Journal of Petroleum Science and Engineering 175 (2019) 519–535

Contents lists available at ScienceDirect

Journal of Petroleum Science and Engineering

journal homepage: [Link]/locate/petrol

On the numerical simulation of sand transport in liquid and multiphase T

pipelines
Mariella Leporinia,∗, Alessandro Terenzib, Barbara Marchettic, Francesco Corvarod,
Fabio Polonarad,e
a
Politecnico di Milano, Italy
b
Saipem, SpA, Italy
c
University degli Studi e-Campus, Como, Italy
d
University Politecnica delle Marche, Ancona, Italy
e
Consiglio Nazionale delle Ricerche, Construction Technologies Institute, Viale Lombardia 49, San Giuliano Milanese, (MI), 20098, Italy

A R T I C LE I N FO A B S T R A C T

Keywords: Sand deposition in pipelines represents one of the problems that can arise in oil and natural gas production and
Sand deposition transport. The main problems that sand can cause include the pipeline obstruction with the relative production
Flow assurance loss, the increase of the lines erosion and corrosion and the compromising of the structural integrity. Moreover,
Oil and gas there is a limit in the amount of sand that can be separated and removed. Therefore, methods and strategies to
Multiphase
reduce the production of sand are key factors for safety and economic risks. In general sand deposition problems
occur in different flow systems such as sand–multiphase mixtures of gas-oil-water or two-phase mixtures. In
order to elaborate an effective method able to predict the amount of sand and to have a reliable prediction of
sand transport velocity and entrainment processes, several factors requires to be considered, including sand
characteristics, sand concentration, flow regimes, fluid properties and pipe properties.
This paper presents the results of a detailed testing on the performance of a new sand transport model
implemented in one-dimensional dynamic multiphase code, performed by comparing numerical results with
experimental data. This study deals with both liquid-solid flow as well as gas-liquid-solid flow. The results
demonstrate a good agreement between numerical and experimental data for the sand-liquid flow and sand
transport in stratified gas-liquid flow, while show that for the sand transport in gas-liquid slug flow improve-
ments are necessary.

1. Introduction formation sand grains, pore pressure reduction could lead to disin-
tegration of sand particles (Dabirian et al. 2016a,b). Sand deposition
In oil and gas field operations, deposition of solid particles and/or can occur both in single phase (oil or gas) and multiphase (oil-gas)
water into transport pipelines is very common and can cause significant pipelines (Zorgani et al., 2018).
flow assurance problems during the production phase (Giacchetta et al., The principal strategies implemented in the oil and gas industry to
2017). In particular, the deposition of any form of products can lead to resolve the problem of sand deposition in pipelines can be categorized
the pipeline section reduction with a significant loss of production into two main groups: sand deposit removal and sand deposition pre-
(Dall’Acqua et al., 2017). One typical kind of deposit which causes flow vention (Rahmati et al., 2013). The removal is carried out through
assurance problems is sand normally produced from near wellbore mechanical devices called pig, a solid object with the diameter smaller
formation when the forces binding the sand particles are lessened or than the pipeline inner diameter, which passes through the pipeline to
absent. Various causes of sand production can be identified: sand pro- scrape off the sand deposit driven by the pressure difference between its
duction in a well depends on the degree of consolidation of the sand head and tail. However, mechanical pigging can not be utilized without
particles, high reservoir fluid viscosity flowing into a well may lead to a proper sand deposition prediction: if the deposit thickness is too thick,
sand production due to the frictional drag force being applied to the the pig can go stuck inside the pipeline making the situation worse,

∗
Corresponding author.
E-mail addresses: [Link]@[Link] (M. Leporini), [Link]@[Link] (A. Terenzi), [Link]@[Link] (B. Marchetti),
[Link]@[Link] (F. Corvaro), [Link]@[Link] (F. Polonara).

[Link]
Received 1 March 2018; Received in revised form 3 December 2018; Accepted 20 December 2018
Available online 26 December 2018
0920-4105/ © 2018 Elsevier B.V. All rights reserved.
M. Leporini et al. Journal of Petroleum Science and Engineering 175 (2019) 519–535

2. Mathematical modelling

The most common transient thermo-hydraulic simulation program

for multiphase flow in pipelines used in the Oil and Gas field, named
OLGA (by Schlumberger) (Giacchetta et al., 2014), implements a new
model of solid transport in multiphase flow. This model incorporates
the physics of suspended particle load, as well as the bed load for
particle balance, concentration, velocity, and mass suspension. It con-
siders formation of a bed as well as cross-sectional distribution of the
particles between the different fluid layers, in addition to the axial
particle transport (Schlumberger, 2016).
In this model, conservations equations are solved for the mass of
particles in two fields: the particles suspended in the fluid layers, and
the particles in the bed. The bed layer consists of two sub-layers: a static
Fig. 1. Sand flow regime in horizontal pipelines. part (called stationary bed) and a moving part (called moving bed or
bed load).
while if the deposit is too thin, the mechanical pigging can not scrape For the suspension, conservation equations are solved for the total
off all the precipitated wax. Another removal technique to avoid the mass of suspended particles. This makes it possible to account for
intrusion of sand into the pipeline is the use of down-hole sand exclu- mixing of particles between the oil, water and gas layers. The en-
sion systems such as gravel packs and screens; however, these techni- trainment and deposition rates of the particles determine the size of the
ques may cause a significant loss in productivity. For these reasons, the bed, while the porosity of the bed (available as user input) determines
development of sand deposition prevention strategies has become a the amount of fluid trapped in the bed. It is suggested by code devel-
conventional method for the exploitation of the oil & gas resources. opers to use a bed porosity value equal to 0.35 (Schlumberger, 2016).
A good sand deposition prevention technique is, of course, based on The model assumes that the fluid does not flow through the stationary
the proper prediction of the behavior of the sand particles in single and bed. Particles are assumed to be fully dispersed in the three phases (gas
multiphase conditions in order to properly predict the sand transpor- layer, oil layer and water) at pipeline inlet. The particles velocity is
tation flow regime and the critical sand deposition velocity. In fact, a defined using a drift velocity or slippage defined as the velocity of the
production system affected by sand and correctly designed operates particles relative to the velocity of the carrier phase. Only gravity in-
above the critical sand deposition velocity, and solid particles are dis- duced slippage is considered, and therefore particles flowing horizon-
persed in fluid phases (considering the critical sand deposition velocity tally will move with the velocity of the carrier phase if no deposition is
defined as the mean stream velocity required to prevent the accumu- considered. The vertical slippage between particles and fluid, Vsl, is
lation of a layer of sliding particles at the bottom of horizontal pipe). calculated depending on the Reynolds number; for laminar flow
Basically, four flow regimes can be identified for the solid-liquid slurry (Re < 3), Eq. (1) is solved (Schlumberger, 2016).:
flows in horizontal pipe (Fig. 1): static bed (stationary bed), moving bed 2
(moving dunes), saltation (heterogeneous flow), suspension (homo- Vsl = 0.32673 (ρs − ρf ) ⎛ ds μ ⎞
⎜ ⎟

f (1)
geneous flow). ⎝ ⎠
According to Dabirian et al. (2015), the same classification can be where ds is the particle diameter, ρs is the density of the solid particles,
adopted also for multiphase pipelines under stratified flows, with two ρf is the density of the fluid, μf is the viscosity of the fluid, Vsl is the slip
additional sub-patterns of dilute solids at the walls and concentrated velocity (Bourgoyne et al., 1991). The Reynolds number for particles is
solids at the wall. However, the four classic sand flow regimes remain based on Vsl and particle diameter: Re= (ρf Vsl ds/μf). For transitional
the predominant. The transition between the flow regimes is identified flows (3 < Re < 300), the empirical correlation for Vsl is defined by
by a critical velocity. Different definitions of the critical velocity exist. Eq. (2), while for fully developed turbulent flows (Re > 300), the
Thomas (1964) introduced the minimum transport condition (MTC) as empirical correlation for Vsl is represented by Eq. (3):
“the mean stream velocity required to prevent the accumulation of a
Vsl = 0.7086 ds (ρs − ρf )0.667 (ρf μf )−0.333 (2)
layer of sliding particles at the bottom of horizontal pipe”. Wood (1979)
and other authors defined the deposit velocity as the limit velocity for
the stationary bed. Various authors like Stevenson and Thorpe (2003), ⎡ (ρs − ρf ) ⎤ 0.5
Vsl = 2.9452 ⎢ds ⎥
such as Salama (2000), Dabirian et al. (2016a), defined the critical ρf (3)
⎣ ⎦
deposition velocity as the transition velocity between the deposit and
The particles velocity in a layer is linearly scaled down to zero when
non-deposit flow regimes. The various definitions of the critical velocity
the concentration in the layer is between 0.7 and 0.8. The viscosity of
have been introduced by many authors on their classification of sand
the fluid is corrected due to the suspended particles. For this, the
flow patterns. The standard classification used in the oil and gas in-
Krieger-Dougherty correction (Krieger and Dougherty, 1959) factor is
dustry is the one proposed by Thomas (1964). In general, the transition
used:
velocity depends on main factors, including, sand characteristics (size
and shape), pipeline properties (diameter, inclination and roughness), μf θ ⎞ −2.0025
= ⎛1 −
flow regimes, and fluid properties. μ0 f ⎝ 0.75 ⎠ (4)
The present study presents a numerical sensitivity on the main
factors influencing sand transport in solid-liquid flow, a reproduction of where θ is the volume fraction of suspended particles in the fluid.
the liquid-solid flow regime map including reference transition velo- As introduced, there are three equilibrium conditions of the particle
cities and a detailed multiphase solid-gas-liquid flow modelling. In transport mode: static bed (particles do not move at all), bed load (some
particular, this paper reports the results of a detailed testing on the of the bed is moving but no particles entrained in the fluid layer above
performance of a new sand transport model implemented in one-di- the bed), and suspended load (some or all of the particles are entrained
mensional dynamic multiphase code, performed by comparing numer- in the fluid layers). The code models the incipient condition of sus-
ical results with experimental data. pension as the point where the shear velocity at the bed surface be-
comes greater than the settling velocity of the particles in the fluid in
contact with the bed:

520
M. Leporini et al. Journal of Petroleum Science and Engineering 175 (2019) 519–535

1
2
⎡ 4gdp ⎛ ρp ⎞⎤
VM 2 = 17 ⎢ − 1⎟ ⎥
3CD ⎜ ρf (12)
⎣ ⎝ ⎠⎦
where the subscript p refers to the particle.
In addition, it is very close also to the definition of critical sus-
pension velocity UC proposed by Danielson (2007):
5
9
⎡ ⎛ ρp ⎞⎤
UC = Kv−1/9dp1/9 ⎢gD ⎜ − 1⎟ ⎥
ρf (13)
⎣ ⎝ ⎠⎦
where D is the pipe diameter and K is an experimentally determined
constant, equal to approximately 0.23 based on SINTEF data.

3. Simulated cases description

In this section, the studies relating to both sand-liquid and sand-gas-

liquid flow simulated in this study and, therefore, the main input data
Fig. 2. Forces acting on a particle (Schlumberger, 2016). for the OLGA model are described.

3.1. SINTEF data

τf
v∗ = > Vsl cos(α )
ρf (5) Danielson's (Danielson, 20017) experimental work aimed to study
∗ the critical condition of sand slurry flow under different fluid condi-
where τf is the shear stress and v is the shear velocity of the fluid at the
tions:
bed surface. On the other hand, the incipient condition of bed load is

• Gas phases: air, nitrogen, SF6

calculated doing a balance of the forces acting on a particle as shown in

• Liquid phases: Exxsol D80, water. Table 1 gives the properties of

Fig. 2.
In Fig. 2, FL is the lift force, FD is the drag force, FB is the buoyancy
Exxsol D80 and water at experimental conditions.
force, FG is the gravity force, α is the pipe inclination and Φ is the angle
of repose. The effects of normal (Ni) and friction (fi) forces are neglected
Table 2 presents the test matrix of the study.
at the initiation of the rolling of the particle. The angle of repose is
defined as the steepest angle of descent or dip relative to the horizontal
3.2. Cranfield data
plane to which a material can be piled without slumping (usually taken
equal to 30°).
Yan (2010) and Al-lababidi et al. (2012) presented experimental
For the particle to move upwards, the following condition needs to
works to investigate the sand transport characteristics and identify the
be satisfied at the contact point P1:
sand minimum transport condition (MTC) in sand–water and sand–
FL sin(Φ) + FD cos(Φ) > (FG − FB )sin(Φ+α ) (6) air–water flows in horizontal and +5 deg inclined pipelines at Cran-
field University.
For the particle to move downwards, the following condition needs The used sand volume fraction, Cv, ranged from 1.61 × 10−5 up to
to be satisfied at the contact point P2: 5.38 × 10−4. These are typical sand concentrations experienced in oil
pipelines (0.000005–0.00005 v/v, 5lb/1000bbl―50lb/1000bbl).
FL sin(Φ) + FD cos(Φ) < − (FG − FB )sin(Φ−α ) (7)
Sometimes, the maximum sand concentration can reach 200lb/1000bl
The forces are calculated as follows: (i.e. 0.0002 v/v) and even 500 lb/1000bbl (i.e. 0.0005 v/v) due to shut
down or maintenance.
1 Two pipe sizes have been tested: 2in and 4in. The 2 inch sand-air-
FG = ρ gπ ds 3
6 s (8)
water facility is designed and constructed using ABS plastic (Class E)
pipe of 50 mm inner diameter (2 inch). The pipe length is 17 m. The 4
1
FB = ρ gπ ds 3 inch multiphase test facility is designed to operate under different
6 f (9)
multiphase flows, with and without sand. For the experimental in-
vestigations conducted, air-water is used as the test fluid with and
du
FL = 1.615uds 2 μf ρf without sand. The test section is made of 4 inch (ID = 0.1 m) stainless
dy (10) steel (SS316 L) pipes totaled 40 m in length. Table 3 presents the test
matrix of the studies.
1 Yan (2010) and Al-lababidi et al. (2012) conducted experiments
FD = CD ρf u2ds 2
8 (11)
with different liquids: water, CMC (Carboxy Methyl Cellulose) solutions
where u is the fluid velocity in the velocity sub-layer above the bed and (7, 20 cP) and oil (105, 200, 300 cP). CMC is a water-soluble polymer
CD is the drag coefficient. Here, the Saffman's model (Johnson, 1998) is used in synthetic detergents, drilling fluids, textiles, cosmetics, etc.
used for the lift force.
It should be noticed that the physical meaning of v* is the same of Table 1
the VM2 defined by Govier & Aziz (Govier and Aziz, 1972) as the ve- Measured fluid properties for Exxsol D80 and water.
locity at and above which the mixture flows in the asymmetric sus- Density (kg/m3) Viscosity (cP) Surface Tension (mN/m)
pension pattern; the velocity below which solids form a moving bed on
Exxsol D80 Water Exxsol D80 Water
the bottom of the pipe. In particular, VM2 is calculated by the authors
through the following Eq. (12) based on the semi-theoretical analysis of 795.3 999.8 2.17 1.33 34.4
Newitt et al. (1955):

521
M. Leporini et al. Journal of Petroleum Science and Engineering 175 (2019) 519–535

Table 2 available in literature relating to the solid-gag-liquid flow regime. Task

SINTEF experimental test matrix. 1 has been carried out in order to verify the model with respect to solid-
Variable Range Units liquid mixtures pipeline flow data. In particular the Task 1 has been
split into two sub-activities: Task 1.1 and Task 1.2.
Pipe inner diameter 0.069 m
Loop length 215 m
4.1.1. Task 1.1. sensitivity on main factors influencing sand transport
Particle size 280 μm
Sand concentration, Cv 0.3 v/v
On the basis of the considerations made above, a sensitivity on the
Superficial gas velocity, uSG 0.1–8.0 m/s main factors influencing the sand transport has been carried out. To do
Superficial liquid velocity, uSW 0.01–2 m/s this, experimental data found in literature have been used. In parti-
Maximum pressure 8 bara cular, the following parameters have been investigated:

1 Sand concentration
Table 3
2 Pipeline properties: diameter, inclination
Cranfield experimental test matrix.
3 Fluid flow rate
Variable Range Units 4 Sand properties (size)
5 Transporting medium properties: viscosity and density
Pipe diameter 2, 4 in
Particle size 200 μm
Particle density 2650 The studies of Danielson (2007) and Yan (2010) have been con-
Sand concentration, Cv 0.000005–0.00005 v/v sidered complete and suitable for the scope of Task 1.1 and they have
Superficial gas velocity, uSG 0.02–10 m/s been reproduced by OLGA code.
Superficial liquid velocity, uSW 0.05–1 m/s

4.1.2. Task 1.2. reproducing liquid-solid flow regime map including

3.3. Tulsa data reference transition velocities
During the Task 1.2, the liquid-solid flow regime map including
Dabirian et al. (2016a,b; 2017) presented a comprehensive study of reference transition velocities has been reproduced on the basis of the
sand flow regimes in air e PAC (Polyanionic cellulose) water stratified simulations carried out in Task 1.1. In particular, the generalized phase
flow in horizontal pipelines for various sand concentrations up to diagram for suspension transport in liquid-solid flow shown in Fig. 3
10000 ppm. Six sand flow regimes have been observed, namely, fully and reported in the fundamental book of Govier and Aziz (1972) has
dispersed solid flow, dilute solids at wall, concentrated solids at wall, been reproduced.
moving dunes, stationary dunes and stationary bed. Critical sand de-
position velocities have been determined based on the transition be- 4.2. Task 2. multiphase solid-gas-liquid flow regime
tween moving (concentrated solids at wall/moving dunes, as appro-
priate) and stationary (stationary dunes/bed, as appropriate) sand Task 2 has been devoted to the simulations of the multiphase solid-
particles. The viscosity of water has been increased to 5 cP by using PAC gas-liquid flow regime. In order to do this, the main experiment relating
to investigate the effect of viscous sublayer. The experiments have been to the sand transport and deposition in multiphase lines have been si-
carried out in a multiphase flow loop with 0.097-m PVC pipe, built at mulated. In particular, the following works have been considered sui-
the University of Tulsa, which utilizes glass beads representing the solid table for Task 2:
particles. The experimental conditions have been selected to ensure that
the particles are transported in the 0.097-m pipe in a stratified flow • Yan (2010) and Al-lababidi et al. (2012). Yan's work is the PhD
regime. To ensure stratified flow, the experiments are conducted at
superficial liquid velocities of 0.05, 0.1 and 0.12 m/s and superficial gas
velocities ranging from 6 m/s to 15.5 m/s. Table 4 presents the test
matrix of the studies.

4. Materials and methods

In this section, the main important tasks carried out during the
present activity are summarized.

4.1. Task 1. solid-liquid flow regime

Task 1 covers the study of the solid-liquid flow regime. In fact, it

was decided to investigate also this important aspect of the sand
transport and deposition in addition to the simulation of the data

Table 4
Experimental test matrix.
Variable Range Units

Pipe inner diameter 0.097 m

Particle specific gravity 2.475 –
Particle size 45-90, 125–250, 425-600 μm
Sand concentration, Cv 250–10000 ppm
Superficial gas velocity, uSG 5.5–15.5 m/s
Superficial liquid velocity, uSW 0.05, 0.1, 0.12 m/s
Liquid viscosity 5 cP
Fig. 3. Usual solid/liquid two phase flow patterns (Govier and Aziz, 1972).

522
M. Leporini et al. Journal of Petroleum Science and Engineering 175 (2019) 519–535

From the above results, it is possible to note that there is very little
influence of pipe angle on sand bed hold-up in the sand-liquid flow, as it
was found also by other authors (Yan, 2010).
It is important to point out that the default values of the OLGA sand
transport model have been used for the angle of repose (30°) and fluid
diffusivity factors and a bed porosity value equal to 0.35 has been as-
sumed, as suggested by OLGA developers. Due to the fact that the bed
porosity affects the volume of the bed, the sand holdup has been
evaluated in two ways: it has been plotted the relative OLGA output
variable (volume fraction of bed layer calculated by OLGA) and has
been calculated through the output variable height of the bed (h in
Fig. 5) by applying Eq. (14):
AS
HS =
APIPE (14)

Fig. 4. Danielson (2007) sand-water experiments. OLGA results vs experi- where HS is the sand holdup, AS is the area occupied by the sand and
mental data. sand holdup vs superficial water velocity. APIPE is the pipe section area. AS has been evaluated by Equations (13)
and (14) depending on the cases shown in Fig. 5. The same results have
thesis on which Al-lababidi's work is based. Al-lababidi's study has been obtained.
been indicated as an open data set used by OLGA's developers to As = r2 (π-2δ) + r sinδ (h-r) (15)
validate the sand transport model implemented in the code
• As = δ r - r sinδ (r-h)
2
University of Tulsa (Dabirian et al. (2016a,b; 2017) (16)

The SINTEF solid-liquid experiments also include pressure gradient

5. Results and discussion information. A plot of the measured vs. predicted pressure gradient by
OLGA is given in Fig. 6; a good fit to experimental data can be observed.
In this section, the results obtained for the Tasks indicated in pre- As outlined also by Danielson (2007), Fig. 6 shows an interesting
vious paragraph are presented. feature, which is that the slope pressure gradient with respect to flow
rate is negative from USL = 0, where USL is the superficial water ve-
5.1. Task 1. solid-liquid flow regime locity, to roughly USL = 0.47, i.e. the transition velocity (equivalent
pressure drop concept (Yan, 2010). When USL increases above 0.47,
5.1.1. Task 1.1. sensitivity on main factors influencing sand transport then the slope pressure gradient becomes positive. A negative slope of
[Link]. SINTEF data. Danielson (2007) reported sample data obtained pressure gradient with respect to flow rate is known to cause unstable
at SINTEF laboratory for sand hold-up as a function of superficial water behavior in multiphase flow, resulting in what is called terrain slugging
velocity, at inclination angles of −1.35°, 1.00°, and 4.00° from – the periodic accrual of liquid which is then pushed out of the system
horizontal. Measurements of both the sand bed height and the as a large slug. This has obvious analogy to the formation of moving
pressure drop have been reported by Danielson. The same dunes in sand-liquid flow. At very low rates of USL (about USL = 0.2 m/s
experimental conditions have been simulated by OLGA sand transport in the SINTEF data), the sand bed stabilizes, perhaps due to the tran-
model. sition from turbulent to laminar flow above the bed.
Fig. 4 shows the comparisons between experimental data and pre- It can be concluded that the experimental data presented in the
dictions from OLGA. The numerical results obtained by OLGA of the Danielson work and obtained at SINTEF multiphase laboratory, relating
sand-hold up for different liquid velocities are in excellent agreement to sand-liquid transport, are in perfect agreement with the numerical
with the experimental data. From Fig. 4, it can be seen that the sand bed results obtained by OLGA sand transport model.
begins to form at the critical carrying velocity of the liquid, which
appears, from the figure, to be at approximately 0.47 m/s. It is apparent
[Link]. Cranfield data. In Yan (2010) and Al-lababidi et al.’s (2012)
from Fig. 4 that once the sand bed starts to form, the increase in sand
works, experimental complex data related to both sand-liquid and sand-
hold-up as a function of decreasing water velocity is quite linear.
gas-liquid flow in 2inch and 4inch pipes are presented. In this section,
Table 5 shows the comparisons between experimental data and
the comparisons between OLGA numerical results and experimental
predictions from OLGA for the horizontal pipe case. Also in this case,
data are presented for the sand-liquid case.
the calculated values agree well with the experimental ones.
Before to show the results, it is important to point out that Yan
(2010) and Al-lababidi et al. (2012) adopted the definition of King et al.
Table 5
Danielson (2007) sand-water experiments. OLGA results vs experimental data.
(2000) for the sand minimum transport condition (MTC): “the mean
horizontal case. sand holdup for various superficial water velocity. stream velocity required to prevent the accumulation of a layer of
sliding particles on the bottom of horizontal pipe”. This definition is
usw h (OLGA) HOLs (OLGA) HOLs (EXP) Δ
different from the one provided by Salama (2000) and Govier and Aziz
m/s m – – % (1972) for the suspension or deposition critical velocity. Yan (2010)
0.10 0.047 0.75 0.77 2.60 and Al-lababidi et al. (2012) evaluated the MTC by visual method.
0.15 0.042 0.65 0.68 4.41 Table 6, Table 7, Table 8 and Table 9 show the comparisons be-
0.20 0.038 0.58 0.58 0.00
tween experimental data and predictions from OLGA in terms of sand
0.25 0.032 0.47 0.47 0.00
0.30 0.027 0.38 0.38 0.00 flow regimes for different sand concentrations (5, 15, 50, 100, 200 and
0.40 0.017 0.20 0.19 5.26 500 lb/bbl) for the 2inch case. Both the numerical and experimental
0.47 0.000 0.00 0.00 0.00 results are related to both horizontal and +5° inclined pipes as a con-
0.50 0.000 0.00 0.00 0.00 firm that in gas-liquid flow there is very little influence of the pipe
0.60 0.000 0.00 0.00 0.00
inclination. OLGA always predict a transition between suspension flow
0.70 0.000 0.00 0.00 0.00
0.80 0.000 0.00 0.00 0.00 regime to moving bed flow regime at a liquid velocity of 0.29–0.30 m/s
and this one can be considered the critical velocity, different from the

523
M. Leporini et al. Journal of Petroleum Science and Engineering 175 (2019) 519–535

Figure 5. Cases for the calculation of HOLs.

Table 7, Tables 8 and 9, it can be observed that for liquid velocities

equal to 0.29–0.30 m/s, i.e. the critical suspension velocity predicted by
OLGA, the experiments show the transition from “Scouring sand dunes
with less particles moving” to “Less particles moving”, a transition very
close to the one simulated by OLGA from suspension to moving load.
Table 10, Table 11 and Table 12 show the comparisons between
experimental data and predictions from OLGA in terms of sand flow
regimes for the 4inch case and for the same sand concentrations (5, 15,
50, 100, 200 and 500 lb/bbl) of the 2inch case. Also in this case, the
numerical and experimental results are the same for both the horizontal
and +5° inclined pipes as a confirm that in gas-liquid flow there is very
little influence of the pipe inclination. For the 4inch case, OLGA always
predict a transition between suspension flow regime to moving bed flow
regime at a liquid velocity of 0.32–0.33 m/s and this one can be con-
sidered the critical velocity, different from the MTC definition (see
Table 13).
Fig. 6. Danielson (2007) sand-water experiments. OLGA results vs experi- Once again, there is not correspondence between the experimental
mental data. horizontal case. pressure gradient vs superficial water velocity. data and the numerical results because they refer to different meanings
of the transition velocity. However, it can be observed that OLGA
Table 6 predicts that the critical velocity increases slightly for uphill flows,
Cranfield data. Experimental results 1. 2in sand-water horizontal and +5° in- according to what experimentally found by Shook and Roco (1991).
clined flow. OLGA results vs experimental data. Flow regime vs liquid super- In order to investigate the liquid viscosity effect on sand transport
ficial velocity. characteristics, Yan (2010) conducted experiments with different li-
quids: water, CMC (Carboxy Methyl Cellulose) solutions (7, 20 cP) and
usw Sand Concentration = 50 lb/1000 bbl
oil (105, 200, 300 cP).
OLGA Sand Regime Exp Sand Regime Table 14 shows the comparisons between experimental data and
predictions from OLGA in terms of experimental observed sand
m/s – –
minimum transport velocities (MTC) and numerical critical deposition
1 Suspension Suspension
0.90 Suspension Suspension velocity predicted by OLGA for the 4inch case and for different sand
0.80 Suspension Suspension concentrations (50 and 200 lb/bbl) and fluid viscosities. From Table 14,
0.70 Suspension Suspension it can be seen that there is not agreement between the experimental
0.60 Suspension Suspension MTC and the numerical critical velocity for the motivation explained
0.55 Suspension Moving streaks (MTC)
0.47 Suspension Dense sand streaks
above. Moreover, it can be noticed that critical velocity predicted by
0.45 Suspension Small dunes OLGA is the same for the cases with water and CMC solutions (all cases
0.40 Suspension Developed sand dunes with turbulent flow).
0.35 Suspension – When the flow became laminar (three cases with oil), the sand
0.30 Suspension –
minimum transport velocity and the critical velocities decrease as the
0.29 Moving bed –
0.25 Moving bed – fluid viscosity increase. This is due to the increasing shear force acting
0.20 Moving bed – on the sand particles from the liquid and the decreasing settling velocity
0.15 Moving bed – of sand particles due to the high viscous fluid. In particular, the main
0.10 Stationary bed – reason for decreasing the critical velocity with increasing viscosity is
0.05 Stationary bed –
the effect of the drag force. Under laminar flow, increasing the viscosity
is resulted to increasing the drag coefficient, correspondingly increasing
MTC definition. the drag force. Therefore, lower critical velocity is required to maintain
From Table 6, it seems that there is no agreement between experi- moving the particle along the pipe. This finding is also consistent to
mental data and OLGA numerical results. But this is not true. In fact, as similar work done by Gillies et al. (1997) which using water (1 cP) and
described in section 2, the moving bed regime predicted by OLGA oil (78 cP) and with the experimental results of Yan (2010).
means that some of the bed is moving but no particles are entrained in From Table 14, it is also possible to see that, at the same liquid
the fluid layer above the bed. The experimental data show that for viscosity, the sand minimum transport velocity have been observed
velocities lower than the observed MTC, dunes continue to develop and, always increased with the increase of sand concentration unlike the
therefore, particles continue to move above the bed. By analyzing also critical velocity predicted by OLGA which is the same with the increase
of sand concentration, as it can be found applying the models of

524