Journal of Petroleum Science and Engineering 129 (2015) 7787

Contents lists available at ScienceDirect

Journal of Petroleum Science and Engineering

journal homepage: www.elsevier.com/locate/petrol

Study of asphaltene deposition in wellbores during turbulent ow

D.M. Paes a, P.R. Ribeiro a,n, M. Shirdel b, K. Sepehrnoori b
a
b

State University of Campinas, Campinas, SP, Brazil

The University of Texas at Austin, Austin, TX, USA

art ic l e i nf o

a b s t r a c t

Article history:
Received 30 October 2014
Accepted 3 February 2015
Available online 24 February 2015

During petroleum production, asphaltene particles can precipitate from the crude oil due to pressure,
temperature, and composition changes along the uid path from the reservoir to the surface. Once
precipitated, those particles can deposit in the inner surface of production tubings, restricting the available
ow area and reducing ow rates. To enable a better understanding of that complex mass transfer problem, a
new methodology was proposed in this paper. The methodology involved a comprehensive review of
fundamental concepts of the mass transfer and particle deposition theories, placing the asphaltene deposition
within a more general context of particle deposition during turbulent ow. Six published particle deposition
models (Lin et al., 1953. Ind. Eng. Chem. 45 (3), 636640. http://dx.doi.org/10.1021/ie50519a048; Friedlander
and Johnstone, 1957. Ind. Eng. Chem. 49 (7), 11511956. http://dx.doi.org/10.1021/ie50571a039; Beal, 1970.
Nucl. Sci. Eng. 40, 111; El-Shobokshy and Ismail, 1980. Atmos. Environ. 14 (3), 297304. http://dx.doi.org/
10.1016/0004-6981(80)90063-3; Papavergos and Hedley, 1984. Chem. Eng. Res. Des. 62, 275295.; Escobedo
and Mansoori, 1995. Paper SPE 29488 presented at the SPE Production Operations Symposium, Oklahoma City,
Oklahoma, 24 April. http://dx.doi.org/10.2118/29488-MS) were studied and validated with four published
aerosol experimental data sets (Friedlander, 1954. Deposition of Aerosol Particles from Turbulent Gases. Ph.D.
Dissertation, University of Illinois, Urbana, Illinois (July 1954); Wells and Chamberlain, 1967. Br. J. Appl. Phys. 18,
17931799. http://dx.doi.org/10.1088/0508-3443/18/12/317; Liu and Agarwal, 1974. J. Aerosol Sci. 5 (2),
145155. http://dx.doi.org/10.1016/0021-8502(74)90046-9; Agarwal, 1975. Aerosol Sampling and Transport.
Ph.D. Dissertation. University of Minnesota, Minneapolis, Minnesota (June 1975)). Based on the results of the
study, Beals (1970. Nucl. Sci. Eng. 40, 111) model was selected as the most suitable to predict particle
deposition and was considered adequate also to predict asphaltene deposition (limiting its application to
similar ranges of Reynolds numbers, Schmidt numbers and dimensionless relaxation times in relation to those
covered in the validation study). Finally, that model was applied in a sensitivity analysis to evaluate the most
important parameters and transport mechanisms governing asphaltene deposition in wellbores.
& 2015 Elsevier B.V. All rights reserved.

Keywords:
Asphaltene
Flow assurance
Production
Petroleum

1. Introduction
Asphaltenes are dened as the crude oil fraction that is soluble
in aromatic solvents (e.g. toluene) but insoluble in light alkanes (e.g.
n-pentane). Under reservoir conditions, they tend to remain dispersed
in the oil as a colloidal suspension. During petroleum production,
changes in oil temperature, pressure and composition can disturb the
stability of the colloidal suspension and lead to asphaltene precipitation. That process can happen both in the reservoir (due to normal
depletion or to the injection of incompatible uids in EOR operations)
and in the wellbore(due to changes in the thermodynamic conditions
of produced uids). The primary precipitates have sizes around several
nanometers and,after agglomerating to each other, reach tens of micra

Corresponding author.
E-mail address: ribeiro@dep.fem.unicamp.br (P.R. Ribeiro).

http://dx.doi.org/10.1016/j.petrol.2015.02.010
0920-4105/& 2015 Elsevier B.V. All rights reserved.

(Eskin et al., 2009). The asphaltene particles density is usually

evaluated as 1200 kg/m3.
Once precipitated, the solids can be deposited along the uid path
from the reservoir to the surface, leading to operational problems,
safety hazards and an overall decrease in production rates. Kokal and
Sayegh (1995) published an extensive literature survey on eld experiences with asphaltene deposition in wellbores. In most cases the
problems occurred in the early production stages of the wells after a
short initial period of high ow rates (turbulent ow). The deposits
were observed to be restricted to tubing depths where uid pressure
was above the oil bubble point, indicating that the multiphase ow
with gas somehow hindered deposition. Historically, light oils, with
low asphaltene content, have showed themselves to be more prone to
develop problems with asphaltenes than heavy oils. That behavior can
be explained by the fact that heavy oils, rich in asphaltene content,
tend to be rich also in resin content, and the resins act as peptizing
agents to stabilize the asphaltenes suspended in the oil (Leontaritis
and Mansoori, 1988).

78

D.M. Paes et al. / Journal of Petroleum Science and Engineering 129 (2015) 7787

Nomenclature
C
Cavg
C
C0
DB
dt
dp
fF
KB
Kd
Kd
mp
N
N0
NSc
NRe
s
t
te

instantaneous particle concentration (at y) (g/cm3)

average particle concentration on ow (g/cm3)
time averaged mean particle concentration (at y)
(g/cm3)
uctuating particle concentration (at y) (g/cm3)
Brownian diffusivity (cm2/s)
tube diameter, cm [in.]
particle diameter, cm, [m]
Fanning friction factor
Boltzmann constant, 1.38  1016 g-cm2/K-s2
transport
coefcient
or
particle
deposition
velocity (cm/s)
dimensionless particle deposition velocity
particle mass, g
radial particle ux (at y) (g/cm2-s)
radial particle ux at an innitesimal distance from
tube wall (g/cm2-s)
Schmidt number
Reynolds number
particle stopping distance (cm)
time (s)
lifetime of the near-wall eddies (s)

Escobedo and Mansoori (1995) published the rst asphaltene

deposition model found in literature. The model was developed for
vertical turbulent streams and was based on previous models used to
predict aerosol deposition (microscopic liquid or solid particles
dispersed on air currents). The authors considered that the asphaltene
particles were transported to pipe walls by a combination of diffusive
and convective mechanisms. Because of the inexistence of published
data for asphaltene deposition, they validated the model with aerosol
experimental data, collected by Friedlander and Johnstone (1957).
A different approach was taken by Ramirez-Jaramillo et al. (2006),
who modeled asphaltene deposition using concepts derived from the
wax deposition theory. They attributed the radial transport of asphaltene particles exclusively to molecular diffusion, with a major dependence on the temperature prole inside the pipe. That approach was
later criticized by other authors (Eskin et al., 2009) and did not have
continuity in literature, since the eld experience shows little inuence of uid temperature on asphaltene deposition.
Jamialahmadi et al. (2009) performed experiments to measure
asphaltene deposition in a ow-loop apparatus. They veried that
after an initial period of pure deposition, particle re-entrainment on
ow started to occur due to the erosion of the deposits. Unfortunately,
those two events (particle deposition and re-entrainment) could not
be measured separatelyand the deposition rates published were the
net result of them. To represent those deposition rates, the authors
used equations developed for aerosol deposition (Cleaver and Yates,
1975), which accounted for diffusive and convective mechanisms.
In addition to that, they used a correlation to account for particle reentrainment on ow (Watkinson and Epstein, 1970). The results
obtained agreed with the experimental measures, but some parameters in the re-entrainment correlation had to be adjusted for that.
Shirdel et al. (2012) studied the application of some particle
deposition models (Friedlander and Johnstone, 1957; Beal, 1970;
Cleaver and Yates, 1975; Escobedo and Mansoori, 1995) to predict
published experimental data. In the rst part of the paper, they used
Friedlander and Johnstones (1957) aerosol data set, which was exempt of particle re-entrainment. In the second part, they used
Jamialahmadis et al. (2009) asphaltene data set, which represents
the net result of deposition and re-entrainment. Based on the studies

tp
t
T
u
U
u
u0
u0RMS
un
V0
v
v
v0
v0RMS
x
y

particle relaxation time (s)

dimensionless relaxation time
uid temperature, K [1C]
instantaneous axial uid velocity (at y) (cm/s)
average ow velocity, cm/s [m/s]
time averaged mean axial uid velocity (at y) (cm/s)
uctuating axial uid velocity (at y) (cm/s)
RMS value of u' (cm/s)
friction velocity (cm/s)
radial particle velocity at the beginning of the freeight (at y s) (cm/s)
instantaneous radial uid velocity (at y) (cm/s)
time averaged mean radial uid velocity (at y) (cm/s)
uctuating radial uid velocity (at y) (cm/s)
RMS value of 0 (cm/s)
particle position during free-ight (cm)
distance from pipe wall (cm)
eddy diffusivity (at y) (cm2/s)
dynamic viscosity, g/cm-s poise [cP]
kinematic viscosity (cm2/s)
uid density, g/cm3 [kg/m3]
particle density, g/cm3, [kg/cm3)
shear stress at y (g/cm-s2)

performed, the authors concluded that both Beals (1970) and Escobedo
and Mansooris (1995) models were adequate to predict asphaltene
deposition.
1.1. Objectives
The main objective of the present work was to investigate the
contribution of diffusive and convective mechanisms to promote the
radial transport of asphaltene particles, complementing the rst part of
Shirdels et al. (2012) study with a more detailed investigation of
particle-uid interaction and with more published models and experimental data sets. Exclusively smooth vertical tubings were considered,
which nullied the inuence of pipe roughness and gravitational forces
on transport rates. Electrostatic effects were not addressed either,
although they are important for asphaltene deposition and their investigation is recommended to complement the present study.
More specically, the three objectives of this research were:
(i) to theoretically investigate the radial transport of asphaltene
particles, clarifying the main mechanisms that contribute for this
complex problem, (ii) to select an accurate model from literature
to predict asphaltene deposition rates, and, (iii) to perform a
sensitivity analysis with the selected model to identify the most
important parameters governing asphaltene deposition.
1.2. Methodology
The literature survey performed showed that the main line of
research applied to model asphaltene deposition is that derived from
the aerosol theory. That line considers convective and diffusive
mechanisms acting to promote the radial transport of asphaltenes,
with limited inuence of uid temperature in the process (contrary to
what proposes the line derived from the wax deposition theory). It
should be noticed, however, that there are substantial differences
between asphaltene and aerosol deposition, especially because of
the medium in which the particles are dispersed (liquid and gas,
respectively). Those differences should be properly addressed before
aerosol models and experimental data sets are applied in the study of
asphaltene deposition. Such analysis has not been done by previous

D.M. Paes et al. / Journal of Petroleum Science and Engineering 129 (2015) 7787

79

authors, and is performed in this paper as a part of the methodology

used. That methodology involved the following steps:
1. Reviewing fundamental concepts of momentum transfer in turbulent ow. Before starting to study particle deposition, it is
important to characterize and understand the medium in
which those particles are dispersed. In this step, the following
subjects were studied: turbulent eddies, hydrodynamic layers,
and momentum ux equations;
2. Reviewing fundamental concepts of mass transfer in turbulent ow.
The interaction between small submicron particles and the
uid was investigated, identifying and discussing the diffusive
and convective transport mechanisms that govern the radial
transport of those particles to pipe walls;
3. Reviewing fundamental concepts of the aerosol theory. Here, the
interaction of larger particles and the uid was investigated,
complementing the study performed in step 2. Applying concepts such as the particle stopping distance and the dimensionless relaxation time, the deposition of small and large particles
was compared and discerned;
4. Selecting, studying and implementing particle deposition models.
Models selected in the literature were carefully reviewed to
identify the transport mechanisms they each incorporated. Particular attention was given to verify whether any assumptions or
simplications were made during their formulation that would
cause the models to predict asphaltene deposition inadequately.
5. Selecting, studying and organizing aerosol deposition experimental data. Aerosol experimental works were selected in the literature
to provide data for a validation study of the models identied in step
4. The experimental methods used by the authors were carefully
reviewed and the published data organized on a standard format.
6. Validation of models. The models were applied to predict experimental deposition rates and errors were assessed. The results of the
analysis, along with the study performed in step 4, supported the
choice of one of the models to be used for asphaltene deposition.
7. Applying the selected model to predict asphaltene deposition. A
sensitivity analysis was performed to identify the impact of
uid and particle properties in asphaltene deposition rates. The
results obtained were discussed on the basis of the transport
mechanisms reviewed in steps2 and 3.
The studies performed on the above steps are described in the
next sections as follows: Fundamentals of Momentum and Mass
Transfer (steps 1, 2 and 3), Study of Particle Deposition (steps4,
5 and 6) and Study of Asphaltene Deposition (step 7).

2. Fundamentals of momentum and mass transfer

du
dy

and shows local axial and radial velocities represented in terms of a

time averaged and a uctuating value (u u0 and v v v0 , respectively). Since the ow is axial to pipe, the time averaged radial component, v, is null. By denition the mean values of the uctuating
velocities, u0 and v0 , are also zero. It should be noticed, however, that
neither the root mean square (RMS) value of those oscillatory components u0 ;RMS and v0RMS , nor their mean product, u0 v0 , are null (Bird et al.,
2002, 158163).
The turbulent eddies increase momentum transfer through the
uid, adding a new term, known as Reynolds stress, to Eq. (1):

where is the uid viscosity and u is the local uid velocity at a

distance y from pipe surface. Note that although Eq. (1) is written
in rectangular coordinates, it can be applied in the near-wall
region of circular tubes, where y{dt (dt is the tubing diameter).
Turbulent ow occurs for high speed streams, when Reynolds
numbers are greater than 10,000. It is characterized by the existence
of turbulent eddies, which are local random uctuations in the direction
and intensity of uid velocity. Fig. 1 illustrates those turbulent structures

du
u0 v0 :
dy

By analogy with Newtons law of viscosity, given by Eq. (1),

Boussinesq represented the Reynolds stress term as a function of
du/dy (Bird et al., 2002, 162). Considering this, Eq. (2) can be
written as:

Laminar ow occurs for low speed streams, when Reynolds

numbers are smaller than 2100 (for circular pipes). It is characterized by an orderly movement of the uid. In that ow regime,
momentum is transferred through the uid exclusively by viscous
action, and the shear stress, , can be evaluated with Newtons law
of viscosity, given by

Fig. 1. Features of turbulent ow: turbulent eddies and hydrodynamic regions.

du
du
;
dy
dy

where is the eddy diffusivity of momentum. Contrary to uid

viscosity, the eddy diffusivity is not an intrinsic property of the
uid, but a property of the turbulent ow. Unfortunately, cannot
be evaluated analytically, and the published correlations for this
parameter were obtained experimentally (Lin et al., 1953).
As it is represented in Fig. 1, the turbulent eddies frequency and
intensity varies along the ow section, dening three hydrodynamic
layers. The turbulent core is characterized by the presence of intense
turbulent eddies, which makes the Reynolds stress term in Eqs. 2 and
3 dominant in relation to the viscous term. The buffer layer represents a transition region, where eddies are of medium to low
intensity and momentum is transferred by both viscous and turbulent effects. The sublaminar layer, located close to pipe wall, is a calm
region in which eddies are rare, and momentum is transferred
almost exclusively by viscous action. The sublaminar layer and the
buffer region are considered to form the so called boundary layer.

80

D.M. Paes et al. / Journal of Petroleum Science and Engineering 129 (2015) 7787

the velocity uctuations, the convective term in Eq. (6) can be

expressed as a function of the eddy diffusivity, leading to:

2.1. Particle transport in turbulent ow

The radial ux of particles in turbulent streams arises from a
combination of both diffusive and convective mechanisms. The
diffusive transport results from the collisions between particles and
uid molecules caused by thermal agitation. Those collisions lead to
an erratic movement of the particles, known as Brownian motion.
Because the collisions are more frequent in high concentration zones,
the Brownian motion is responsible for a net ux of particles towards
low concentration zones, as Fig. 2 illustrates. That particle ux, N, can
be evaluated by the Ficks rst law of diffusion, as a function of the
particle concentration gradient, dC/dy:
N DB

dC
dy

The Brownian diffusivity, DB, can be assessed by the Stokes

Einstein equation, given by
K BT
DB
;
3dp

dC
C 0 v0 :
dy

Considering that the particles faithfully follow the turbulent

eddies path so that the concentration uctuations closely follow

dC
dC
:
dy
dy

As a result of eddy diffusion, transport in the turbulent core is

rapid, increasing the dispersion of the particles and leading to a at
concentration prole, similar to that observed for uid velocity
(represented in Fig. 1). In that region the viscous stress term and
the Brownian motion term in Eqs. (3) and (7), respectively, tend to be
negligible in comparison to the Reynolds stress and eddy diffusion
terms. That fact leads to similarities in the momentum and mass
transfer processes, known as Reynolds analogy. Mathematically, the
analogy is represented by the quotient between Eqs. (3) and (7), with
the viscous stress and Brownian motion terms neglected:

where KB is the Boltzmann constant, T is the uid absolute

temperature, and dp is the particle diameter. As Eq. (5) shows, DB
is inversely proportional to dp. Thus, as particles size increase, their
transport by Brownian diffusion tends to diminish while other
transport mechanisms become dominant.
The convective transport results from the interaction between
particles and turbulent eddies. In general, particles tend to be carried
by theeddies, following their path and being transported towards the
pipe wall, as Fig. 3 illustrates. Although it is a convective mechanism,
that type of transport is known in the literature as eddy diffusion.
The eddy diffusion is understood as a consequence of the uctuations
of local uid velocities, u' and '. Those uctuations in uid velocity
induce local oscillations also on particles concentration, C. Thus, C
can be written as a time averaged and an oscillatory component
C C C 0 . The eddy contribution for particle transport is considered
by adding a convective term to the Ficks rst law, as follows (Bird
et al., 2002, 657659):
N DB

N DB

du=dy
dC=dy

Approaching the wall, turbulence is progressively damped until it

disappears in the immediate neighborhood ofpipe surface, and the
transport is almost exclusively by Brownian motion. The main resistance to particle transport occurs in the boundary layer, where the
concentration gradient is maximum, as it was represented in Fig. 2.
The Schmidt number is a dimensionless parameter frequently
used to characterize ows in which there are simultaneous
momentum and mass transfer processes. It is dened by the ratio
between uid kinematic viscosity, , and Brownian diffusivity:
NSc

DB

In the analysis made before, including the modied Ficks law

(Eq. (7)), it was assumed that the particles were carried by the turbulent
eddies without slippage, rigorously following their path. Although this
is a reasonable assumption for the convective transport of molecules or
small particles, it is not suitable for large particles, which may experience slippage in relation to the uid because of their inertia. Two
theoretical parameters are used in the literature to quantify and model
particle-uid slippage: the stopping distance and the relaxation time.
2.2. Stopping distance
The particle stopping distance is dened as the distance a sphere
(mass, mp, diameter, dp, and density, p), with an initial velocity V0,
travels in free-ight through a stagnant uid before it stops because of
drag forces. Assuming that the sphere travels in the Stokes regime, its
force balance results in
2

mp
Fig. 2. Transport mechanisms: Brownian motion. Particles are transported due to
the collisions between them and the uid molecules, caused by thermal agitation.

d x
dx
 3dp ;
dt
dt 2

10

where x is the sphere position at a given time t. The spheres velocity

during free-ight can be evaluated integrating Eq. (10) with the
boundary condition dx/dt(t0)V0 and considering that the spheres
mass is given by d3pp/6:
dx
V 0 e  t=t p ;
dt

11

where tp is the particle relaxation time, dened as

tp

Fig. 3. Transport mechanisms: eddy diffusion. Particles are carried by the turbulent
eddies.

p dp 2
:
18

12

The spheres position can be evaluated integrating Eq. (11) with

the boundary condition x(t 0) 0:



13
x t p V 0 1  e  t=t p :

D.M. Paes et al. / Journal of Petroleum Science and Engineering 129 (2015) 7787

81

The stopping distance, s, can be evaluated from Eq. (13),

considering that x approaches s when t tends to innity:
s tpV 0:

14

The rst authors that applied the stopping distance concept to

model particle deposition were Friedlander and Johnstone (1957).
They veried that the use of Eq. (7) underestimated deposition
rates of large particles and attributed the problem to particle-uid
slippage in the near-wall region. According to their theory,
particles followed the intense eddies in the turbulent core, thus
being launched by them towards the boundary layer with high
momentum. Because the near-wall eddies were of low duration
and intensity, they would not be able to interfere in the particles
path, and the particles would make their nal trip to deposit in
free-ight, as Fig. 4 illustrates.V0 is the particle radial velocity at
the position s, where it starts the free-ight movement. The
authors assumed that this velocity was equal to the RMS value
of the radial uctuating velocity, v0RMS , and used Laufers (1953)
experimental data to correlate it.

Fig. 5. Particle deposition diagram: typical distribution of experimental data.

Changes in dimensionless relaxation times, caused mainly by variations in particle
size and ow velocity, can lead to different deposition regimes.

ow, Cavg:
N0 K d C avg :

2.3. Relaxation time

Eq. (11) shows that the particle relaxation time, tp, is the time
for the velocity of a free-ight particle to decay to 1/e (or 36.8%) of
its initial value. This is regarded as the characteristic time particles
take to respond to changes in uid velocity, being considered a
measure of particle inertia. In general, particle motion is not
affected by eddies with lifetimes shorter than tp. The lifetime of
the near-wall eddies can be evaluated by the equation (Sippola
and Nazaroff, 2002, 14):
te

U 2 f F =2

15

where U is the average ow velocity and fF is the Fanning friction

factor.
The occurrence of slippage in the near-wall region can be
evaluated comparing tp to te. The ratio tp/te is known as dimensionless relaxation time:
t

p dp 2 U 2 f =2
;
18

16

That dimensionless parameter provides a quantitative measure

of particle-uid slippage and can be applied to classify experimental data in three deposition regimes (Sippola and Nazaroff,
2002, 16): diffusion (t o0.1), diffusionimpaction (0.1 rt o10),
and inertia-moderated (t 410).
2.4. Particle deposition regimes
The particle deposition rate, N0, is dened as the amount of
particles depositing in a known area of a pipe during a known
period of time. It is equal to the radial particle ux, N, evaluated at
the wall vicinity (y0). It can be expressed as the product of a
transport coefcient, Kd, by the average particle concentration on

Fig. 4. Transport mechanisms: particle inertia. Particles are launched in free-ight.

17

The transport coefcient incorporates the physical mechanisms

acting to promote particle deposition. That parameter has units of
velocity and is also known as deposition velocity. It can bepmade

dimensionless by dividing it by the friction velocity un U f F =2:

Kd

K d p
U f F =2

18

Particle deposition rates are frequently reported in the form of Kd

versus t . When those parameters are used to plot experimental data
in a logarithmic graph, the result is a characteristic s-shaped curve,
similar to that represented in Fig. 5. Eq. (16) shows that t is a
function of both uid and particle properties. In the following
discussion about the different trends and deposition regimes observed
on Fig. 5, it will be considered that only particle diameter varies, so
that changes in t should be attributed to changes in particle size.
The diffusion regime occurs for small particles, with t o0.1. Such
small values of dimensionless relaxation times indicate that particles
respond to turbulent uctuations in a much shorter period of time
than theeddies lifetime. Hence, particles are easily carried by the
eddies and particle-uid slippage is almost inexistent. Thus, particle
inertia tends to be negligible while Brownian motion and eddy
diffusion are dominant in that regime. Increasing particle sizes
(or t ) decreases the deposition velocities, since the Brownian
diffusivity is inversely proportional to particle diameter (see Eq. (5)).
The diffusionimpaction regime occurs for particles of intermediate sizes, with 0.1rt o10. In that range of t , the time particles take
to respond to turbulent uctuations is of the same magnitude order of
the near-wall eddies lifetime. As a result, slippage starts to occur
within the boundary layer, and particle inertia starts to be dominant
while Brownian motion turns out to be less important. As the sizes of
particles increase, deposition velocities increase sharply because
particle stopping distance is proportional to d2p (see Eq. 14).
The inertia-moderated regime occurs for large particles, with
t Z10. Such large values of dimensionless relaxation time indicate
that the time particles take to respond to turbulent uctuations is
much longer than the eddies lifetime. As a result, slippage starts to
occur not only in the near-wall region, but also in the turbulent core.
Increasing particle sizes, their transport in the turbulent core (by eddy
diffusion) become less efcient and they reach the boundary layer
with lower speeds, what reduces the deposition velocities.
To verify the regimes in which asphaltene deposition occurs,
dimensionless relaxation times were assessed for the ranges of
eld parameters shown in Table 1. The diameters of the asphaltene
precipitates were assumed to vary between 1 nm and 30 mm. Only
turbulent ow was considered in the analysis, with ow rates ranging

82

D.M. Paes et al. / Journal of Petroleum Science and Engineering 129 (2015) 7787

Table 1
Ranges of eld parameters.
Parameter

Range considered

dp(m)
U (m/s)
(cP)
(kg/m3)
p (kg/m3)
T (1C)
dt (in.)
NRe
NSc
t

0.001 to 30
1 to 7
0.8 to 40
700 to 900
1 200
50 to 200
1 to 4
10  103 to 1  106
9.2  102 to 8.1  1010
1013 to 10

from 100 to 30,000 bpd and Reynolds numbers from 10,000 to

1  106. Combining the parameters of Table 1, Schmidt numbers were
found to vary from 9.2  102 to 8.1  1010 while t varied from
1  1013 to 10, showing that asphaltene deposition may occur
preferentially in the diffusion and diffusionimpaction regimes, as it
was illustrated in Fig. 5. It should be noticed, however, that the
inertia-moderated regime can also happen if unusual light uids,
large particles, or high ow rates occur.

3. Study of particle deposition

After reviewing basic concepts of momentum and mass transfer in
turbulent ow, six particle deposition models were selected (Lin et al.,
1953; Friedlander and Johnstone, 1957; Beal, 1970; El-Shobokshy and
Ismail, 1980; Papavergos and Hedley, 1984; Escobedo and Mansoori,
1995) to study the different approaches used in literature to account
for the transport mechanisms reviewed in this paper. To allow the
comparison of the models performance, four aerosol deposition
experimental works (Friedlander, 1954; Wells and Chamberlain,
1967; Liu and Agarwal, 1974; Agarwal, 1975) were selected from the
literature and used for a validation study.
3.1. Particle deposition models
The methodology to model particle deposition involves three
steps: writing equations for the particle ux, N, in each hydrodynamic region (sublaminar layer, buffer region and turbulent core);
integrating those equations with the aid of proposed assumptions
and boundary conditions; mathematically manipulating the results
to nd an equation for the deposition velocity, Kd, as a function of
particle and uid properties.
Lin et al. (1953) studied the deposition of small submicron
particles. They proposed that those particles were radially transported by two mechanisms: (i) Brownian motion: the collisions
between particles and uid molecules, caused by thermal agitation,
would lead to an erratic movement of the particles towards low
concentration zones (pipe walls vicinity), (ii) Eddy diffusion: the
turbulent eddies of the uid would carry the particles from the
turbulent core towards low turbulence zones (pipe walls vicinity).
Considering the inexistence of particle-uid slippage for such small
particles, the authors modeled the ux N in the boundary layer with
the modied Ficks law (Eq. (7)). In the turbulent core, the Reynolds
analogy (Eq. (8)) was applied. To integrate those equations, the
particle concentration at the wall was equaled to zero, assuming that
all particles that reach the wall were attached and that no reentrainment occurred.
Friedlander and Johnstone (1957) studied the deposition of large
particles. They proposed that those particles were radially transported
by two mechanisms: (i) Eddy diffusion: particles would be carried by
the turbulent core eddies, being launched by them towards the

boundary layer with high velocities, (ii) Particle inertia: at a certain

position within the boundary layer (equal to the stopping distance),
particles would initiate a free-ight movement towards pipe walls
(because the near-wall eddies would not have enough intensity to
interfere in the motion of such large particles). Considering that the
transport of the particles caused by thermal agitation is impaired by
increases in their sizes, the authors modeled the ux N in the
boundary layer using the modied Ficks law with the Brownian
motion term neglected. In the turbulent core, the Reynolds analogy
was applied. To incorporate particle inertia to the model,the stopping
distance was used as the lower integration limit of the modied Ficks
law. The particle concentration at that position was assumed null, due
to the no re-entrainment hypothesis.
Beal (1970) combined aspects of previous theories to develop a
model suitable for both small and large particles. The author considered that small particles would be transported by Brownian motion
and eddy diffusion, as it was proposed by Lin et al. (1953), while large
particles would be transported by eddy diffusion and particle inertia,
as it was proposed by Friedlander and Johnstone (1957). To model the
ux N, the Reynolds analogy was applied in the turbulent core and the
modied Ficks law (without any simplication) in the boundary layer,
using the stopping distance as the lower integration limit. Contrary to
Friedlander and Johnstones (1957) assumption, the author postulated
that the particle concentration at the stopping distance was not null.
Instead, it was assumed a constant concentration in the free-ight
region, which would reduce deposition rates.
El-Shobokshy and Ismail (1980) also studied the deposition of
small and large particles. They attributed the radial transport of those
particles to the same mechanisms proposed by previous authors: a
combination of Brownian motion and eddy diffusion for small
particles and a combination of eddy diffusion and particle inertia for
large particles. While Friedlander and Johnstone (1957) and Beal
(1970) considered that particle-uid slippage would occur only in
the free-ight region, the authors assumed that it occurred though the
entire boundary layer. Particles would be transported by the turbulent
eddies from the center of the pipe to the position of the stopping
distance, always with some degree of slippage. Once they reach the
stopping distance, they would make their nal trip towards pipe walls
in free-ight. To incorporate those assumptions to the model, the
authors proposed that uid and particles had different diffusivities
(p a ). The ux N in the boundary layer was then modeled with the
modied Ficks law, using p in place of (the equations of Liu and Ilori
(1973) were used to correlate the particle diffusivity with the uid
diffusivity). The stopping distance was used as the lower integration
limit of that equation and a constant particle concentration was
assumed in the free-ight region, as it was proposed by Beal (1970).
In the turbulent core, the authors postulated that the particles were
homogeneously dispersed (CCavg) because of the intense turbulence.
Papavergos and Hedley (1984) adopted a different methodology to
nd equations for the deposition of small and large particles. They
reviewed a wide set of experimental data and organized them in
graphs of dimensionless deposition velocities versus dimensionless
relaxation times (Kd versus t ). The selected data were then separated
in different regimes, according to t values: diffusion regime (small
particles, with t o0.1), diffusionimpaction regime (intermediate
size particles, with 0.1rt o10), and inertia-moderated regime (large
particles, with t Z10). The equations to predict deposition rates in
each regime were found by adjusting curves to the selected data.
Escobedo and Mansoori (1995) also investigated the deposition
of small and large particles. Small particles were considered to be
transported by Brownian motion and eddy diffusion. Large particles were considered to acquire high momentum in the turbulent
core due to the action of the turbulent eddies, making their nal
trip to reach pipe walls in free-ight. The main difference of this
model in relation to previous ones is the application of the
modied Ficks law not only in the boundary layer, but also in

D.M. Paes et al. / Journal of Petroleum Science and Engineering 129 (2015) 7787

the turbulent core (previous authors used the Reynolds analogy in

this region). The authors used the stopping distance as the lower
integration limit of that equation and a constant particle concentration was assumed in the free-ight region, as it was proposed
by Beal (1970).
After studying the theory of the models, we performed a
detailed review of their formulation. Starting from the differential
equations for N, all integrations and mathematical developments
were repeated to arrive on the nal equations for the deposition
velocity. That study was of great importance because it allowed a
clear understanding of the assumptions and simplications made
by the authors during the deduction of each model.

3.2. Particle deposition experimental data

The studied experimental data refer to the deposition of aerosol
particles to smooth vertical surfaces during turbulent ow. In all
studied experiments, the authors monitored and avoided particle reentrainment, taking measures to guarantee that all particles that
reached pipe walls were kept attached. Table 2 summarizes the main
features of particles and tubes used in the experiments. It also
contains the ranges of NRe, NSc and t for each experiment, showing
that the three deposition regimes were covered.
In general, the laboratory apparatus used to investigate aerosol
deposition consisted of an aerosol generator, responsible for putting
the small solid or liquid particles in suspension in the air stream,
followed by an observation section (composed by the tubes listed on
Table 2), where deposition was monitored. At the end of each run,
the amount of particles attached to pipe walls, from which deposition
rates were calculated, was determined in three different manners.
Friedlander (1954) used a microscope to count the particles deposited on a known area of the tubes. Wells and Chamberlain (1967)
previously tagged the particles with radioactive material and
assessed the activity on tube walls (demountable sections) in a well
of a beta scintillation counter. Liu and Agarwal (1974) and Agarwal
(1975) tagged the particles with uranine, a uorescent dye, and used
a uorometer to evaluate the amount of uranine on tubes.
The results of those experiments were published using different
parameters, notations and measurement units. To organize and
enable the comparison of the data obtained by the different authors,
the reviewed results were converted to a standardized format (Kd
versus t ) and plotted in Fig. 6. Although some scattering is veried,
the experimental data in that gure tend to be distributed in the
characteristic s-shaped curve, presented before on Fig. 5. According
to the values of dimensionless relaxation time, the selected data can
be divided into three deposition regimes: diffusion (8 data points),
diffusionimpaction (75 data points) and inertia-moderated (49 data
points).

83

3.3. Validation study

The criterion used for the comparison of the models predictions with experimental data was the average error, dened as
Pn
E% 100

i1






=K d_exp
K d_exp  K d_mod

;
n

19

where K d_exp
and K d_mod
are, respectively, the dimensionless
deposition velocities measured experimentally and predicted by
the models; n represents the number of points used for the error
evaluation.
The rst three rows of Table 3 show the average errors of the
models separated by deposition regime. As it can be seen, the best
predictions in the diffusion regime were provided by Papavergos
and Hedley (1984) and Beal (1970), with average errors of 20.1%
and 23.5%, respectively. In the diffusionimpaction regime, the
best results were provided again by the same models, with
average errors of 37.6% and 38.7%, respectively. None of the models
presented reasonable predictions in the inertia-moderated regime,
where Friedlander and Johnstone (1957) had an average error of
56.3% while other models exhibited errors greater than 70%.
Dimensionless relaxation times calculated before for the parameters
of Table 1 ranged from 1013 to 10, indicating that asphaltene deposition
occurs preferentially in the diffusion and diffusionimpaction regimes
(represented in Fig. 5). Small precipitates tend to be deposited in the
diffusion regime, in which Brownian motion and eddy diffusion arethe
dominant transport mechanisms acting. As particles size increase, the
transition to the diffusionimpaction regime occurs, and Brownian
motion becomes negligible while particle inertia acquires signicant
importance. Fig. 7and the fourth row of Table 3 show the validation of
the models considering those two regimes together. The best results

Fig.6. Particle deposition diagram: distribution of the reviewed experimental data.

Table 2
Studied aerosol experiments.
Particles
Iron (0.8; 1.57; 1.81; 2.63 m)
Aluminum (1.81 m)
Lycopodium spores (30 m)
Wells and Chamberlain (1967) Aitken nuclei (0.17 m)
Friedlander (1954)

Liu and Agarwal (1974),

Agarwal (1975)

Tri-cresyl-phosphate droplets (0.65;

1.1; 2.1 m)
Polystyrene spheres (5 m)
Olive oil (1.4 to 21 m)

Tubes

NRe

NSc
3

t
5

Glass (0.54; 0.58; 1.305; 2.5 cm) From 3.1  10 to

Brass (1.38; 2.5 cm)
3.5  104

From 3.7  10 to
4.5  107

From 0.22 to 17

Annular space*: Brass rod

(OD 1.27 cm)
Cooper tube (ID 3.81 cm)

From 1  103 to
5  104

From 1.1  105 to

3.2  106

From 4.6  10-4

to 11.9

Glass (0.327 to 1.27 cm)

Cooper (1.38 cm)

From 3.5  103 to

6  104

From 1.8  106 to

1.4  1011

From 0.21 to 774

n
The deposition surface used by Wells and Chamberlain (1967) was a brass rod placed axially in a cooper tube. For all calculations it was considered a hydraulic diameter
of (3.811.27) 2.54 cm.

84

D.M. Paes et al. / Journal of Petroleum Science and Engineering 129 (2015) 7787

Table 3
Average errors of the studied particle deposition models.
Deposition regimes

Lin et al.
(1953)

Friedlander and Johnstone

(1957)

Beal
(1970)

El-Shobokshy and Ismail

1980

Papavergos and Hedley

1984

Escobedo and Mansoori,

1995

Diffusion
Diffusionimpaction
Inertia-moderated
Diffusion and diffusion
impaction

30.5%
na*
na*
na*

na**
55.4%
56.3%
na**

23.5%
38.7%
470%
37.2%

4 70%
47.5%
na
52.0%

20.1%
37.6%
470%
35.9%

38.8%
470%
na
68.6%

Lins et al. (1953) model is not applicable in the diffusionimpaction and inertia-moderated regimes, because it does not account for particle inertia.
Friedlander and Johnstones (1957) model is not applicable in the diffusion regime, because it does not account for Brownian motion.
El-Shobokshy and Ismail (1980) and Escobedo and Mansoori (1995) did not propose deposition equations when the stopping distance is in the turbulent core.

nn

Table 4
Base case of asphaltene deposition.

Fig. 7. Particle deposition diagram: validation of the studied models with the
selected experimental data.

were obtained by Papavergos and Hedley (1984) and Beal (1970), with
average errors of 35.9% and 37.2%, respectively. Although Papavergos
and Hedley (1984) presented a lower average error, we recommend the
use of Beal (1970) due to the following reasons: it was developed on the
basis of consolidated physical assumptions, while Papavergos and
Hedleys (1984) equations are simple curve ttings to experimental
data; it provided a smooth transition between diffusion and diffusion
impaction regimes, while the model of Papavergos and Hedley (1984)
presented a discontinuity at t 0.1.

4. Study of asphaltene deposition

According to the previous section, Beals (1970) model was considered the most suitable to predict particle deposition in the
diffusion and diffusionimpaction regimes. Although the model was
validated with aerosol data, the study of its theory showed that no
simplications or assumptions that would make it inappropriate to
liquid streams were adopted during its formulation. In this paper, we
assumed that the model is adequate to predict the radial transport of
asphaltene particles, provided that it is used in similar ranges of NRe,
NSc and t in relation to those covered on the validation study
(10  103 rNRe r6  104; 1.1  105 rNSc r1.4  1011; 0.21rt r10
considering only the diffusion and diffusionimpaction regimes in
turbulent ow).
The model was then applied in the study of asphaltene deposition. For that study, a hypothetic base case was proposed. The
parameters shown in Table 4 were chosen considering the typical
scenario in which deposition occurs: high speed turbulent streams of
light oil. The deposition velocity calculated with the selected model
for those data was 1.19  10-4 cm/s. Starting from the base case, two

Parameter

Value

dp (mm)
U (m/s)
(cP)
(kg/m3)
p (kg/m3)
T (1C)
dt (in.)

2
5
2
740
1200
85
3

sensitivity analysis were performed: (i) deposition velocities were

assessed applying a 710% variation to each of the entry parameters
shown in Table 4, (ii) deposition velocities were assessed for a
broader range of entry parameters, shown in Table 1.
Table 5 brings all relevant information about the rst sensitivity
analysis: range of entry parameters (rst column), range of calculated
deposition velocities (second column), and percentage variation of
deposition velocities in relation to the base case (third column). The
results obtained showed that the main parameters governing deposition rates were: ow velocity (Kd percentage variation of 49.3%),
particle diameter (34.6%), uid viscosity (32.8%), and uid density
(29.8%). Pipe diameter and uid temperature had a secondary
importance (7.1% and 0.4%, respectively). Those results are in
accordance with those obtained in the second sensitivity analysis,
which were presented in Figs. 813. The variation of precipitates
diameter, oil viscosity and ow velocity over the broader ranges
showed on Table 1 had a great impact on deposition rates, causing
changes in deposition velocities of more than four orders of
magnitude. In the next paragraphs, the inuence of each entry
parameter in the transport of the asphaltenes is discussed considering the results obtained in the two sensitivity analysis.
Fig. 8 shows the impact of asphaltene sizes upon deposition,
considering four average ow velocities. The variation of precipitate
diameters in the interval 1 nmrdp r30 m made deposition velocities vary by more than four orders of magnitude. To support the
interpretation of those results, two additional curves were plotted for
U1 m/s in Fig. 8. The dotted line represents Beals (1970) model
with particle inertia neglected while the dashed line represents the
model with Brownian motion neglected. Starting at the left, the
deposition of small precipitates is entirely controlled by Brownian
motion and eddy diffusion. Deposition rates decrease with particle
diameter, because Brownian motion is inversely proportional to the
size of the transported solids. However, increasing the size (mass)
increases momentum so that particle inertia gains importance. The
deposition rates reach a minimum, beyond which, the process is
essentially momentum controlled. That minimum occurs roughly
between 0.5 and 5 m (depending on ow velocities) and characterizes the transition from the diffusion to the diffusionimpaction
regime. Thus, a reasonable approximation for asphaltene deposition
is that submicron particles tend to deposit in the diffusion regime

D.M. Paes et al. / Journal of Petroleum Science and Engineering 129 (2015) 7787

85

Table 5
Sensitivity analysis.
Parameter

Range considered

Calculated Kd
(10-4 cm/s)

Percentage variation (%)

dp (mm)
U (m/s)
(cP)
(kg/m3)
T (1C)
dt (in.)

1.8 to 2.2
4.5 to 5.5
1.8 to 2.2
666.0 to 814.0
76.5 to 93.5
2.7 to 3.3

0.10 to 1.41
0.92 to 1.51
1.41 to 1.02
1.02 to 1.37
1.19 to 1.19
1.23 to 1.15

34.6
49.3
32.8
29.8
0.4
7.1

Fig. 9. Sensitivity analysis: impact of changes in ow velocity upon deposition

velocities.

Fig. 8. Sensitivity analysis: impact of changes in precipitates diameter upon

deposition velocities.

while particles of micra or tens of micra tend to deposit in the

diffusionimpaction regime.
Considering that approximation, the following methodology was
applied to express results on the next graphs: submicron particles
(0.001 and 0.01 m), depositing in the diffusion regime, were represented by dotted lines; particles of micra and tens of micra (10 and
30 m), depositing in the diffusionimpaction regime, were represented by solid lines; 1.3 m particles, depositing in the transition
region between those two regimes, were represented by dashed lines.
Fig. 9 shows the inuence of ow velocity upon deposition rates.
The variation of the parameter in the interval 1rUr7 m/s raised
deposition velocities of large particles (10 and 30 m) by more than
four orders of magnitude. Greater ow velocities lead to higher
turbulence intensities, what increase transport rates in the turbulent
core by eddy diffusion. As a consequence, large particles reach the
boundary layer with higher velocities, and their deposition in freeight due to particle inertia is enhanced. Submicron particles also tend
to reach the boundary layer with higher velocities. However, they are
so small that their momentum is still limited, and they keep following
the turbulent vortices till reaching the tubing surface. The slight
increase in deposition velocities observed for those particles is caused
exclusively by the increase of eddy diffusion.
Fig. 10 shows the inuence of oil viscosity upon deposition rates.
The variation of that parameter in the interval, 0.8r r40 cP,
reduced deposition velocities by more than four orders of magnitude.
Small submicron particles transport by Brownian motion is impaired
by higher viscosities, because Brownian diffusivity is inversely proportional to the parameter. In systems composed by large particles,
deposition rates decay even faster, because increases in uid viscosity
raise the drag forces acting on the free-ight particles, reducing their
stopping distance.
Fig. 11 presents the inuence of oil density upon deposition rates.
That gure portraits a limited impact of that parameter on particle
transport, contrary to what was observed in the rst sensitivity
analysis, when changes of 710% in oil density made deposition

Fig. 10. Sensitivity analysis: impact of changes in oil viscosity upon deposition
velocities.

Fig. 11. Sensitivity analysis: impact of changes in oil density upon deposition
velocities.

velocities vary 29.8%. It is expected that increases in oil density raise

deposition, since eddiesactivity is favored by higher densities ( is
directly proportional to ). The less pronounced effect of that parameter in the second sensitivity analysis is due to the fact that the
range in which uid density can vary in industry is considerably narrower than the range in which other parameters can vary

86

D.M. Paes et al. / Journal of Petroleum Science and Engineering 129 (2015) 7787

Fig. 12. Sensitivity analysis: impact of changes in oil temperature upon deposition
velocities.

wider set of experimental data in relation to previous works (including Shirdels et al., 2012).
Beals (1970) model was assumed suitable to represent asphaltene
deposition provided that it was used in similar ranges of NRe, NSc and
t in relation to those for which it was validated. Sensitivity analysis
performed with the model showed that deposition velocities can vary
up to four orders of magnitude with asphaltenes diameter, ow
velocity and uid viscosity. Other parameters, such as tubing diameter, oil density and temperature, had minor inuence on deposition rates. Small primary asphaltene precipitates tend to deposit in
the diffusion regime, in which Brownian motion and eddy diffusion
are the dominant transport mechanisms. Large aggregates tend to
deposit in the diffusionimpaction regime, in which eddy diffusion
and particle inertia are the main mechanisms acting. The transition
between those regimes was found to occur roughly around 1 m and
was characterized by a minimum in deposition rates.
The deposition of large asphaltene aggregates (dp 41 m) can be
readily diminished by reducing production rates (ow velocity),
because both eddy diffusion and particle inertia are minimized by
that preventive measure. The same does not occur with small
submicron precipitates, for which deposition rates are little affected
by reductions in ow velocity (because particle inertia tends to be
negligible for those particles, independent of ow velocity).
Light oils tend to be more problematic, not only for asphaltene
precipitation, but also for its deposition. The radial transport of both
small and large particles is enhanced in uids of low viscosity,
because Brownian motion and particle inertia are favored when drag
forces applied by the uid on particles are minimized. This fact is in
accordance with historical observation that light oil elds are more
prone to develop problems with asphaltene deposition.
This paper contributed for a better understanding of asphaltene
deposition. However, there is still much to be investigated about this
complex mass transfer problem. The inclusion of electrostatic, thermal
and adhesion/reentrainment effects in the model of Beal (1970) would
make it more complete and accurate to predict deposition rates.

Fig. 13. Sensitivity analysis: impact of changes in tubing internal diameter upon
deposition velocities.

Acknowledgement

(precipitates diameter, ow velocity and oil density can vary more

than 100%).
Fig. 12 shows the inuence of oil temperature upon deposition
rates. The variation of that parameter in the interval, 50rTr200 1C,
did not affect deposition signicantly. For small precipitates, higher
temperatures led to slight increases in transport rates, which were
caused by the growth of thermal agitation and, consequently, Brownian
motion. For large aggregates, changes in oil temperature did not affect
deposition rates. Fig. 13 shows the inuence of tubing diameter upon
deposition rates. That parameter does not affect the studied transport
mechanisms, and did not have impact on deposition velocities.

5. Conclusions
In this paper, six particle deposition models were selected from
literature and validated with four experimental data sets, also from
literature. According to calculated average errors and to the study of
the models theory, Beal (1970) was selected to be used in the
diffusion and diffusionimpaction regimes. That choice is in agreement with previous results presented by Shirdel et al. (2012), who
elected the models of Beal (1970) and Escobedo and Mansoori (1995)
the most suitable to predict deposition. The preference by the rst
model in the present paper is due to the lower average error it
presented in the validation study performed, which considered a

The authors would like to acknowledge the Petroleum Engineering

Department of the Universidade Estadual de Campinas (UNICAMP)
and the Coordenao de Aperfeioamento de Pessoal de Nvel Superior
(CAPES, Brazil) for providing the necessary resources for this research.
The support from the Center for Petroleum and Geosystems Engineering at the University of Texas at Austin is also greatly appreciated.
References
Agarwal, J.K., 1975. Aerosol Sampling and Transport Ph.D. Dissertation. University
of Minnesota, Minneapolis, Minnesota, June 1975.
Beal, S.K., 1970. Deposition of particles in turbulent ow on channel or pipe walls.
Nucl. Struct. Eng. 40, 111.
Bird, R.B., Stewart, W.E., Lightfoot, E.N., 2002. Transport Phenomena. John Wiley &
Sons, New York, NY.
Cleaver, J.W., Yates, B., 1975. A sublayer model for the deposition of particles from a
turbulent ow. Chem. Eng. Sci. 30 (8), 983992. http://dx.doi.org/10.1016/
0009-2509(75)80065-0.
El-Shobokshy, M.S., Ismail, I.A., 1980. Deposition of aerosol particles from turbulent
ow onto rough pipe wall. Atmos. Environ. 14 (3), 297304. http://dx.doi.org/
10.1016/0004-6981(80)90063-3.
Escobedo, J., Mansoori, G.A.. 1995. Solid particle deposition during turbulent ow
production operations. In: Paper SPE 29488 Presented at the SPE Production
Operations Symposium, Oklahoma City, Oklahoma, 24 April. http://dx.doi.
org/10.2118/29488-MS.
Eskin, D., Ratulowski, J., Akbarzadeh, K., et al. 2009. An approach to modeling
asphaltene deposition in a turbulent pipe ow on the basis of a Couette device
experimental data. In: Proc., Multi-Scale Modeling Symposium for Industrial
Flow Systems, Melbourne, Australia, 78 December.
Friedlander, S.K., 1954. Deposition of Aerosol Particles from Turbulent Gases Ph.D.
Dissertation. University of Illinois, Urbana, Illinois (July 1954).

D.M. Paes et al. / Journal of Petroleum Science and Engineering 129 (2015) 7787

Friedlander, S.K., Johnstone, H.F., 1957. Deposition of suspended particles from

turbulent gas streams. Ind. Eng. Chem. 49 (7), 11511956. http://dx.doi.org/
10.1021/ie50571a039.
Jamialahmadi, M., Soltani, B., Muller-Steinhagen, H., et al., 2009. Measurement and
prediction of the rate of asphaltene deposition of occulated asphaltene
particles from oil. Int. J. Heat Mass Transfer 52, 46244634. http://dx.doi.org/
10.1016/j.ijheatmasstransfer.2009.01.049.
Kokal, S.L., Sayegh, S.G. 1995. Asphaltenes: the cholesterol of petroleum. In: Paper
SPE 29787 Presented at the SPE Middle East Oil Show, Bahrain, 1114 March.
http://dx.doi.org/10.2118/29787-MS.
Laufer, J., 1953. The Structure of Turbulence in Fully Developed Pipe Flow. Technical
Note 2954. National Advisory Committee for Aeronautics, Washington (June
1953).
Leontaritis, K.J., Mansoori, G.A., 1988. Asphaltene deposition: a survey of eld
experiences and research approaches. J. Petrol. Sci. Eng. 1 (3), 229239. http:
//dx.doi.org/10.1016/0920-4105(88)90013-7.
Lin, C.S., Moulton, R.W., Putnam, G.L., 1953. Mass transfer between solid wall and
uid streamsmechanism and eddy distribution relationships in turbulent
ow. Ind. Eng. Chem. 45 (3), 636640.
Liu, B.Y.H., Agarwal, J.K., 1974. Experimental observation of aerosol deposition in
turbulent ow. J. Aerosol Sci. 5 (2), 145155. http://dx.doi.org/10.1016/00218502(74)90046-9.

87

Liu, B.Y.H., Ilori, T.A., 1973. Aerosol deposition in turbulent pipe ow. Environ. Sci.
Technol. 8 (4), 351356. http://dx.doi.org/10.1021/es60089a001.
Papavergos, P.G., Hedley, A.B., 1984. Particle deposition behavior from turbulent
ows. Chem. Eng. Res. Des. 62, 275295.
Ramirez-Jaramillo, E., Lira-Galeana, C., Manero, O., 2006. Modelingasphaltene
deposition in production pipelines. Energy Fuels 20 (3), 11841196. http://dx.
doi.org/10.1021/ef050262s.
Shirdel, M., Paes, D., Ribeiro, P., et al., 2012. Evaluation and comparison of different
models for asphaltene particles deposition in ow streams. J. Petrol. Sci. Eng.
8485, 5771. http://dx.doi.org/10.1016/j.petrol.2012.02.005.
Sippola, M.R., Nazaroff, W.W. 2002. Particle Deposition from Turbulent Flow:
Review of Published Research and its Applicability to Ventilation Ducts in
Commercial Buildings. Lawrence Berkeley National Laboratory Report: LBNL51432.
Watkinson, P., Epstein. 1970. Particulate fouling of sensible heat exchanges. In:
Fourth International Heat Transfer Conference, Versailles, France, 31 August5
September.
Wells, A.C., Chamberlain, A.C., 1967. Transport of small particles to vertical surfaces.
Br. J. Appl. Phys. 18, 17931799. http://dx.doi.org/10.1088/0508-3443/18/12/
317.