entrainment may play a
the energy equation governing the oil flow in the pipe, coupled role when surface tension is low. The amount of deposit in the
with closure laws for the velocity profile in the oil. The deposit pipe (relative to the pipe volume) decreases with an increase of
layer is assumed to be immobile, and the mechanism of shear the oil flow rate and an increase of the pipe diameter, with the
removal has been later implemented through empirical models, Reynolds number correlating well the steady-state film
21798 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

thickness values. The viscosity and rheological properties of Computational cells with Fc = 1 (i.e., Fd = 0) are filled with the
the gel-like layer have a strong impact on the deposit mobility. crude oil, cells where Fc = 0 (Fd = 1) are filled with the deposit,
The rest of this paper is organized as follows. In Section 2, whereas cells where 0 < Fc, Fd < 1 are cells cut by the interface.
we present the details of the numerical framework; the results In each cell of the domain, the properties of this mixture fluid
of a representative simulation of flow with deposition are can be computed as an average over the two phases, weighted
introduced in Section 3; in Section 4, the numerical model is by the respective volume fractions, for example, the density
validated by comparison with experimental wax deposition
data from the literature; the results of the parametric study are ρ = ρc Fc + ρd Fd (1)
presented in Section 5; in Section 6, the numerical database is
utilized to extract closure laws for the velocity and temperature where ρc and ρd denote the density of the crude oil and of the
in the film flow region and a prediction model for the steady- deposit, respectively.
state deposit thickness is proposed; concluding remarks are Crude oils are complex mixtures of several hydrocarbon
provided in Section 7. species. From a numerical perspective, it is impractical to track
the concentration of each single component as the flow evolves
2. NUMERICAL FRAMEWORK within a pipe. Experimental evidence suggests that high
molecular weight paraffin waxes are the major constituents of
2.1. Fluid Model. The main interest of the present work is solid deposits from crude oils1,22,40 because of their lower
to investigate the fluid mechanic and thermal interactions solubility in the liquid phase as the pipe wall temperature drops
between the crude oil that flows in the pipeline and the deposit due to the cold environment. At early stages, only a few
layer that is formed at the pipe wall as a result of wax percent of precipitated wax content in the deposit is sufficient
deposition. Hence, the temporal evolution of the interface to gel the fluid, with liquid oil entrapped in a matrix of wax
separating the crude oil and deposit plays a crucial role, and it crystals.1 Therefore, we adopt a pragmatic fluid model where
is imperative to model accurately the mass, momentum, and the crude oil is composed of two main pseudo hydrocarbon
energy exchanges occurring at the interface. Figure 1 shows a components: oil, which lumps low carbon number non-wax
components, with overall mass fraction Yc,o; dissolved wax,
which lumps all high carbon number species, of total mass
fraction Yc,w, with Yc,o + Yc,w = 1. Likewise, we consider a gel-
like deposit where both solid wax crystals and liquid oil
trapped in their network structure coexist, with mass fractions
Yd,w and Yd,o, respectively (Yd,o + Yd,w = 1). The densities
appearing in eq 1 are therefore calculated as

ij Y Yc,w yzz ij Y Yd,w yzz


ρc = jjjj zz ,
zz ρd = jjjj zz
zz
−1 −1

jρ jρ
k c,o c,w { k d,o d,w {
Figure 1. Schematic of crude oil entering at a temperature Tc within a c,o d,o
+ +
pipeline where the wall temperature Twall is below the wax appearance ρ ρ
temperature TWAT. A thermal boundary layer is formed at the pipe (2)
wall. The radial temperature gradient generates an outward directed
heat flux q. A local crude-to-deposit mass transfer of magnitude j is
with the subscripts c and d indicating crude oil and deposit
established and a deposit layer grows at the pipe wall. nwall and nint phases, respectively, while o and w denote the non-wax and
indicate, respectively, the normal vector to the wall and to the oil− wax pseudocomponents. Note that the mass fractions of each
deposit interface. The streamwise coordinate is denoted as z, while y pseudocomponent vary across the computational domain. The
indicates the radial distance from the pipe wall. four densities appearing in eq 2 are, in general, a function of
the local temperature.
In summary, the fluid model adopted makes use of two
schematic of the flow configuration under analysis. We treat variables to define the volume fraction of each phase within the
the crude oil and the deposit as two immiscible phases, and we computational cell (Fc and Fd) and four additional variables to
utilize a volume-of-fluid (VOF) method38 to capture the identify the mass fraction of each pseudo-component within
dynamics of the interface separating the phases, throughout the each phase (Yc,o, Yc,w, Yd,o, and Yd,w); see Table 1. These six
flow domain. The VOF method has the advantage of advecting variables (though only three are independent because Fc + Fd =
fluid interfaces while preserving mass to machine accuracy, as 1, Yc,o + Yc,w = 1 and Yd,o + Yd,w = 1) vary as a function of space
opposed to other well-established surface-tracking methods and time across the computational domain, and their values are
such as the level-set or front tracking, and hence this makes it calculated by the solution algorithm, as indicated in the next
particularly suitable for simulating flows with phase change. section.
Accordingly, the crude oil−deposit two-phase mixture is 2.2. Governing Equations. Within the VOF algorithm,
treated as a single mixture fluid; we define the fraction of the volume fraction of the primary phase, here the crude oil, is
the computational cell volume occupied by the crude oil as Fc transported as a passive tracer by the flow field and hence Fc is
and that occupied by the deposit as Fd, with Fc + Fd = 1. extracted as a solution of the following transport equation

Table 1. Fluid Model Utilized in This Work and Related Variables

21799 DOI: 10.1021/acs.iecr.9b05250


Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

∂(ρc Fc) κ = −∇·nint, where nint = ∇Fc/|∇Fc| is the local interface unit
+ ∇·(ρc Fc u) = −(jo + jw )δ normal vector (see Figure 1).
∂t (3)
The molecular viscosity of the mixture fluid μ is computed
where t indicates time. Because the VOF method adopts a as
single-fluid formulation of the two-phase flow, single velocity, μ = μc Fc + μd Fd (8)
pressure, and temperature field are shared among the phases
and u represents the mixture fluid velocity vector. The volume with
fraction of the deposit Fd is simply calculated as 1 − Fc. If mass
μc = μc,o Yc,o + μc,w Yc,w , μd = μd,o Yd,o + μd,w Yd,w
is exchanged between the crude oil and the deposit, the right-
hand side of eq 3 is not zero. Two crude-to-deposit mass (9)
transfer mechanisms are considered: (1) the deposition of the where μc,o, μc,w, μd,o, and μd,w are the dynamic viscosities of the
non-wax pseudocomponent, of magnitude jo (units kg/(m2 individual pseudocomponents. Waxy crude oils exhibit very
s)); (2) the deposition of the wax pseudocomponent, jw. The complex rheological behavior, from Newtonian fluid character-
mass transfer is considered positive when it occurs from the istics when T > TWAT to non-Newtonian behavior at lower
crude oil to the deposit, thus motivating the negative sign at temperatures, with orders of magnitude increase of the
the right-hand side of eq 3. The delta function appearing in eq viscosity.43,44 Also, the increase of the wax content of the
3 represents the interfacial area density, that is, the area of the gel-like deposit with time leads to hardening of the deposit,
crude−deposit interface within the computational cell divided also known as ageing.1 Despite the fact that the rheology of
by the cell volume and has units of 1/m. It is constructed to be crude oils below the WAT has been extensively studied45−47
non-zero only at the oil−deposit interface;39 details about δ and several constitutive equations to describe their non-
and the deposition model are provided in Section 2.3. Newtonian behavior have been proposed (e.g., Cross,48
The mass fractions of the non-wax pseudocomponents Casson,49 Herschel−Bulkley,50 and Coussot51 models), the
within the crude oil (Yc,o) and within the deposit (Yd,o) are values of the empirical constants present in these equations are
calculated by solving the following conservation equations generally not known for crude oils of industrial interest.
Furthermore, these laws cannot be applied in the present
∂(ρc FcYc,o)

ÄÅ ÉÑ
+ ∇·(ρc Fc uYc,o) context where the crude oil and the gel-like deposit are

ÅÅ i μt yz ÑÑ
j
∂t
Å Ñ
modeled as distinct phases and specific values of the deposit
j
= ∇·ÅÅÅFcjjρc Γc,o + z
zz∇Yc,oÑÑÑ − jo δ
viscosity μd are needed. In the absence of field data, in the
ÅÅ j z ÑÑ
ÅÇ k t{ ÑÖ
present work, we assume simple temperature-dependent values
Sc (4) of the individual viscosities appearing in eq 9. However, for a
few selected cases (see Section 5.5), we evaluate the effects of
∂(ρd FdYd,o) non-Newtonian fluid characteristics and deposit ageing by

ÄÅ ÉÑ
+ ∇·(ρd Fd uYd,o) adopting a rheological model based on thixotropic fluids.15,51
ÅÅ i μt yz ÑÑ
j
∂t
Å zz∇Y ÑÑÑ + j δ
The mixture temperature field T is obtained by solving the
= ∇·ÅÅÅFdjjjρd Γd,o + z ÑÑ o
ÅÅ j Sct z{ ÑÑÖ
energy transport equation
ÅÇ k
d,o
(5) ∂(ρcpT )

ÄÅ ÑÉÑ
+ ∇·(ρcp uT )
ÅÅ ij μt yz ÑÑ
∂t
ÅÅ
where Γc,o and Γd,o are diffusion constants. Turbulence is
j
= ∇·(λeff ∇T ) + ∇·ÅÅÅ∑ Hi , jFi jjρi Γi , j + zz∇Yi , j ÑÑÑÑ
z
ÅÅ i , j j z ÑÑ
modeled via a Reynolds-averaged Navier−Stokes (RANS)

ÅÅÇ k Sct { ÑÑÖ


method; see Section 2.4; μt is the turbulent viscosity and Sct =
μt /ρΓ t the turbulent Schmidt number (Γ t : turbulent
diffusivity), Sct = 0.7.41 The mass fractions of the wax
pseudocomponents are evaluated as Yc,w = 1 − Yc,o and Yd,w = 1 + (jo ΔHfus,o + jw ΔHfus,w )δ (10)
− Yd,o.
where cp is the constant pressure specific heat, λeff = λ + λt is
The mixture fluid velocity u and pressure field p are obtained
the effective thermal conductivity, while the sum at the right-
by solving the single-fluid continuity and momentum equations
hand side sums over the four pseudocomponents, i = c, d and j
with a projection method (see Section 2.6), respectively
= o, w; the specific enthalpy of the pseudospecies (i, j), Hi,j, for
∂ρ constant specific heat is calculated as Hi,j = cpi,j(T − Tref) where
+ ∇·(ρ u) = 0
∂t (6) Tref is a reference temperature value. The last term on the
right-hand side of eq 10 implements the enthalpy source due
and to the interfacial mass transfer and ΔHfus,j (j = o, w) denotes
∂(ρ u) the specific enthalpy of fusion of the lumped species. The
+ ∇·(ρ uu) specific heat and molecular thermal conductivity of the mixture
∂t fluid are evaluated based on the pseudocomponents specific
= −∇p + ∇·[μeff (∇u + (∇u)T )] + ρ g + σκ ∇Fc (7) values as done for the molecular viscosity; see eqs 8 and 9; the
turbulent thermal conductivity is computed as λt = cpμt/Prt,
where g denotes the gravity acceleration vector and μeff = μ + where Prt is the turbulent Prandtl number (Prt = 0.8541).
μt is the effective dynamic viscosity of the fluid. The last term 2.3. Deposition Model. Wax deposition is a complex
at the right-hand side of eq 7 implements the surface tension phenomenon which combines phase equilibria, phase tran-
force according to the continuum surface force algorithm,42 sition, thermodynamics, and heat and mass transfer processes.
with σ being the surface tension coefficient and κ being the We assume that deposition is mainly driven by molecular
local oil−deposit interface curvature; the latter is calculated as diffusion, and we implement a deposition model based on
21800 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

Figure 2. (a) Moles of liquid per mole of mixture Lx and (b) nondimensional liquid concentration derivative ηi vs temperature predicted by the
model. Conditions are from Svendsen:13 Mo = 215 kg/kmol, Mw = 530 kg/kmol, ΔHfus,o = 162.3 kJ/kg, ΔHfus,w = 203.4 kJ/kg, Tfus,o = 272 K, Tfus,w
= 341 K, K0,o = 1, K0,w = 1, Yc,o = 0.85, Yc,w = 0.15.

thermodynamic chemical equilibria, originally proposed by We show here the results obtained using the model above
Svendsen,13 to express the source terms jo and jw in the flow for a fluid with two pseudo hydrocarbon components; all the
equations above. Accordingly, the deposition rate of the i-th properties are from Svendsen:13 molecular weight Mo = 215
species (i = o, w for the bi-component mixture considered kg/kmol and Mw = 530 kg/kmol; enthalpy of fusion ΔHfus,o =
here) is calculated starting from the Fick’s law 162.3 kJ/kg and ΔHfus,w = 203.4 kJ/kg; fusion temperature
Tfus,o = 272 K and Tfus,w = 341 K; and constant mass fractions
dci ∂T of components at the liquid state Yc,o = 0.85 and Yc,w = 0.15.
ji = Γm
dT ∂n int (11) The number of moles of liquid per mole of mixture as a
function of temperature is reported in Figure 2a. It can be seen
where Γm is an average diffusion constant and ci is the mass that, as the temperature increases, the fraction of liquid in the
concentration of the component i at the liquid state in the mixture increases because the solubility of solid components
mixture, which under the assumption of thermodynamic grows. Eventually, when the mixture temperature rises above
equilibrium is only a function of the temperature. Following the value of T = 319 K, the mixture is all in the liquid state (Lx
Svendsen,13 the term dci/dT can be expressed as = 1), thus indicating that TWAT = 319 K. The resulting liquid
dci ρ concentration derivative ηi is plotted in Figure 2b; at lower
= ηi temperatures, T < 300 K, the non-wax component contributes
dT T (12)
more to the deposition rate as Ko ≪ Kw, thus making the
where ηi represents the dimensionless variation of the mass deposit richer in the light (low carbon number) species. At
concentration of species i at the liquid state in the mixture per higher temperatures, the situation is reversed and the deposit
unit temperature (ηi > 0) and has the following analytical layer is enriched with higher carbon number components. The

ÄÅ É
plots in Figure 2 match exactly those included in Svendsen13

M ΔH Ñ
expression
ÅÅ Ñ
ÅÅT (1 + ψ )2 ∂Lx + ψ i fus,i ÑÑÑK i
(Figures 3 and 4 therein), thus verifying the model

ÅÅÇ R gT Ñ ÑÖ
implementation.
∂T Note that the thermodynamic approach does not account
ηi = Yc, i 2 for the influence of the cooling rate on phase transition, which
(1 + ψK i) (13)
may decrease the onset temperature of crystallization
where Ki is the equilibrium constant between the i-th solid and significantly below the WAT,52 and thus represents a potential

ÅÄÅ ÑÉÑ
liquid phases direction for future developments of the present deposition

ÅÅ MiΔHfus, i ij T yzzÑÑÑ
Å j
model.

K i = K 0, i expÅÅÅ jj1 − zzÑÑ


ÅÅ R gT jk Tfus, i z{ÑÑÑ
In the numerical implementation, the spatially and

ÅÇ ÑÖ
temporally varying mixture fluid temperature and mass
(14) fractions of the two lumped hydrocarbon components,
extracted from the solution of the flow equations discussed
with K0,i being the liquid-to-solid activity coefficient ratio13 and in Section 2.2, are used as input parameters for the calculation
Mi being the molecular weight, while Rg and Tfus,i indicate, of the local deposition rates
respectively, the universal gas constant and the fusion
temperature of species i. In eq 13, Lx represents the number ρηo
of moles of liquid in 1 mol of fluid, and ψ is the ratio of solid to jo = Γm ∇T · n
T (15a)
liquid number of moles, ψ = (1 − Lx)/Lx. In the present case
where only two pseudocomponents are considered, Lx and ρηw
∂Lx/∂T have analytical expressions that only depend on Ki, jw = Γm ∇T · n
T (15b)
ΔHfus,i, Tfus,i, the number of moles of each species, and the local
fluid temperature. For the sake of brevity, these relationships When a deposit layer already exists at the pipe wall, n in eqs
are not reported here; the interested reader can refer to 15a and 15b refers to the unit normal vector to the crude−
Svendsen13 and Sileri et al.15 deposit interface, nint. In this configuration, the delta function
21801 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

appearing in eqs 3−5 and 10 is evaluated as δ = |∇Fc|, twice as large. Finally, the low-Re correction essentially damps
according to the VOF method. When the pipe wall is clean, n the turbulent viscosity depending on the local turbulence
refers to the wall normal vector, nwall. In this situation, the Reynolds number.58
volume fraction gradient is not representative of the oil− In summary, within the k−ω SST turbulence model with
deposit interface area density and the latter is estimated low-Re correction, the turbulent viscosity is calculated with the
geometrically as δ = Aint/Vcell; Aint denotes the interface area following expression

ÄÅ É
1 Ñ
which, under the assumption that the pipe wall is covered by
ω maxÅÅÅ ΩG , ÑÑ
ρk 1

ÅÅÇ a1ω α *Ñ
Ñ
an infinitesimally thin layer of deposit, is estimated as the area
ÑÖ
μt =
of the boundary face of the near-wall cell. Note that, if the
temperature gradient is positive when outward directed, that is, (16)
if the wall is warmer than the fluid, then eqs 15a and 15b yield where k and ω denote the turbulence kinetic energy and
a deposit-to-crude mass transfer, that is, the deposited mass specific dissipation rate, respectively. In eq 16, the term ρk/ω
dissolves back to the crude oil. derives from the original k−ω model; ΩG/(a1ω) implements
During the simulation, at each time instant, jo, jw, and δ are the SST stress limiter,54 where Ω is the absolute value of the
calculated for each computational cell of the domain based on vorticity, a1 is a constant and G is a function of the distance
the procedure outlined above, and their values are utilized to from the wall, ranging from 1 near the wall to 0 far from it; the
evaluate the deposition source terms appearing at the right- term 1/α* operates the turbulence damping based on the
hand sides of eqs 3−5 and 10. Hence, at each time instant, the values of the local turbulence Reynolds number Ret = ρk/μω.
mass of oil and wax pseudocomponents that pass from the oil The turbulence kinetic energy and specific dissipation rate
phase to the deposit phase in each cell is (joδ)VcellΔt and (jwδ) are obtained by solving two transport equations;54 all the
VcellΔt, respectively, with Vcell being the volume of the cell and original values of the constants have been retained.
Δt the time-step of the simulation. 2.5. Boundary Conditions. We simulate wax deposition
2.4. Turbulence Model. The two-phase flow of liquid oil within horizontal pipes of circular cross section, with diameter
and gel-like deposit at the wall can be essentially regarded as an D and length L. The pipe is modeled with a three-dimensional
annular flow, where a high-speed oil core drives by shear a film geometry, except for the validation case (see Section 4) where
of more viscous fluid at the wall. In the present work, we span a a two-dimensional axisymmetric configuration is preferred.
range of crude oil Reynolds numbers of Rec = ρcUcD/μc = 20− The computational mesh is structured in the 2D case and
15 000 (Uc is the average velocity of crude oil at the inlet and block-structured when 3D. Near-wall refined mesh elements
D is the pipe diameter), thus entering in the fully turbulent are employed to capture the boundary layer and the thin film
regime, whereas the flow in the thin and very viscous deposit dynamics.
layer is most likely laminar. Therefore, we need a turbulence The boundaries of the geometry model and respective
model with appropriate solution of the laminar−turbulent conditions consist of:
transition. Furthermore, the deposit thickness hd spans from
• Inlet: only crude oil enters the flow domain, thus Fc = 1
zero at the channel inlet to values of hd/D ≈ 0.01, which are on
at the inlet. A fixed value of the mass fraction of the non-
the order of the viscous sublayer thickness. This requires the
wax pseudo species Yc,o is set. A constant flow rate of
near-wall region to be accurately captured with the computa-
crude oil with average speed Uc is imposed. When Rec =
tional mesh down to the viscous sublayer, thus ruling out the
ρcUcD/μc ≤ 2000, a fully developed laminar profile is
use of wall functions.
set. When Rec > 2000, a series of preliminary single-
Large-eddy simulations (LES) are emerging as a powerful
phase adiabatic steady-state simulations are run with
tool to model turbulence in multiphase gas−liquid flows,53 but periodic inlet−outlet conditions to achieve a fully
LES of two-phase liquid−gel flows are far less established. developed turbulent velocity profile. The final velocity
Therefore, in this work, a RANS turbulence modeling and turbulence profiles become the inlet boundary
approach has been preferred, by adopting the k−ω shear− conditions for the successive unsteady, diabatic simu-
stress transport (SST) model originally developed by lations. For the pressure field, the pressure gradient in a
Menter,54 with a low Reynolds number correction. Among direction orthogonal to the boundary is set to zero. A
the several RANS models available in the literature, the k−ω55 constant inlet temperature is imposed.
is the model of choice for boundary-layer flows and it seems
appropriate in the present case, where strong momentum and • Pipe wall: a no-slip and no-penetration velocity
energy exchanges occur between the flow and the pipe wall. condition is set, together with a constant wall temper-
The SST version blends the original k−ω formulation in the ature Twall.
near-wall region to a high-Reynolds number k−ϵ model in the • Outlet: zero gradient conditions are set for the velocity,
outer part of the boundary layer and introduces a stress limiter temperature, volume fraction, and pseudo-species mass
in the expression of the turbulent viscosity, based on the fractions, while the pressure is given a reference p = 0
Bradshaw et al.56 assumption that the shear stress in a value.
boundary layer is proportional to the turbulent kinetic energy. Preliminary single-phase steady-state simulations are run to
The stress limiter plays a very important role as it reduces the obtain temporally developed fields of velocity and temperature
magnitude of the Reynolds shear stress near the wall, thus to be utilized as initial conditions for two-phase simulations.
decreasing the magnitude of the wall-fluid exchanges, Unless otherwise specified (see Section 4), the pipe is initially
compared to the standard k−ω model. Da Riva et al.57 utilized “clean”, that is, no deposit is present, and hence Fc = 1
the k−ω SST model to simulate annular flow condensation in everywhere at t = 0.
a minichannel and obtained predictions of the heat transfer 2.6. Discretization Methods. The entire numerical model
coefficient at the wall in excellent agreement with experiments, comprises seven scalar unknowns, that is, Fc, Yc,o, Yd,o, p, T, k,
whereas the standard k−ω model exhibited values that were and ω, and one vectorial unknown, u. The mathematical
21802 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

problem is closed by seven scalar equations, as shown in eqs the fluid has to travel less than one mesh cell during each time-
3−6, 10, and two equations for the turbulence quantities, plus step, which limits the time-step size to values on the order of
one vectorial equation, as shown in eq 7. The equations are 10−4 s. The direct numerical simulation of hours of wax
solved utilizing the commercial simulation package ANSYS deposition process would therefore require ∼109 time-steps,
Fluent v.18.1, which adopts a finite-volume discretization which is not possible with the available computational
method. The volume fraction equation is discretized in time resources. In order to achieve appreciable values of the
with a first-order explicit scheme, and the convective term is deposition rate within time scales that can be achieved with
computed through a geometrical reconstruction of the fluxes numerical simulations (≲102 s), the diffusion constant Γm
across the faces of each interfacial cell.59 All remaining driving the interfacial mass transfer has been increased by 3
equations are marched in time with a first-order implicit orders of magnitude compared to the original value used in
formulation. A variable time-step technique is chosen, with a Svendsen13 (Γm = 2.76 × 10−10 m2/s); see the last column of
maximum Courant number of 0.5; tests with smaller values did Table 2.
not exhibit appreciable differences in the two-phase flow The pipe is modeled using a 3D geometry. At t = 0, no
dynamics. The convective terms of mass fraction, velocity, and deposit is present. At the onset, wall deposition starts because
turbulence equations are discretized with a third-order of the radial temperature gradient and because the wall
MUSCL (Monotonic Upstream-centered Scheme for Con- temperature is below the WAT. A video of the simulation is
servation Laws)60 scheme, while all the diffusion terms are available in the Supporting Information. An illustrative image
discretized with a second-order central finite-difference of the flow within the pipe is provided as Abstract Graphic,
scheme. The convective term in the energy equation is while sequential images illustrating fluid and thermal dynamics
formulated with a power-law61 model. The PISO (Pressure are presented in Figure 3. The snapshots are taken at t = 60 s,
Implicit Splitting of Operators)62 algorithm is utilized to when the average thickness of the deposit layer is
obtain pressure and velocity fields via a coupled solution of the approximately constant and the flow can be regarded as
continuity and momentum equations, while a staggered-grid steady. A film of deposit covers the entire pipe wall and its
approach is used to obtain pressure values on the faces of the
perimetric distribution is approximately uniform, as gravita-
computational cells. At each time-step, the discretized
tional effects are minimal because of the small difference
equations are solved iteratively till the scaled residuals fall
between crude and deposit densities. Waves on the crude−
below a tolerance value of 10−4 (10−9 for the energy equation).
deposit interface begin forming when the deposit film becomes
The deposition model described in Section 2.3, the
temperature-dependent fluid properties, and run-time process- sufficiently thick and then travel downstream; waves are
ing algorithms are all implemented by means of user-defined continuously formed as time elapses (see snapshots in Figure
subroutines. 3), and they are not necessarily axisymmetric, but their
perimetral extension may be limited to only a portion of the
3. SIMULATION OF WAX DEPOSITION wall circumference. The maximum velocity of the flow at the
pipe centerline is on the order of 1 m/s, that is, about 4 times
In this section, we present the results of a representative larger than Uc at the inlet, because of the reduction in the pipe
numerical simulation to illustrate the capability of the cross section. The velocity of the flow in the deposit layer is
numerical model. The flow conditions are taken from the
much smaller, on the order of 1 mm/s, because of the higher
work of Svendsen:13 the channel is horizontal (g = 9.81 m/s2,
viscosity. Nonetheless, the flow within the travelling waves
acting downward), with a diameter of D = 9 mm (radius R =
exhibits a substantially larger speed depending on the height of
D/2) and length L = 40D; crude oil (Fc = 1) enters the pipe at
the wave, and it may reach values as high as 0.2 m/s, with the
a temperature of Tc = 303 K, while the wall is at Twall = 278 K;
waves “surfing” over an almost stagnant layer of very viscous
the mass fractions of the two pseudo hydrocarbon components
at the channel inlet are Yc,o = 0.85 and Yc,w = 0.15; the deposit. The reduction of the viscosity with increasing
Reynolds number of the flow based on the conditions at the temperature impacts the wave dynamics because on the top
inlet is Rec = 193, to which corresponds an average velocity of of the waves, the fluid is warmer and hence less viscous. The
the oil at the inlet of Uc = 0.22 m/s; the values of the molecular contours of the temperature field on the channel cross section
weights, fusion enthalpies, and temperatures of the two emphasize the presence of a large radial temperature gradient
pseudocomponents are specified in Section 2.3; the wax in the proximity of the pipe wall. The rate of mass deposition is
appearance temperature is TWAT = 319 K. non-zero only at the interface; it reaches higher values where
All the properties of the mixture fluid and its components the film is the thinnest because of the combined effect of larger
are reported in Table 2. The wax deposition model given by temperature gradients and lower temperatures, the latter
Svendsen13 did not need any property of the deposit other resulting in larger values of ηo (see Figure 2b).
than its density. We choose the specific heat and thermal Data extracted from the simulation are plotted as a function
conductivity of the deposit to be the same as that of the crude of time in Figure 4. The figure shows the temporal evolution of
oil. The gel-like deposit is arbitrarily set to be 100 times more the (a) deposit thickness, (b) deposition rate, (c) deposit mass,
viscous than the crude oil. The surface tension coefficient of and (d) pressure drop through the channel. As time elapses,
crude oil/asphaltene, σ = 0.04 N/m,63 is selected. The other the deposit layer grows because the rate of deposition exceeds
thermophysical properties of the pseudocomponents involved the rate of deposit removal by the bulk phase. When the
in the deposition model, that is, molecular mass, fusion deposit is sufficiently thick, such that deposition and removal
temperature, and enthalpy, are set as reported in Section 2.3. rates balance out, the average thickness of the deposit becomes
Svendsen’s original model is one-dimensional and was constant; this is achieved after about 40 s of simulation time
developed to predict the growth of the deposit layer after (Figure 4a). A theoretical value of the growth rate of the
several hours of operation. In the present numerical model, in deposit thickness at the beginning of the simulation can be
order to correctly capture the two-phase interface dynamics, estimated as
21803 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

Table 2. Fluid Properties Used in the Baseline Numerical Simulationa

a
Temperature is expressed in degrees kelvin. Γi,j (i = c, d; j = o, w) refer to the diffusion constants used in eqs 4, 5, and 10, while Γm indicates the
average diffusion constant driving deposition; see eqs 15a and 15b.

Figure 3. (a) Temporal evolution of the oil−deposit interface (Fc = 0.5 iso-contour, depicted in white) and velocity field, extracted on a vertical
centerline plane. Flow is from left to right. In order to aid visualization, the velocity fields within the oil and deposit phases have different color
scales. (b) Corresponding temperature field and interface profile extracted on the cross-section indicated in (a) with a vertical dashed red line. (c)
Total deposition rate, (jo + jw)δ, and interface profile on the same cross section as (b). From top to bottom, the images are 0.05 s apart. The
simulation conditions are as follows:13 horizontal pipe, D = 9 mm, Uc = 0.22 m/s (Rec = 193), g = 9.81 m/s2 (acting downward), Tc = 303 K, Twall =
278 K, TWAT = 319 K; at the inlet, Fc = 1, Yc,o = 0.85 and Yc,w = 0.15; the fluid properties are reported in Table 2.

21804 DOI: 10.1021/acs.iecr.9b05250


Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

Figure 4. Simulation results vs time: (a) average thickness of the deposit hd/R in the domain, evaluated as hd /R = 1 − 1 − Vd /(πR2L) , where
1
Vd is the deposit volume, Vd = ∫ VFd dV; (b) average deposition rate of each pseudo species ji ,ave = ∫ j
Sint Sint i
dSint , with Sint being the area of the
crude−deposit interface; (c) total mass of each pseudo species deposited in the domain, mi,tot = ∫ VFdρd,iYd,i dV; (d) pressure difference between
inlet and outlet sections. The inset in (a) shows a close-up of the deposit thickness at the initial time instants. In (d), the pressure drop is rescaled
by the single-phase value at t = 0, Δpsp. The simulation conditions are reported in the caption of Figure 3.

dhd j(t = 0+) 1 ρ(ηo + ηw ) ∂T for Γm = 2.76 × 10−9 m2/s, which approaches the experimental
(t = 0+) = = Γm values measured by Bidmus and Mehrotra19 for wax−solvent
dt ρd ρd Twall ∂n wall
mixtures (jave = 30−300 mg/(m2 s)). Last, Figure 4d indicates
(17) that the pressure necessary to maintain a constant flow rate of
where Γm, ρ, ρd, ηo, and ηw must be evaluated at the Twall, and crude oil within the pipe increases as the very viscous deposit
∂T/∂nwall is an average temperature gradient at the wall. Using layer grows, up to a value three times larger than that for the
the single-phase simulation data for ∂T/∂nwall, an initial growth clean pipe.
rate of (1/R)(dhd/dt) ≈ 0.03 1/s is predicted. This value
matches very well the numerical data, see the inset in Figure 4. VALIDATION
4a, and as the wax layer grows, its growth rate, dhd/dt, In order to validate the numerical model, we consider the wax
decreases because of an increase in the interfacial temperature deposition data given by Singh et al.22 The authors conducted
(because the oil−deposit interface moves away from the pipe wax deposition experiments using a Garden Banks condensate
wall) and to the onset of removal. At the initial time instants, as operating fluid, circulated in turbulent flow conditions in a
the deposition rate of non-wax hydrocarbons is orders of D = 16.5 mm internal diameter pipe. The flow loop featured a
magnitude larger than that of waxes (ηo = 8.2 while ηw = pipe-in-pipe heat exchanger where warm waxy oil transited in
0.008) as suggested by Figure 4b, and the deposit film is the inner pipe (at T < TWAT) and colder water flowed
enriched with light species; see Figure 4c. As the wax layer countercurrently in the outer pipe, thus generating an outward
thickens and the interfacial temperature grows, ηo drops while directed heat flux. Tests were performed for a duration of 24 h;
ηw increases (see Figure 2b) and the deposition rates stabilize the deposit thickness was indirectly calculated during the
at a steady situation where jo,ave ≈ 6.7jw,ave. These steady values experiment from pressure drop measurements, and these
of the deposition rates depend strongly on the film thickness, values verified versus direct deposit mass measurements
that is, on the balance between deposition and removal rates, performed after texp = 2.67, 8, 16, and 24 h. From a numerical
which determines the temperature of the crude−deposit perspective, it is not feasible to run unsteady multiphase
interface. In absolute units, at the steady-state, we obtain jave simulations for hours of simulation time. In order to overcome
≈ 10 g/(m2 s); simulations repeated with smaller values of Γm the difference between the time scales of the two-phase flow
yielded reduced deposition rates, down to jave ≈ 0.7 g/(m2 s) and of the deposition process, we perform simulations by
21805 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

patching an initial deposit layer along the entire pipe wall, with properties of the deposit, which are set equal to those of the
average thickness (and wax fraction) corresponding to the crude oil, except for the viscosity μd that will be the subject of a
experimental data measured at a certain texp. Then, we run the parametric analysis. Singh et al.22 reported that the deposit
unsteady simulation for a sufficiently long time so as to achieve critical carbon number of the operating fluid was ∼30 and
a meaningful value of the deposit growth rate dhd/dt, which considered the total fraction of alkanes with carbon number
can be compared with a benchmark value extracted from an above 30 (n-C30+ fraction) as the wax fraction of the fluid.
empirical correlation for hd(t) proposed by Singh et al.22 Accordingly, in the numerical model, we assume that the non-
The film thickness measured in the experiments is on the wax pseudo component is made of the n-C30− alkanes
order of hd/R = 0.02 and below; therefore, a very fine near-wall fraction, while wax is made of the remaining n-C30+ fraction.
mesh is required. In order to limit the computational effort, we Based on the Garden Banks composition reported in the
adopt a 2D axisymmetric geometry. The domain mesh, Appendix of Singh et al.,22 we evaluated the molecular mass,
depicted in Figure 5, employs near-wall cell refinement with fusion temperature, and enthalpy of each lumped component
as a weighted average of the values of the individual
hydrocarbons. The resulting values, introduced within the
deposition model, are as follows: Mo = 213 kg/kmol, Mw = 556
kg/kmol, Tfus,o = 271 K, Tfus,w = 349 K, ΔHfus,o = 177.9 kJ/kg,
and ΔHfus,w = 189.6 kJ/kg. The wax appearance temperature is
TWAT = 309 K.22
As a preliminary study, we performed single-phase steady-
state simulations of the crude oil flow alone at ṁ c = 20 and 40
kg/min. The average wall shear stress and heat transfer
coefficient matched the experimental measurement with an
accuracy within 5%, thus validating the single-phase flow
model. Afterward, we ran six different two-phase flow cases: for
each value of ṁ c, we conducted one simulation starting with a
clean wall (corresponding to texp = 0), and two simulations
starting with an existing deposit layer of thickness and wax
content corresponding to texp = 2.67 and 8 h in the
experiments. The results of the six cases are illustrated in
Figure 6. For each case, different values of the deposit viscosity,
unknown in the experiments, were tested. In the simulations,
the average deposit growth rate is extracted as a linear fit of the
Figure 5. Sketch of the two-dimensional axisymmetric mesh used for hd(t) curve when t > 1 s. For the experiments, the deposit
the validation case near the inlet section. The figure illustrates the growth rate is calculated from the logarithmic fit provided by
near-wall cell refinement and the interface of the deposit layer the authors, hd(t) = A ln(texp + B) + C (A, B, and C are
initialized for the simulation run with a crude oil flow rate of ṁ c = 20 constants defined in Singh et al.22).
kg/min and texp = 2.67 h.
When the simulation starts with a clean wall, as shown in
Figure 6a,d, hd grows linearly with time because all the
recursive cell splitting to ensure accuracy both in the radial and variables on the right-hand side of eq 17 are essentially
streamwise directions. When the simulation starts with an constant during the early stages of wax deposition. The values
existing deposited layer, a film profile thickening along the of the deposit viscosity do not influence the deposition process
streamwise direction is patched; see the white line in the inset because removal has not yet taken place. The deposition rate
of Figure 5. The mesh setup ensures that the film is always increases with the crude oil flow rate ṁ c because the more
within the finest region of the mesh to well capture the efficient heat convection at the wall leads to a higher wall
interfacial exchanges. We ran preliminary tests with different temperature gradient. The deposition rates extracted from the
mesh configurations, and we observed that the recursive cell numerical simulations for both ṁ c = 20 and 40 kg/min match
splitting maximizes accuracy and prevents excessive computa- exactly the theoretical values evaluated by means of eq 17, but
tional cost. slightly overestimate the rates calculated from the fit to the
From the experimental test matrix, we selected the data experimental data. However, the fit provided by Singh et al.22
obtained with a crude oil flow rate of ṁ c = 20 and 40 kg/min, was obtained by correlating the experimental film thickness
corresponding to Rec ≈ 7400−14 800. Because of the data at texp ≥ 2.67 h, and hence it does not necessarily
axisymmetric model, the gravitational force is neglected. The extrapolate well to calculate the deposition rate when texp is of
inlet oil temperature is Tc = 298 K, while the wall temperature, the order of seconds. Furthermore, the wall temperature is
which in the experiment is estimated from the coolant assumed to be constant in the numerical model, while in the
temperature at the oil outlet,22 is Twall = 289 K at the test experiment this may actually increase along the test section,
section outlet, but it is likely to increase in the upstream thus resulting in a reduction of the average deposition rate.
direction. For simplicity, and in the absence of more accurate When texp > 0, the deposition rate decreases with time and
data, in the numerical model the Twall is maintained constant at the simulations capture well this trend, as shown in Figure
289 K. The properties of the crude oil, ρc, μc, cpc, λc, and Γm, are 6b,c. At texp = 2.67 h, deposit viscosity values of μd ≤ 0.6 ×
set as functions of the temperature according to the 104μc yield a reduction of the deposit thickness versus time,
correlations reported by Singh et al. 22 The thermal and therefore they are assumed to be below the actual deposit
conductivity of the deposit is calculated by means of Maxwell viscosity in the experiment, where hd increases with time. The
correlation.64 No data are available for the remaining best match with the experimental data is achieved with μd =
21806 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

Figure 6. Comparison of the results of our simulations with data from the Singh et al.22 experimental setup. The plots show the average deposit
thickness hd vs time, calculated as indicated in the caption of Figure 4, when the simulation starts with (a,d) a clean wall (texp = 0), (b,e) with an
initial deposit of thickness equal to that measured in the experiments after texp = 2.67 h, and (c,f) texp = 8 h. In each graph, the curves refer to
different values of μd; when an increase of the deposit thickness is observed, the average value of the deposit growth rate dhd/dt is reported. The
textbox on the bottom-right of each graph indicates the deposit growth rate extracted from the experiments. The simulations conditions are as
follows: D = 16.5 mm, Tc = 298 K, Twall = 289 K, TWAT = 309 K; the fluid properties are set as indicated in the text.

0.75 × 104μc, while a twice larger value (μd = 1.5 × 104μc) is In summary, given the complexity of the flow problem to be
necessary to obtain agreement with the experiment at texp = 8 emulated with numerical simulations and the possible sources
h. This hints at ageing effects in the deposit, that is, in order for of discrepancy between numerical and experimental initial and
the deposit thickness to continue growing against the forces boundary conditions, the deposit growth rates obtained with
promoting removal, its viscosity must increase with time. the present numerical model agree quite well with the
Although no specific viscosity values are available in the experimental measurements. The values of the deposit
literature for the deposit phase of the Garden Banks oil, the viscosity that allow matching computational and experimental
values obtained in this study, which are of the order of μd ≈ 10 deposition rates are coherent throughout the study and are
Pa s, seem plausible when compared to the viscosity of waxy consistent with the available literature.
crude oils at low temperatures.65 Analysis of the deposition
5. IMPACT OF FLOW CONDITIONS ON WAX
data for ṁ c = 40 kg/min, as shown in Figure 6e,f, reveals the
DEPOSITION
occurrence of trend cross-over in both experiments and
simulations: at low texp, the thickness growth rate increases In this section, we discuss the results of a parametric study
with the flow rate, whereas at larger texp this trend is reversed. aimed at investigating the impact of selected flow parameters
The same trend has been observed with other crude oils,21 and and fluid properties on the wax deposition and removal
process. The simulation setup illustrated in Section 3
it is attributed to the governing effect of heat convection at the
represents the reference case (case 1); then, the effects of
early stages of wax deposition, where ∂T/∂n increases with ṁ c,
the inlet average crude oil velocity (Uc), channel diameter (D),
whereas at later stages, when the deposit is thicker, higher flow surface tension coefficient (σ), deposit rheology (μd), and inlet
rates promote shear-induced deposit removal, thus yielding crude oil temperature (Tc) are investigated separately by
lower deposit thicknesses. varying the flow parameters one at a time from the reference
Although the variations of hd versus time reported in Figure case 1. A summary of the operating conditions investigated for
6 are very small, the VOF method with geometrical each simulation case is provided in Table 3. All the remaining
reconstruction of the cell-face fluxes is still robust, thanks to fluid properties are set as reported in Section 3 and Table 2.
its mass conservation capabilities. In our simulations, this has Below, the results of a mesh sensitivity analysis are reported
been tested by verifying that the integral of the mass source first, followed by the results of the parametric study.
terms calculated over the whole domain coincides with the net 5.1. Mesh Sensitivity Study. An analysis of the mesh
variation of the deposit mass in the domain, starting from t = 0 sensitivity was first performed for single-phase flow and heat
when removal does not occur yet. transfer using constant fluid properties in order to compare the
21807 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

Table 3. Summary of the Operating Conditions Investigated in Section 5a

a
All the remaining fluid properties are set as indicated in Section 3 and Table 2. The wax appearance temperature is TWAT = 319 K. In cases 13−15,
deposit ageing is modeled and ϵ, θ are parameters within the model, see eq 21. In the last column, Rec = ρcUcD/μc, where ρc and μc are evaluated at
Tc.

numerical results with well-established correlations predicting


the heat transfer performance. Here, we discuss the results for
two different meshes, a “coarse” and a “fine” grid. Both the
meshes are three-dimensional, and a block-structured config-
uration with hexahedrons cells is adopted; see the images in
the insets of Figure 7. The central block is discretized with
cubes of side of 0.08R; the lateral blocks are refined toward the
pipe wall, down to a minimum mesh thickness along the radial
direction of 0.01R and 0.002R for the coarse and fine grids,
respectively. The sensitivity test is run for values of the crude
oil Reynolds number Rec ranging from 193 to 5000, where Rec
= 5000 is the largest Reynolds number analyzed in the
systematic study. All the fluid properties are set as indicated in
Table 2 and evaluated at the inlet temperature. The numerical
results, in terms of average Nusselt number, are compared with
the prediction method of Gnielinski66 for hydrodynamically
developed flow in a circular pipe, with a constant wall
temperature. This method employs a laminar flow correlation
for Rec < 2300 and a turbulent flow correlation when Rec > 104,
with a linear interpolation for intermediate values of Rec. As
depicted in Figure 7, the average Nusselt numbers obtained
with both the coarse and fine grids match very well when the Figure 7. Results of the grid sensitivity test performed on single-phase
flow is laminar; the deviations with the empirical correlation simulations run with different oil flow rates and constant fluid
are below 5%. However, for turbulent flow conditions (Rec = properties: average Nusselt number within the tube,
5000), we observed an excessive smearing of the thermal 1 ∂T / ∂n
Nuave = πL ∫ T − T dS , where ∂T/∂n is the temperature derivative
S
boundary layer with the coarse mesh, which resulted in a wall b
normal to the wall, S = πDL is the total wall surface, and Tb is the fluid
smaller heat transfer coefficient compared to the benchmark bulk temperature. The numerical results are compared with the
data. The fine mesh yields results which agree well with the Gnielinski66 prediction method for heat transfer in laminar−turbulent
prediction from Gnielinski,66 thus suggesting that the thermal flow through pipes. The insets illustrate the computational mesh on
boundary layer at the wall is well resolved. The thermal the channel cross section for the coarse and fine grids used.
boundary layer thickness at the channel exit, measured when
the fine mesh is employed (Rec = 5000), is of about 0.03R and
therefore too thin to be adequately captured with the coarse 0.084 (coarser mesh) and hd/R = 0.09 (finer), with about 7
mesh. (coarser) and 15 (finer) cells discretizing the film region.
A mesh sensitivity test for a two-phase flow with deposition Previous numerical studies of thin-film flows and heat transfer
was then run for case 3 of Table 3 (Rec = 2000) using the two suggested that a minimum of seven cells are required across
meshes discussed above. The terminal steady-state film the film to obtain converged results,39,67 which agrees well with
thicknesses, averaged throughout the domain, were hd/R = the present observation. Hence, in order to maintain a safe
21808 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

Figure 8. Results for different crude oil flow rates: (a) average deposit thickness and (b) time-trace of the height of the film measured from the
bottom of the pipe, at z = 30D from the entrance. The inset in (a) shows a close-up of the deposit thickness at the initial time instants. In (b), the
data refer to steady flow conditions and the reference t = 0 is arbitrary. The simulations conditions are (cases 1−4 in Table 3): D = 9 mm, Uc =
0.22−5.7 m/s, μd = 100μc, σ = 0.04 N/m, Tc = 303 K, Twall = 278 K.

margin, the simulations presented in this section were run with the most unstable wave amplitude ΔhKH in a liquid−gas
the finer mesh when Rec ≥ 2000 and with the coarser one for stratified flow
lower Reynolds numbers (thicker films).
5.2. Effect of the Crude Oil Flow Rate. Starting from 3πσ(ρ1 + ρ2 )
case 1, we increase the inlet velocity of the crude oil from Uc = ΔhKH =
ρ1ρ2 |u1 − u 2|2 (18)
0.22−5.7 m/s, thus increasing Rec from 193 to 5000; see the
summary in Table 3 (runs 2−4). For each test condition, with ρ1 being the density of the continuous phase, which can
preliminary steady-state simulations are run to achieve be rewritten in nondimensional form by defining a Weber
developed velocity and temperature fields. The evolution of number of the relative motion, Werel = ρ1|u1 − u2|2D/σ
the average deposit film thickness within the pipe is reported as
a function of time in Figure 8a. It can be seen that during the ΔhKH 6π (1 + ρ1 /ρ2 )
early stages of wax deposition, the deposition rate increases =
with Rec because of the enhanced heat convection that R Werel (19)
promotes larger wall temperature gradients. However, larger
values of Rec also promote deposit removal, and therefore the with Werel = 200−104 under the conditions presently studied
terminal steady-state deposit thickness increases when (velocity of the deposit, u2, set to zero and u1 set to the
reducing the oil flow rate. This yields a cross-over among maximum speed at the channel axis). Although the equation
the t−hd curves for different flow rates, which was also above does not necessarily apply to the present flow
observed in the flow loop experiments by Panacharoensawad configuration, the predicted value of the wave amplitude for
Rec = 193 agrees well with the simulation data, ΔhKH = 0.2R
and Sarica21 and Singh et al.22 Under the set of parameters
(predicted) versus ≈0.25R (simulation), while it under-
analyzed, a continuous film of deposit covers the entire pipe
estimates the numerical results at larger Rec. Especially at the
walls and the removal mechanism is via interfacial shear, and
lower Reynolds numbers, where the deposit thickness and the
no entrainment of deposit fractions in the core flow is
wave amplitude is larger, the mass transported by the fast
observed.
waves may yield a non-negligible contribution to the deposit
From the onset of the two-phase flow, the total amount of
removal compared to the mass transported within the slowly
deposit within the pipe increases until deposition and removal
moving film.
rates balance out and hd reaches a plateau (Figure 8a).
We have extracted the time trace of the film heights also
Although this is true at the pipe scale (when hd is averaged
from the top of the channel (not reported in the figures) to
throughout the flow domain), the local thickness of the film
verify the impact of gravitational effects on the perimetral
still exhibits significant variations in time due to the occurrence
distribution of the deposit film. The liquid film on the top is on
of large-amplitude waves. Figure 8b depicts the time trace of average 12% thinner than that at the bottom when Rec = 193,
the height of the film measured from the bottom of the pipe on while the difference decreases below 10% for larger Reynolds
a cross section located at 30D from the entrance. The values of numbers. Cioncolini and Thome71 compiled a large databank
the base film thickness (distance from wall to wave trough) are of experimental gas−liquid annular flow data and proposed a
very close to the average film thickness hd for each Rec, but the liquid film circumferential asymmetry prediction model based
wave amplitudes (distance between wave base height and wave upon the competition between inertia and gravitational forces.
peak height) are of magnitude comparable to hd, so that locally They defined a Froude number of the flow as
the height of the film may be over twice its average thickness.
The wave amplitudes decrease when increasing Rec, in Fr = ρ1u12 /(ΔρgD) , with u1 being the velocity of the gas
agreement with existing models for liquid−gas annular core flow, and concluded that differences in the top and
flows.68−70 Chandrasekhar68 ascribed the appearance of bottom film thicknesses above 10% arise when Fr < 10. This
waves to the relative motion of the fluids and developed a criterion seems applicable also in the present case, where Fr =
theory based on the Kelvin−Helmholtz instability to predict 5.3 and 11.4 for, respectively, Rec = 193 and 500.
21809 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

5.3. Effect of Channel Diameter. We investigate the waves are observed along the center of the channel, where the
impact of the pipe size on the deposition dynamics by varying velocity of the crude oil is faster. Despite very different flow
D in the range D = 1−20 mm (cases 5−7). Wax deposition patterns observed when varying Uc and D, the steady-state film
within capillaries (D ≤ 1 mm) is relevant as a model for thickness values for laminar flow conditions collapse quite well
asphaltene deposition in reservoir pores,72 where low Reynolds along a straight line when plotted versus Rec on a logarithmic
numbers are encountered compared to flows in transportation scale (Figure 9a), thus suggesting a power-law dependence,
pipelines. In this study, we keep the inlet oil velocity constant best fit hd/R ≈ Rec−0.28, between nondimensional film thickness
(Uc = 0.22 m/s, see Table 3), so that Rec = 21−429. The values and Reynolds number of the oil flow.
of the film thicknesses at the terminal steady-state regime are 5.4. Effect of Surface Tension. The surface tension force
plotted in Figure 9a as a function of Rec. Smaller diameters has a stabilization effect on the flow because it becomes more
difficult to perturb the interface, and therefore it is expected to
have a significant impact on the distribution of the deposited
film over the pipe wall. In order to evaluate this effect, the
surface tension coefficient is varied within the range σ = 0−
0.08 N/m (cases 8−10), with σ = 0.04 N/m representing the
baseline case 1. The average deposit thickness within the pipe
obtained at the terminal steady regime increases from hd/R =
0.15 when surface tension is absent, to hd/R = 0.22 when σ =
0.08 N/m, while completely different flow patterns are
established; see Figure 10. In the absence of surface tension,

Figure 10. Snapshots of the deposit distribution (Fc ≤ 0.5) within the
pipe for (top to bottom) σ = 0 and 0.08 N/m, colored by the local
velocity magnitude; part of the deposit is clipped to allow
visualization; x and y refer to generic horizontal and vertical
coordinates, respectively. The remaining conditions are (cases 8−
10): D = 9 mm, Uc = 0.22 m/s, μd = 100μc, Tc = 303 K, Twall = 278 K.

the flow configuration is annular wavy; droplets of deposited


Figure 9. (a) Average film thicknesses at steady conditions vs crude phase are continuously detached from the top of the interfacial
oil Reynolds number obtained when varying the pipe diameter (D = waves and entrained in the pipe core flow. When σ = 0.02−
1−20 mm) and oil inlet velocity (Uc = 0.22−5.7 m/s). The deposit 0.04 N/m (see Figure 3), droplet entrainment disappears and
thickness prediction model is presented in Section 6. (b) Snapshots of the capillary forces promote the formation of a thicker deposit
the deposit distribution (Fc ≤ 0.5) within the pipe for different D, film. At the largest surface tension value tested (slightly above
colored by the local velocity magnitude; part of the deposit is clipped that of air−water), the flow transitions to a stratified wavy
to allow visualization; x and y refer to generic horizontal and vertical regime. This result is consistent with existing studies on gas−
coordinates, respectively. The flow conditions are (cases 5−7): μd =
100μc, σ = 0.04 N/m, Tc = 303 K, Twall = 278 K.
liquid and liquid−liquid flows,73 which show that finite surface
tension values tend to shift the stability boundary of stratified
(smaller Rec) yield larger relative deposit thicknesses as flows toward larger superficial velocities.
capillary forces overcome inertia (see next section) and higher 5.5. Effect of Deposit Rheology. In this section, we
investigate the impact of the deposit rheology by testing both
radial temperature gradients occur (as dT/dr ≈ (Tc − Twall)/
time-independent and time-dependent (ageing) viscosity
R). Figure 9b illustrates the deposit distribution within the
models (runs 11−15). The exposure of the deposit layer to
channel at different D, and it can be seen that the deposition a range of temperature and shear stress histories triggers
patterns differ significantly. When D = 1 mm, gravitational physical and chemical processes which alter its molecular
effects are negligible (Fr = 22.7), and the flow is essentially structure and properties.74 In the specific case of wax
axisymmetric, so that interfacial waves appear as ring waves. deposition, the thermal gradient across the deposit may result
When D = 5 mm, top and bottom average film thicknesses still in an internal mass flux which yields a continuous increase of
match, although the waves are no longer axisymmetric. When the deposit wax content.1 This leads to hardening of the
D = 20 mm, gravitational effects become important (Fr = 3.2), deposit as time elapses, that is, ageing of the deposit. Deposit
and the flow pattern appears as stratified, with the heavier ageing in fluid flow simulations can be implemented by
deposit phase flowing at the bottom of the channel. Interfacial adopting a rheological model that takes into account the
21810 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

Figure 11. (a) Average deposit thickness vs time with ageing (μd function of temperature, shear rate, and time) and without ageing effects (μd only
a function of temperature); the simulations conditions are (cases 11−15): D = 9 mm, Uc = 0.22 m/s, σ = 0.04 N/m, Tc = 303 K, Twall = 278 K. (b)
Total mass of each pseudocomponent deposited in the domain for different inlet oil temperatures; solid lines refer to non-wax species md,o and
dashed lines to wax species md,w; the remaining conditions are (cases 16−18): D = 9 mm, Uc = 0.22 m/s, σ = 0.04 N/m, μd = 100μc.

variation of the deposit viscosity which depends on the flow the asymptotic values of the viscosity, which in turn makes the
history15,75 flow in the film slower, thus reducing the shear rate and further
increasing Λ(t → ∞). As a result, we observe an indefinite
μd = μd,0 (1 + Λn) (20) increase of μd and build-up of the deposit content in the pipe
where Λ(t) is a structure parameter that describes the local as time elapses, and steady conditions are never achieved
strength of the microscopic network of wax crystals, μd,0 is the within the simulation time.
asymptotic viscosity value corresponding to a negligible role of 5.6. Effect of Operational Temperatures. With the aim
the network interactions (Λ → 0), and n is a constant that of evaluating the effect of the temperature on the two-phase
characterizes the fluid behavior under stress. The structure flow, simulations are run by varying the inlet oil temperature Tc
parameter evolves with the flow history and we assume that it and wall temperature Twall (runs 16−18), but keeping the same
results from the competition between two opposite processes, difference Tc − Twall = 25 K in order to maintain an
a structuration process where the network of crystals is approximately constant radial temperature gradient (which
spontaneously formed and a de-structuration process where drives the deposition process). Figure 2b shows that ηo ≫ ηw
the network is continuously disrupted as a result of the flow, when T < 300 K and vice versa at larger temperatures, and
according to the following constitutive relationship51,76,77 hence the deposit wax content is expected to increase with the
temperature. We observe that the steady-state deposit
dΛ 1 thickness decreases monotonically when increasing the
= − εΛ|γ |̇
dt θ (21) operating temperatures (figure not reported here), in agree-
ment with previous studies.1,23,78 This behavior cannot be
where θ is a characteristic time of restructuration of the attributed solely to the decreasing trend of ηo + ηw with the
material, ϵ is a material-dependent constant, and γ̇ is the local temperature because this is non-monotonic at larger temper-
shear rate. Note that, if steady-state regime is achieved, it atures. Indeed, there is a significant change of the deposit
follows that Λ = 1/(ϵθ|γ̇|) and μd = μd,0[1 + (ϵθ|γ̇|)−n]. In the viscosity when temperatures are increased, as μd decreases by 1
present analysis, we have set n = 1 (ideal yield stress fluid), μd,0 order of magnitude across the range of temperatures
= μc(T), and we tested three different combinations for ϵ and considered, thus significantly promoting film removal. Figure
θ, starting from the values adopted in Sileri et al.15 for 11b depicts the composition of the deposited mass versus time,
asphaltene. All the other flow parameters are left unchanged and the trends are in agreement with those of η (Figure 2b),
from the baseline case; see Table 3. The average film thickness with the wax content of the deposit increasing from 10 to 90%
height obtained for simulations with ageing effects are plotted as Tc rises to 323 K. Singh et al.1,78 also reported an increase of
versus time in Figure 11a, together with the results of the deposit wax content for higher wall temperatures, although
simulations run with the standard temperature-dependent of smaller entity than that observed here, because of a smaller
model. As expected, the deposit viscosity has a remarkable range of variation of the operating temperatures and a reduced
influence on the flow dynamics as it directly impacts the content of dissolved wax in the oil.
mobility of the deposit film. For the two sets of parameters (θ,
ϵ) = (1, 0.01) and (0.1, 0.1), the steady-state viscosity values
are essentially the same because Λ(t → ∞) ≈ 1 (|γ̇| = 4Uc/R ≈ 6. PREDICTION MODEL FOR THE DEPOSIT
200 1/s at the wall of a clean pipe in these conditions), so that THICKNESS
μd(t → ∞) ≈ 1.5μc. This explains the ∼15% thicker film The direct application of the numerical model to study field
observed at steady-state (orange and green curves in Figure pipelines is limited by the relatively short deposition time that
11a) when compared to the case run without ageing effects and can be achieved via interface-resolving simulations. Nonethe-
μd = μc (blue curve). Decreasing the factor ϵθ by an order of less, the simulation results can be used to derive simplified
magnitude (θ = 1, ϵ = 0.001, yellow line in Figure 11a) yields models for the flow and heat transfer in the deposit layer. In
much larger viscosity values at later time instants, μd ≈ 100μc this section, we propose a prediction model for the steady-state
and above. This happens because lower values of ϵθ increase deposit thickness by developing closure laws for the velocity
21811 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

Figure 12. (a) Velocity and (b) temperature profiles along a vertical centerline located at z/D = 25 from the pipe inlet; see the schematic at the
top-left of (a); y indicates the radial distance from the pipe wall. The red dashed line locates the crude−deposit interface, here yint/R = 0.17,
identified as the Fc = 0.5 point. The insets on the bottom-right in (a) and in (b) show a close-up of the profiles within the deposit film (where Fc =
0). The simulation conditions are (case 1): D = 9 mm, Uc = 0.22 m/s, μd = 100μc, Tc = 303 K, Twall = 278 K.

and temperature in the film flow region extracted from the increase. For simplicity, the flow within the film is modeled as a
numerical database. Couette flow, with the following linear velocity profile
The accumulation of mass in the deposit as time elapses can τ
be expressed via a balance of deposition and removal rates uz(y) = int y
μd (24)
dmd
= ṁ dep − ṁ rem where τint is the shear stress at the interface and μd is an
dt (22)
average value of the film viscosity. In the present case, the bulk
where, assuming that the pipe wall is covered by a uniformly oil flow is laminar and the shear stress at the interface can be
deposited film of thickness hd, md = ρdπLhd(2R − hd) and ṁ dep calculated by assuming a parabolic velocity profile within the
= j2π(R − hd)L. j and the removal terms require assumptions cylindrical core region occupied by the oil
ij R yz Uc
τint = 4μc jjj zz
j R − hd zz R − hd
about their specific driving phenomena. If deposition via 2

k {
molecular diffusion is assumed, j can be predicted using the
Svendsen13 model (25)

ρ(ηo + ηw ) dT where the equation above takes into account the reduction of
j = Γm the oil flow area due to the presence of the deposit. The shear
Tint dy rate on the oil side of the interface predicted with this model is
yint (23)
duz/dy = 325 1/s, which agrees well with the value extracted
where ρ is the mixture density. For simplicity, the oil−deposit from the simulation, duz/dy = 365 1/s, indicated in Figure 12a.
interface is considered to be parallel to the pipe wall; the flow Introducing eq 25 into eq 24, a linear velocity profile for the
and heat transfer within the deposited film are described using flow in the film is derived. This compares quite well with the
a rectangular coordinate system z−y, where z is the streamwise simulation data as it can be seen in the inset of Figure 12a. In
coordinate and y is the radial distance from the pipe wall, as the case of turbulent oil flow, τint in eq 24 should be calculated
indicated in the schematic of Figure 1; as such, dT/dy in eq 23 using an appropriate correlation for the interfacial shear stress.
represents the temperature gradient in a direction perpendic- Based on the linear viscosity increase observed in the film in
ular to the wall. Equation 23 requires knowledge of the the simulation, we also derived an analytical solution for a
interfacial temperature Tint and the temperature derivative Couette flow with a viscosity μd varying linearly with y;
normal to the interface. Therefore, a thermal and fluid however, this did not differ in an appreciable way from that
mechanical model for the flow in the pipe is necessary, obtained with a constant μd. If we assume that the removal of
together with an adequate description of the removal process. the deposit layer occurs exclusively because of the shear
To this end, we analyze the velocity and temperature data of exerted by the oil flow, which generates a streamwise deposit
the numerical simulation indicated as case 1 in Table 3, and flow rate, the theory above can be used to evaluate analytically
Figure 12a depicts the velocity and temperature profiles along the removal term of eq 22
a vertical centerline located at z/D = 25 from the pipe inlet,
extracted at a time instant when the flow is steady-state. It can hd τint ρ τintπRhd 2
be seen that the streamwise velocity uz within the film is very
ṁ rem = 2ρd πR ∫0 μd
y dy = d
μd (26)
small, below 1% of the average oil velocity Uc. The fluid in the
film moves in the streamwise direction because of the There are some interesting similarities between eq 26 and
interfacial shear exerted by the bulk oil. The shear rate within the empirical correlations adopted in previous studies for
the film is not constant, as it grows from duz/dy = 1.82 1/s at calculating deposit removal by shear. ṁ rem is linearly propor-
the pipe wall to 3.7 1/s near the interface. This can be ascribed tional to the wall shear stress,29−32 while the term hd2 accounts
to an observed linear decrease of the deposit viscosity by a for the existing deposit mass.29,30 In Ramirez-Jaramillo et al.,29
factor of 2 across the film, induced by the temperature the removal rate is proportional to a term e−1/T, while in
21812 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

Correra et al.32 and Nazar et al.,30 this is inversely proportional where αint is the heat transfer coefficient at the interface and
to the wax fraction of the deposit, which increases with time; in the bulk oil temperature is approximated with its inlet value Tc.
eq 26, these dependencies are both embedded in μd, as larger This leads to the following analytical expressions for the
temperatures yield lower deposit viscosity (i.e., enhanced interface temperature
mobility), while μd increases with time because of deposit
ageing. αint(Tc − Twall)
Tint = Twall + hd
The temperature profile across the deposit film is linear, as it λd + αinthd (28)
is apparent from Figure 12b. This suggests that heat transfer
across the film is governed by radial heat conduction, which is and temperature gradient
consistent with the very low fluid velocity observed. Figure 13
dT αint(Tc − Twall)
=
dy λd + αinthd
yint (29)

which can be directly used in the deposition term, eq 23, and


to calculate ηo and ηw therein. This linear temperature profile
for the film is compared with the simulation data in Figure
12b; the interface heat transfer coefficient αint is estimated by
means of a correlation for hydrodynamically developed laminar
flow in a pipe,66 using z/D = 25 as the development length,
which yields Nuint = 10.2 and αint = λdNuint/(D − 2hd) = 179
W/(m2 K). The theoretical profile slightly underestimates the
interface temperature and temperature gradient (Figure 12b);
the best agreement with the numerical data would be obtained
with a value of Nuint about twice as large, corresponding to a
development length of about 1D when calculated by means of
Gnielinski66 correlation.
Figure 13. Interface temperature and film thickness profiles along the The complete model can be used to predict the deposit
circumferential direction on a cross section located at z/D = 25 from thickness at steady state; by setting dmd/dt = 0 in eq 22, such
the pipe inlet. The zero angle reference is at the bottom of the cross that ṁ dep = ṁ rem, it follows that
section. The flow conditions are the same as those indicated in the
caption of Figure 12. ρ(ηo + ηw ) αint(Tc − Tw )
Γm 2π (R − hd)L

μc ij R yz Uc
Tint λd + αinthd

jj zz
μd jjk R − hd zz{ R − hd
displays the oil−deposit interface temperature and film 2
thickness along the perimeter of the same cross section and = 4ρd πRhd 2
time instant studied in Figure 12; it can be seen that the film (30)
thickness varies considerably along the pipe wall, with the
interface temperature profile following very well that of hd, thus which is implicit in hd and has to be solved numerically. We
further confirming the dominant effect of heat conduction. have calculated the predicted values of hd starting with the
Based on this observation, the temperature profile within the operating conditions of case 1 and varying Rec within the range
film is calculated analytically by solving a one-dimensional heat Rec = 20−2000; the results were included in Figure 9a. The
conduction problem. At the pipe wall, T(y = 0) = Twall. At the model captures well the trend obtained with the simulations,
oil−deposit interface, the boundary condition can be extracted with a tendency to overestimate the numerical data, from 20%
by imposing continuity of temperature and heat flux with the deviation when Rec = 20, to 50% when Rec ≈ 2000. This
bulk oil flow, as done for the velocity field. We tested two fully deviation may act as a factor of safety when applying the
developed profiles for the temperature in the bulk oil, deriving present model to the prediction of the deposit thickness in
one from a parabolic velocity profile and the other from a plug field pipelines. The differences between model and simulation
flow profile. However, in both cases, we observed a remarkable results may be attributed to: (i) use of a correlation developed
underestimation of the heat flux and, therefore, of the for single-phase flow in a smooth tube to predict the interfacial
temperature gradient at the interface. As a matter of fact, the heat transfer coefficient, which underestimates αint and
temperature profile extracted from the simulation (Figure 12b) therefore yields lower interfacial temperatures (see Figure
exhibits a rather flat shape in the bulk flow region, compared to 12b) that promote deposition; (ii) the steady-state model does
fully developed profiles which are blunter at the pipe axis. This not account for the dynamics of the traveling waves and the
is likely due to the longer length required to develop the mass that they transport; (iii) the comparison is done on the
temperature profile in a stratified flow compared to a single- spatially average film thickness, while in the simulations the
phase pipe flow, and to the presence of interfacial waves that film grows gradually from the channel inlet. Nonetheless, eq 30
continuously disrupt the temperature boundary layer over the provides a mechanistic framework to estimate the impact of
interface. Therefore, we preferred to adopt a convective the flow conditions on the wax deposition and removal
boundary condition at the interface mechanisms; potential improvements may focus on more
faithful representations of the deposit rheology and more
dT accurate descriptions of the two-phase dynamics and heat
λd = αint[Tc − T (y = yint )]
dy transfer, for example, including wave formation and the
yint (27) associated mass transport.
21813 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

7. CONCLUSIONS Video of the numerical simulation of wax deposition


presented in Section 3 (AVI)


We have developed a numerical model to investigate the
dynamics of wax deposition and removal within oil trans-
portation pipelines. Liquid-like crude oil and solid-like deposit AUTHOR INFORMATION
are treated as two immiscible phases separated by an interface, Corresponding Author
and each phase is considered to be a bicomponent mixture of *E-mail: mirco.magnini@nottingham.ac.uk.
wax and non-wax species. The model is based on the coupled ORCID
solution of the unsteady mass, momentum, energy, and species Mirco Magnini: 0000-0002-9481-064X
transport equations for the two-phase flow and the interface
Notes
dynamics is explicitly captured. The rate of deposition is
The authors declare no competing financial interest.


evaluated through a chemical-equilibria model, and it depends
on the local temperature and mass fractions at the interface. A ACKNOWLEDGMENTS
model for thixotropic fluids is utilized to account for the ageing
of the deposit in time. This numerical framework enables new O.K.M. acknowledges funding from PETRONAS and the
insight on the complex interaction between interface dynamics, Royal Academy of Engineering for Research Chair. The
fluid mechanics, thermodynamics, and heat transfer during the authors are grateful to G. Gonçalves (Imperial College
wax deposition process and allows the impact of the flow London) for valuable discussions and comments during the
development of this work.


conditions on the deposit removal mechanisms to be studied in
detail.
The simulation tool has been first validated versus
NOMENCLATURE
experimental wax deposition data available from the literature. Roman Letters
Afterward, we have studied the effect of the oil flow rate, pipe A area [m2]
diameter, surface tension, deposit rheology, and temperature c concentration [kg/m3]
on the deposit formation, and we observed that this may cp constant pressure specific heat [J/(kg K)]
appear as a continuous film covering the pipe wall and present D diameter [m]
fast traveling waves, or be completely stratified, depending on F VOF volume fraction [-]
the relative importance of gravitational, inertial, capillary, and Fr Froude number, ρU 2/(ΔρgD) [‐]
viscous forces. The dominant deposit removal mechanism is
via the interfacial shear exerted by the bulk oil flow, although g gravitational acceleration [m/s2]
deposit entrainment may occur as surface tension weakens. H specific enthalpy [J/kg]
The deposit builds up faster for lower flow rates and smaller h height [m]
tubes, and the average film thickness at the steady-state j deposition flux [kg/(m2 s)]
correlates well with the Reynolds number of the bulk flow, for K, K0 equilibrium constants in the deposition model [-]
different oil velocities and channel diameters. The deposit k turbulence kinetic energy [m2/s2]
rheology has a remarkable impact on the flow dynamics, and L length [m]
ageing effects can lead to significant structure-building within Lx moles of liquid per mole of fluid [-]
the deposit, eventually resulting in an indefinite increase of M molecular weight [kg/kmol]
deposit viscosity and thickness. Larger operational temper- m mass [kg]
atures seem to mitigate the deposit build-up, in particular when ṁ dep, ṁ rem deposition/removal rates [kg/s]
the viscosity exhibits a marked dependence on the temper- Nu Nusselt number, αD/λ [-]
ature. n normal direction [-]
The insight gained via the numerical simulations has been Prt turbulent Prandtl number, μtcp/λt [-]
utilized to develop a prediction method to estimate the rates of p pressure [Pa]
deposition and removal based on analytical closures for the q heat flux [W/m2]
velocity and temperature profiles in the deposit film. The R radius [m]
model captures well the dependence of the deposit thickness Rg universal gas constant [J/(mol K)]
on the Reynolds number of the flow with a tendency to Re Reynolds number, ρUD/μ [-]
overpredict the simulation results. Performance can be Sct turbulent Schmidt number, μt/ρΓt [-]
improved by implementing better closure relations, focusing T temperature [K]
in particular on interfacial shear stress and heat transfer. t time [s]
Finally, the fundamental analysis of mass and momentum U, u velocity [m/s]
transfers within the deposited film would benefit from better V volume [m3]
characterization of the rheological properties of wax deposits x horizontal coordinate [m]
from crude oils of industrial use. Y mass fraction [-]


y radial distance from pipe wall or vertical coordinate
[m]
ASSOCIATED CONTENT We Weber number, ρU2D/σ [-]
* Supporting Information
S z streamwise coordinate [m]
The Supporting Information is available free of charge on the Greek Letters
ACS Publications website at DOI: 10.1021/acs.iecr.9b05250. α heat transfer coefficient [W/(m2 K)]
Γ diffusion constant [m2/s]
Description of the video of the numerical simulation Γm average diffusion constant driving deposition [m2/s]
(PDF) γ̇ shear rate [1/s]
21814 DOI: 10.1021/acs.iecr.9b05250
Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

δ VOF delta-function [1/m] (14) Huang, Z.; Lee, H. S.; Senra, M.; Scott Fogler, H. A
ϵ constant in ageing model [-] fundamental model of wax deposition in subsea oil pipelines. AIChE J.
η variation of mass concentration at liquid state per unit 2011, 57, 2955−2964.
temperature [-] (15) Sileri, D.; Sahu, K. C.; Matar, O. K. Two-fluid pressure-driven
θ characteristic time of restructuration [s] channel flow with wall deposition and ageing effects. J. Eng. Math.
2011, 71, 109−130.
κ interface curvature [1/m]
(16) Zheng, S.; Fogler, H. S.; Haji-Akbari, A. A fundamental wax
Λ structure parameter [-] deposition model for water-in-oil dispersed flows in subsea pipelines.
λ thermal conductivity [W/(m K)] AIChE J. 2017, 63, 4201−4213.
μ dynamic viscosity [Pa s] (17) Huang, Z.; Zheng, S.; Fogler, H. S. Wax Deposition:
ρ density [kg/m3] Experimental Characterizations, Theoretical Modeling, and Field
σ surface tension coefficient [N/m] Practices; CRC Press, 2015.
τ shear stress [Pa] (18) Fong, N.; Mehrotra, A. K. Deposition under turbulent flow of
ψ ratio of solid to liquid number of moles [-] wax-solvent mixtures in a bench-scale flow-loop apparatus with heat
ω turbulence specific dissipation rate [1/s] transfer. Energy Fuels 2007, 21, 1263−1276.
Subscripts (19) Bidmus, H. O.; Mehrotra, A. K. Solids deposition during “cold
flow” of wax-solvent mixtures in a flow-loop apparatus with heat
c crude-oil phase transfer. Energy Fuels 2009, 23, 3184−3194.
cell computational cell (20) Parthasarathi, P.; Mehrotra, A. K. Solid deposition from multi-
d deposit phase component wax-solvent mixtures in a benchscale flow-loop apparatus
eff effective with heat transfer. Energy Fuels 2005, 19, 1387−1398.
fus fusion (21) Panacharoensawad, E.; Sarica, C. Experimental study of single-
int oil−deposit interface phase and two-phase water-in-crude-oil dispersed flow wax deposition
o oil pseudo-species in a mini pilot-scale flow loop. Energy Fuels 2013, 27, 5036−5053.
t turbulent (22) Singh, A.; Panacharoensawad, E.; Sarica, C. A mini pilot-scale
WAT wax-appearance temperature flow loop experimental study of turbulent flow wax deposition by
w wax pseudo-species using a natural gas condensate. Energy Fuels 2017, 31, 2457−2478.
wall pipe wall (23) Tinsley, J. F.; Prud’homme, R. K. Deposition apparatus to study


the effects of polymers and asphaltenes upon wax deposition. J. Pet.
Sci. Eng. 2010, 72, 166−174.
REFERENCES (24) Creek, J. L.; Lund, H. J.; Brill, J. P.; Volk, M. Wax deposition in
(1) Singh, P.; Venkatesan, R.; Fogler, H. S.; Nagarajan, N. single phase flow. Fluid Phase Equilib. 1999, 158−160, 801−811.
Formation and aging of incipient thin film wax-oil gels. AIChE J. (25) Hoffmann, R.; Amundsen, L. Single-phase wax deposition
2000, 46, 1059−1074. experiments. Energy Fuels 2010, 24, 1069−1080.
(2) Aiyejina, A.; Chakrabarti, D. P.; Pilgrim, A.; Sastry, M. K. S. Wax (26) Huang, Z.; Lu, Y.; Hoffmann, R.; Amundsen, L.; Fogler, H. S.
formation in pipelines: A critical review. Int. J. Multiphase Flow 2011, The effect of operating temperatures on wax deposition. Energy Fuels
37, 671−694. 2011, 25, 5180−5188.
(3) Rittirong, A.; Panacharoensawad, E.; Sarica, C. An experimental (27) Ramírez-Jaramillo, E.; Lira-Galeana, C.; Manero, O. Numerical
study of paraffin deposition under two-phase gas-oil slug flow in simulation of wax deposition in oil pipeline systems. Pet. Sci. Technol.
horizontal pipes. Offshore Technology Conference, 2015. 2001, 19, 143−156.
(4) Srivastava, S. P.; Handoo, J.; Agrawal, K. M.; Joshi, G. C. Phase- (28) Hernandez, O. C.; Hensley, H.; Sarica, C.; Brill, J. P.; Volk, M.
transition studies in n-alkanes and petroleum-related waxes - A review. Improvements in Single-Phase Paraffin Deposition Modeling. SPE
J. Phys. Chem. Solids 1993, 54, 639−670. Prod. Facil. 2004, 19, 237−244.
(5) Azevedo, L. F. A.; Teixeira, A. M. A critical review of the (29) Ramirez-Jaramillo, E.; Lira-Galeana, C.; Manero, O. Modelling
modeling of wax deposition mechanisms. Pet. Sci. Technol. 2003, 21,
wax deposition in pipelines. Pet. Sci. Technol. 2004, 22, 821−861.
393−408.
(30) Nazar, A. R. S.; Dabir, B.; Islam, M. R. Experimental and
(6) Sarmento, R. C.; Ribbe, G. A. S.; Azevedo, L. F. A. Wax blockage
mathematical modeling of wax deposition and propagation in pipes
removal by inductive heating of subsea pipelines. Heat Transfer Eng.
transporting crude oil. Energy Sources 2005, 27, 185−207.
2004, 25, 2−12.
(31) Eskin, D.; Ratulowski, J.; Akbarzadeh, K. A model of wax
(7) Kelland, M. A. Production Chemicals for the Oil and Gas Industry;
CRC Press, 2009. deposit layer formation. Chem. Eng. Sci. 2013, 97, 311−319.
(8) Chi, Y.; Daraboina, N.; Sarica, C. Investigation of inhibitors (32) Correra, S.; Fasano, A.; Fusi, L.; Merino-Garcia, D. Calculating
efficacy in wax deposition mitigation using a laboratory scale flow deposit formation in the pipelining of waxy crude oils. Meccanica
loop. AIChE J. 2016, 62, 4131−4139. 2007, 42, 149−165.
(9) Rana, D. P.; Bateja, S.; Biswas, S. K.; Kumar, A.; Misra, T. R.; (33) Banki, R.; Hoteit, H.; Firoozabadi, A. Mathematical formulation
Lal, B. Novel microbial process for mitigating wax deposition in down and numerical modeling of wax deposition in pipelines from enthalpy-
hole tubular and surface flow lines. SPE Oil and Gas India Conference, porosity approach and irreversible thermodynamics. Int. J. Heat Mass
2010. Transfer 2008, 51, 3387−3398.
(10) Etoumi, A.; El Musrati, I.; El Gammoudi, B.; El Behlil, M. The (34) Albagli, R. C.; Souza, L. B.; Nieckele, A. O. Reynolds number
reduction of wax precipitation in waxy crude oils by Pseudomonas influence on wax deposition. Offshore Technology Conference, 2017.
species. J. Ind. Microbiol. Biotechnol. 2008, 35, 1241−1245. (35) Bhat, N. V.; Mehrotra, A. K. Modeling of deposition from
(11) Lu, Y.; Huang, Z.; Hoffmann, R.; Amundsen, L.; Fogler, H. S. “waxy” mixtures in a pipeline under laminar flow conditions via
Counterintuitive effects of the oil flow rate on wax deposition. Energy moving boundary formulation. Ind. Eng. Chem. Res. 2006, 45, 8728−
Fuels 2012, 26, 4091−4097. 8737.
(12) Burger, E. D.; Perkins, T. K.; Striegler, J. H. Studies of wax (36) Mehrotra, A. K.; Bhat, N. V. Modeling the effect of shear stress
deposition in the trans Alaska pipeline. J. Pet. Technol. 1981, 33, on deposition from “waxy” mixtures under laminar flow with heat
1075−1086. transfer. Energy Fuels 2007, 21, 1277−1286.
(13) Svendsen, J. A. Mathematical modeling of wax deposition in oil (37) Mehrotra, A. K.; Bhat, N. V. Deposition from “waxy” mixtures
pipeline systems. AIChE J. 1993, 39, 1377−1388. under turbulent flow in pipelines: inclusion of a viscoplastic

21815 DOI: 10.1021/acs.iecr.9b05250


Ind. Eng. Chem. Res. 2019, 58, 21797−21816
Industrial & Engineering Chemistry Research Article

deformation model for deposit aging. Energy Fuels 2010, 24, 2240− units. Proceedings of International Conference on Heat Exchanger Fouling
2248. and Cleaning VIII, 2009; pp 245−251.
(38) Hirt, C. W.; Nichols, B. D. Volume of fluid (VOF) method for (64) Bird, R. B.; Stewart, W. E.; Lightfoot, E. N. Transport
the dynamics of free boundaries. J. Comput. Phys. 1981, 39, 201−225. Phenomena, 2nd ed.; John Wiley & Sons: New York, 2007.
(39) Magnini, M.; Pulvirenti, B.; Thome, J. R. Numerical (65) Wardhaugh, L. T.; Boger, D. V. Flow characteristics of waxy
investigation of hydrodynamics and heat transfer of elongated bubbles crude oils: application to pipeline design. AIChE J. 1991, 37, 871−
during flow boiling in a microchannel. Int. J. Heat Mass Transfer 2013, 885.
59, 451−471. (66) Gnielinski, V. VDI Heat Atlas, 2nd ed.; Springer-Verlag: Berlin
(40) Bidmus, H. O.; Mehrotra, A. K. Heat-transfer analogy for wax Heidelberg, 2010.
deposition from paraffinic mixtures. Ind. Eng. Chem. Res. 2004, 43, (67) Gupta, R.; Fletcher, D. F.; Haynes, B. S. On the CFD modelling
791−803. of Taylor flow in microchannels. Chem. Eng. Sci. 2009, 64, 2941−
(41) Anderson, D. A.; Tannehill, J. C.; Pletcher, R. H. Computational 2950.
Fluid Mechanics and Heat Transfer; Hemisphere Publishing Corpo- (68) Chandrasekhar, S. Hydrodynamic and Hydromagnetic Stability;
ration, 1984. Oxford University Press, 1981.
(42) Brackbill, J. U.; Kothe, D. B.; Zemach, C. A continuum method (69) Holowach, M. J.; Hochreiter, L. E.; Cheung, F. B. A model for
for modeling surface tension. J. Comput. Phys. 1992, 100, 335−354. droplet entrainment in heated annular flow. Int. J. Heat Fluid Flow
(43) Wardhaugh, L. T.; Boger, D. V. Measurement of the unique 2002, 23, 807−822.
flow properties of waxy crude oils. Chem. Eng. Res. Des. 1987, 65, 74− (70) Han, H.; Zhu, Z.; Gabriel, K. A study on the effect of gas flow
rate on the wave characteristic in two-phase gas-liquid annular flow.
83.
Nucl. Eng. Des. 2006, 236, 2580−2588.
(44) Tiwary, D.; Mehrotra, A. K. Phase transformation and
(71) Cioncolini, A.; Thome, J. R. Liquid film circumferential
rheological behavior of highly paraffinic “waxy” mixtures. Can. J.
asymmetry prediction in horizontal annular two-phase flow. Int. J.
Chem. Eng. 2004, 82, 162−174.
Multiphase Flow 2013, 51, 44−54.
(45) Dimitriou, C. J.; McKinley, G. H.; Venkatesan, R. Rheo-PIV
(72) Lawal, K. A.; Crawshaw, J. P.; Boek, E. S.; Vesovic, V.
analysis of the yielding and flow of model waxy crude oils. Energy Fuels Experimental investigation of asphaltene deposition in capillary flow.
2011, 25, 3040−3052. Energy Fuels 2012, 26, 2145−2153.
(46) Dimitriou, C. J.; McKinley, G. H. A comprehensive constitutive (73) Barmak, I.; Gelfgat, A.; Vitoshkin, H.; Ullmann, A. Stability of
law for waxy crude oil: a thixotropic yield stress fluid. Soft Matter stratified two-phase flows in horizontal channels. Phys. Fluids 2016,
2014, 10, 6619−6644. 28, 044101.
(47) Pedersen, K. S.; Rønningsen, H. P. Effect of precipitated wax on (74) Hewitt, G. F.; Coletti, F. Crude Oil Fouling; Gulf Professional
viscosities: a model for predicting non-Newtonian viscosity of crude Publishing, 2015.
oils. Energy Fuels 2000, 14, 43−51. (75) Yang, J.; Matar, O. K.; Hewitt, G. F.; Zhang, W.; Manchanda, P.
(48) Cross, M. M. Rheology of non-Newtonian fluids: A new flow Modelling of fundamental transfer processes in crude-oil fouling.
equation for pseudoplastic systems. J. Colloid Sci. 1965, 20, 417−437. Proceedings of 15th International Heat Transfer Conference, Kyoto,
(49) Casson, N. In Rheology of Disperse Systems; Mill, C. C., Ed.; Japan, 2014.
Pergamon Press: Oxford, 1959; pp 84−104. (76) Coussot, P.; Nguyen, Q. D.; Huynh, H. T.; Bonn, D. Viscosity
(50) Herschel, W. H.; Bulkley, R. Konsistenzmessungen von gummi- bifurcation in thixotropic, yielding fluids. J. Rheol. 2002, 46, 573−589.
benzollosungen. Colloid Polym. Sci. 1926, 39, 291−300. (77) Coussot, P. Rheophysics of pastes: a review of microscopic
(51) Coussot, P.; Nguyen, Q. D.; Huynh, H. T.; Bonn, D. Avalanche modelling approaches. Soft Matter 2007, 3, 528−540.
behavior in yield stress fluids. Phys. Rev. Lett. 2002, 88, 175501. (78) Singh, P.; Venkatesan, R.; Fogler, H. S.; Nagarajan, N. R.
(52) Hammami, A.; Mehrotra, A. K. Thermal behaviour of Morphological evolution of thick wax deposits during aging. AIChE J.
polymorphic n-alkanes: effect of cooling rate on the major transition 2001, 47, 6−18.
temperatures. Fuel 1995, 74, 96−101.
(53) Lakehal, D. Status and future developments of Large-Eddy
Simulation of turbulent multi-fluid flows (LEIS and LESS). Int. J.
Multiphase Flow 2018, 104, 322−337.
(54) Menter, F. R. Two-equation eddy-viscosity turbulence models
for engineering applications. AIAA J. 1994, 32, 1598−1605.
(55) Wilcox, D. C. Turbulence Modelling for CFD; DCW Industries,
Inc.: La Canada, California, 1998.
(56) Bradshaw, P.; Ferriss, D. H.; Atwell, N. P. Calculation of
boundary layer development using the turbulent energy equation. J.
Fluid Mech. 1967, 28, 593−616.
(57) Da Riva, E.; Del Col, D.; Garimella, S. V.; Cavallini, A. The
importance of turbulence during condensation in a horizontal circular
minichannel. Int. J. Heat Mass Transfer 2012, 55, 3470−3481.
(58) ANSYS Fluent 18.1 Theory Guide; ANSYS Fluent Inc., 2017.
(59) Youngs, D. L. In Numerical Methods for Fluid Dynamics;
Morton, K. W., Baines, M. J., Eds.; Academic Press, 1982; pp 273−
285.
(60) van Leer, B. Towards the ultimate conservative difference
scheme. V. A second-order sequel to Godunov’s method. J. Comput.
Phys. 1979, 32, 101−136.
(61) Patankar, S. V. Numerical Heat Transfer and Fluid Flow;
Hemisphere Publishing: New York, 1980.
(62) Issa, R. I. Solution of the implicitly discretized fluid flow
equations by operator-splitting. J. Comput. Phys. 1986, 62, 40−65.
(63) Sileri, D.; Sahu, K.; Ding, H.; Matar, O. K. Mathematical
modelling of asphaltenes deposition and removal in crude distillation

21816 DOI: 10.1021/acs.iecr.9b05250


Ind. Eng. Chem. Res. 2019, 58, 21797−21816

Das könnte Ihnen auch gefallen