Wear 372-373 (2017) 42–53

Contents lists available at ScienceDirect

Wear
journal homepage: www.elsevier.com/locate/wear

Erosion prediction for slurry flow in choke geometry

Farzin Darihaki a, Ebrahim Hajidavalloo a,n, Amir Ghasemzadeh b, Gholam Abbas Safian b
a
Mechanical Engineering Department, Shahid Chamran University of Ahvaz, Ahvaz, Iran
b
NISOC , Ahvaz, Iran

art ic l e i nf o a b s t r a c t

Article history: Erosion damage is common in many industries and one of the flow components which may experience
Received 21 June 2016 severe erosion rates is choke geometry. Choke valves in wellhead or well completion may experience
Received in revised form such conditions. In the present study, Computational Fluid Dynamic (CFD) was employed to calculate
27 November 2016
erosion for a choke geometry constructed of aluminum. The Eulerian-Lagrangian approach was utilized
Accepted 5 December 2016
for continuous and discrete phase calculations in which different turbulence models were implemented
Available online 9 December 2016
to investigate the importance of erosion mechanisms in the choke geometry. Meanwhile, various erosion
Keywords: models were examined to achieve accurate erosion prediction. Afterward, two modifications in the
Multiphase Flow geometry were made to see how erosion will be changed in different sections. It was revealed that the
Liquid-Solid
modifications can affect the dominating erosion mechanisms and newly designed profiles can reduce the
Particle Erosion
erosion rate in different parts of the geometry significantly. Finally, erosion rate for carbon steel material
Choke
CFD which might be found in the oil and gas production systems was compared with the one of original
aluminum case and it was found that thickness loss rate for the carbon steel case is one-third of the
aluminum one in the most parts of the choke and much smaller on contraction plane.
& 2016 Elsevier B.V. All rights reserved.

1. Introduction different geometries including elbow and sudden expansion. Nu-

Erosion caused by solid particles carried with the fluid is
common in many industries. In some cases such as oil and gas
production, it is not possible to eliminate the solid phase. The
extent of erosion damage depends on the flow conditions and the
equipment geometry. Regarding the geometry, cases like elbow
and choke in which the direction of flow-streamlines is changed
considerably, there is the risk of high erosional damage. As a part
of the production system, chokes which might be found in well-
head downstream or well completion are considered as equip-
ment with high erosion rates.
High local erosion rate can cause severe damage to the system
due to equipment failure and even stop the whole operation which
costs lots of money and time. Hence, it is needed to know when
and where erosion is a serious problem and inspection or repair
should be executed.
CFD is a powerful and capable tool to estimate the erosion
extent for real-life, complex geometries and flow phenomena and
it has been used by many researchers for about two decades to
verify erosion models and predict erosion rates.
Edwards [1] employed the general procedure of erosion cal-
culation in a commercial CFD code and evaluated the erosion for
n
Corresponding author.
E-mail address: hajidae@scu.ac.ir (E. Hajidavalloo).
groho [2] studied erosion for gas-solid flow in sudden expansion
geometry both experimentally and numerically. He utilized the
CFD code developed by Edwards [1] and examined the effect of
different expansion ratios, where he found that the erosion rate
will be reduced by increasing the expansion ratio. Based on find-
ings of Founti et al. [3], Nugroho [2] implemented a two-dimen-
sional formulation of Saffman lift force [4] in the particle tracking
routine, then he concluded that this force has a significant effect
on the predicted erosion.
Erosional damage for gas-solid flow in an annular cavity was
examined experimentally and numerically by Wong et al. [5]. They
reported erosion data for each section of the cavity, including
cavity wall, backward-facing step and forward-facing step and
found that there is an over-prediction with calculated erosion rates
for smaller particles of 38 μm. Zhang et al. [6] studied erosion for
liquid-solid flow in an elbow and investigated the behavior of
particles in near wall region. They concluded that the effect of this
region becomes important when smaller particles are present.
Mansouri et al. [7] also studied the erosion damage in an elbow for
different particle sizes. They showed that it is important to take
care of particle behavior in the viscose boundary layer when small
particles with 25 μm in size approach the wall.
From the early days of the research on particle erosion, so many
experiments have been performed. McLaury [8] studied erosion
for choke geometry and did experiments to correlate erosion ex-
tent of a specimen resulting from an impinging gas-solid jet to the

http://dx.doi.org/10.1016/j.wear.2016.12.008
0043-1648/& 2016 Elsevier B.V. All rights reserved.
 F. Darihaki et al. / Wear 372-373 (2017) 42–53 43

impact characteristics including impact speed, impact angle and

particle and wall properties, which is called erosion equation.
Other researchers including Oka et al. [9,10], Zhang [11], Mansouri
et al. [12], Arabnejad et al. [13] also reported different erosion
equations using different experiments.
Mansouri et al. [14] investigated the erosional damage for gas-
solid and liquid-solid jets impacting a specimen with different
angles. They reported different erosion profile for each of these
flows. They found that at normal impact angle, gas-solid flow
Fig. 1. Effect of shear lift force on a particle near the wall.
produces U-shaped erosion pattern and maximum erosion occurs
where the centerline of the jet crosses the specimen, i.e. stagna-
Virtual mass force is calculated using such equation,
tion point. While for the slurry flow erosion profile is W-shaped
and erosion has two peaks that are located at some distance from ⎯⎯⎯→ ρf d
the stagnation point.
Fvm =0. 5
ρp dt
( ν ⃗− νp⃗ )
(5)
The aim of this research is to investigate erosion rate in choke
geometry caused by sand particles carried by liquid. Various Gravitational force ( FG ): Buoyancy can affect the particle mo-
models have been examined to achieve the best agreement with tion and is defined by the following equation,
the experimental data and provide more insights on this phe-
nomenon. The effect of lift force on predicted erosion rate has been ⎯→ (ρp − ρf )
FG = g⃗
studied using a more sophisticated formulation. Also, this paper ρp (6)
explores to what extent modification of the choke entrance can
reduce the erosional damage on different sections of the choke. In Unlike the pressure gradient force and virtual mass force which
the end, a steel material which is common in the industrial sys- are important when the particle density is much smaller than the
tems has been selected and the predicted erosion rate has been fluid density, for the gravitational force, it is the density difference
compared with the original case. that matters.
Shear lift force ( FL ): High velocity gradient which causes an
unbalanced pressure distribution results in a lift force acting on
2. Theory and mathematical models the surface of the particle in the direction which has larger relative
velocity. Shear lift force is also known as Saffman lift force or
Erosion calculation is executed in two steps. At the first, con- transverse lift force. Assuming the particle velocity is greater than
tinuous fluid flow is solved within the computational domain average fluid velocity around the particle, this force is illustrated in
using the Eulerian approach and then the motion of discrete solid
Fig. 1.
particles is calculated in the same domain.
Sommerfeld [15] has introduced the following equation for the
The particle loading is assumed to be low enough so that the
shear lift force,
motion of particles doesn’t affect the continuous phase. By im-
plementing this one-way coupling approach, there won’t be any ⎯→ 0.5 ⎛1⎞ ⎡
0.5

term in the fluid flow equations representing the solid phase. FL =1. 615 ( )
dp2 ρf μ ⎜ ⎟ ⎣ ( ν ⃗−νp⃗ )× ω⃗⎤⎦
⎝ ω⃗ ⎠ (7)
Particle motion, including path and velocity components, is
computed by integrating equations of particle motion. The first Where ω is the vorticity, Mei [16] has derived a correction factor
equation is the velocity definition and the second equation is the for the lift force based on the calculations of Dandy and Dwyer [17]
Newton's second law or the force balance on the particle, in which relative Reynolds number ranges from 0.1 to 100,
⎯→
dvp ⎯→ ⎯→ ⎯⎯⎯→ ⎯→ ⎯→ ⎧
=FD +Fp +Fvm +FG +FL
dt (1) (
⎪ 1 − 0. 3314 β 0.5 ) Re≤40
⎪
⎪ ⎛ ⎞
In which, the terms on the right-hand side represent forces f ( Re , Res )= ⎨ exp ⎜ − Re ⎟+0. 3314 β 0.5 ,
⎪ ⎝ 10 ⎠
acting on the particle per unit mass of it. ⎪
Drag force ( FD ): The drag force is the most dominant force on ⎪
⎩ 0. 0524 ( β Re)
0.5
, Re >40 (8)
the particle motion and is calculated by the following equation,
Where,
18 μ Re
FD= CD
24 ρp dp2
( ν⃗ −νp⃗ )
Res
(2) β=0. 5
Re (9)
Where CD is the drag coefficient and Re is the relative Reynolds
number,
ρf dp2 ω⃗
ρf dp ν ⃗ −νp⃗ Res =
μ (10)
Re=
μ (3)
Therefore, the corrected lift force could be calculated as follows,
Pressure gradient force ( Fp ): This force is the result of pressure
⎯→
⎯ ⎯→
distribution of the fluid around the particle and has the following F L* = f ( Re , Res ) FL (11)
form,
⎯→ ρf Shear lift force with this formulation which covers a wider
Fp = νp⃗ ∇ν ⃗ range of relative Reynolds number, has never been utilized before
ρp (4)
in the erosion calculation studies. A User-Defined Function (UDF)
Virtual mass force ( Fvm ): Acceleration of the fluid around the has been written and employed in this study accounting for the lift
particle act as a force equals to the inertia of this added mass. force and its correction.
44 F. Darihaki et al. / Wear 372-373 (2017) 42–53

2.1. Particle rebound model ( Etl ), defined by the wall thickness loss to the flow-time. Wall
Thickness Loss Rate can be obtained from the Erosion Ratio as,
Every time a particle hits the wall, it will lose some of its mo-
⎡ m⎤ ⎡ kg ⎤ ⎛ m*̇ ⎞ ⎡ ⎤
mentum. Restitution coefficient is a measure for momentum loss Etl ⎢ ⎥ = ER ⎢ ⎥ × ⎜ p
⎟ m
and defined as the ratio of velocity components after the impact to ⎣ s⎦ ⎣ kg ⎦ ⎜ ρ A ⎟ ⎢⎣ s ⎥⎦
⎝ w ⎠ f (16)
the ones before. Particle rebound models are defined as functions
of impact angle and usually developed using experiments on dif- Where mp*̇ is the mass flow rate corresponding to the impacting
ferent materials. Forder et al. [18] have introduced a rebound particle, ρw is the density of wall material and Af is the area of cell
model based on tests using AISI 4130 carbon steel, and particles of face at the boundary where the impact takes place.
diameter 150–300 μm . Restitution coefficients in their model are Five erosion models which have been examined in the current
as follows. study have been introduced in the following part.
McLaury (1996): Using direct impingement test with different
eperpendicular =0. 988 − 0. 78 θ +0. 19 θ 2−0. 024 θ 3+0. 027 θ 4 (12)
impact angles for aluminum, McLaury [8] introduced an equation
which depends on the velocity and angle of impact, and constants
eparallel=1 − 0. 78 θ +0. 84 θ 2−0. 21 θ 3+0. 028 θ 4−0. 021 θ 5 (13) related to the particle shape, and wall material properties.
By utilizing perpendicular and parallel coefficients of restitu- ER=CFs V pn f ( θ ) (17)
tion, velocity components right after the impingement can be
calculated; Where C is the wall material constant, Fs is the particle shape
νn′ = eperpendicular × νn constant taking the values of 0.2, 0.53 and 1 for fully rounded,
(14)
semi-rounded and sharp particles, respectively. Vp is the impact
νt′ = eparallel × νt speed and f ( θ ) is the impact angle function. Impact velocity ex-
(15)
ponent is an empirical constant. The values of the constants in
Careful modeling of particle rebound is not required for the McLaury model have been given in Table 1.
cases in which liquid or a fluid with considerable viscosity is
⎧ aθ 2+bθ; θ≤θ 0
present. Because the particle relaxation time would be short en- f ( θ )=⎨
⎩ x cos2 (θ ) sin (wθ )+y sin2 θ +z; θ > θ 0
⎪

ough and the particle reaches the fluid velocity quickly, the dif- (18)
ferences in the rebound velocity do not affect its path.
The parameter θ0 in impact angle function is the reference
impact angle.
2.2. Erosion equations
Oka et al. (2005): After performing experiments under wide
test conditions, Oka et al. [9,10] reported an erosion equation
During tracking of each particle, impingement data is recorded
which takes account of more parameters.
and used to calculate the erosion at the impact location. This part
of erosion calculation procedure is carried out using the erosion ER= 10−9ρw f ( θ ) E90 (19)
equations.
Erosion equations are mostly introduced in the form of Erosion
Ratio (ER), which is the mass of removed material from the surface ⎛ Vp ⎞k 2 ⎛ D ⎞k3
k
to the mass of particles impacting the wall surface. Erosion Ratio E90=K ( Hv) 1 ⎜ ⎟ ⎜ ⎟
⎝ v′ ⎠ ⎝ D′ ⎠ (20)
can be converted to other forms of erosion with different units. In
the present study, erosion is presented by Wall Thickness Loss Rate

Table 1
Constant parameters of erosion models.

McLaury (1996) C θ0 n a b w x y z
2. 388 × 10−7 10° 1. 73 −34. 79 12. 3 5. 205 0. 174 −0. 745 1
Oka et al. (2005) Particle material K k1 k2 k3 v′⎡⎣ m/s⎤⎦ D′ [ μm] s1 q1 s2 q2
SiO2 65 −0.12 0.038
2.3 ( Hv ) 0.19 104 326 0. 71 0. 14 2. 4 −0. 94

SiC 45 −0.05 3.0 ( Hv )

0.085 0.19 99 326 0. 71 0. 14 2. 8 −1. 00

Glass Bead 27 −0.16 2.1 0.19 100 200 2. 8 0. 41 2. 6 −1. 46

Zhang (2006) C n a1 a2 a3 a4 a5
2. 17×10−7 2. 41 5. 3983 −10. 1068 10. 9327 −6. 3283 1. 4234

Mansouri et al. (2016) C A n n1 n2 n3

4.49 × 10−7 0. 6947 2. 41 0. 2 0. 85 0. 65

Arabnejad (2015) Wall material C1 C2 K Vtrsh ⎡⎣ m/s ⎤⎦

Carbon Steel 1018 5.90 × 10−8 4.25 × 10−8 0.5 5.5
Carbon Steel 4130 4. 94×10−8 3.02 × 10−8 0.4 3.0
Stainless Steel 316 4. 58×10−8 5.56 × 10−8 0.4 5.8
Stainless Steel 2205 3.92 × 10−8 2.30 × 10−8 0.4 2.3
13 Chrome Steel 4. 11×10−8 3.09 × 10−8 0.5 5.1
Inconel 625 4. 58×10−8 4.22 × 10−8 0.4 5.5
Aluminum Alloy 6061 3. 96×10−8 3.38 × 10−8 0.4 7.3
 F. Darihaki et al. / Wear 372-373 (2017) 42–53 45

n2 interaction, including Forder et al. [18] rebound model and the

(
f ( θ )= sin ( θ ) )n ( 1+Hv ( 1−sin ( θ ) ) )
1
(21)
introduced erosion models.

qi
ni =si ( Hv)
Where E90 is the Volumetric Erosion Ratio at the normal angle of
impact, taking units of mm3/kg and Hv denotes the Vickers
hardness of wall material in units of GPa.
In the Oka et al. model, n1, n2 and k2 depend on the particle and
wall material properties, while K , k1, v′ (reference impact speed)
and D′ (reference particle diameter) only refer to the particle
properties. The values of the constants used in Oka et al. model for
different particle materials have been listed in Table 1.
Zhang (2006): Zhang [11] did some experiments with direct
impingement of gas-solid on Inconel 718 alloy and calibrated
McLaury erosion equation. He introduced the following erosion
equation,
−0.59
ER=C × ( HB) Fs f ( θ ) V pn
f ( θ )=a1θ +a2 θ 2+a3 θ 3+a 4 θ 4+a5 θ 5
Where the particle shape constant ( Fs ) is the same as in McLaury
model. Other constants have been provided in Table 1. HB is the
Brinell Hardness number of wall material and can be derived from
Vickers hardness by the following equation,
(22)
3. Case study
McLaury [8] measured the erosion rate in a choke geometry.
The choke was made of several wafers as shown in Fig. 2. After a
couple of hours of flow-time, when the erosion occurs, it is pos-
sible to determine the amount of mass loss by consecutive mea-
surements of the wafers’ mass. Then, assuming uniform erosion on
the wall of each wafer, the thickness loss can be obtained from the
mass loss. The resulted erosion is one-dimensional along the axis
of the choke. The measurements were performed every 5 hours
during 25 hours of flow-time. Structural and flow parameters of
the executed test are given in Table 2.
In the present study, erosion data after 10 hours under test
condition is used for model validation, so the effect of corners
(23) rounding on the flow is insignificant. Fig. 3 shows the wall thick-
ness loss rate reported by McLaury [8]. As it was expected, erosion
in the contraction plane or forward-facing step of the choke has
(24) the largest value. Constriction section also has a considerable
amount of mass loss. This is the section where the free-stream
fluid and hence particles have highest mean velocity. But the ex-
pansion downstream has experienced lower erosion damage.

Hv ⎡⎣ GPa⎤⎦+0. 1023 4. Computational grid and numerical schemes

HB=
0. 0108 (25)

Mansouri et al. (2016): Based on direct impingement test on a Knowing that geometry is symmetrical with respect to the
stainless steel (SS316) with particles carried by liquid, Mansouri choke axis and also the forces acting on the flow has symmetrical
et al. [12] reported the following erosion equation, distribution, one can conclude that calculations could be done
using symmetrical condition around the choke axis. In this study, a
−0.59
ER=C ( HB) Fs f ( θ ) V pn (26) quarter-pipe was considered for simulation, so the computational
cost was reduced considerably. This means that computational
domain has two symmetrical boundaries. Multi-block grid gen-
f ( θ )= (
A sin ( θ ) )n1
(
(1 + Hv n3 1−sin ( θ ) )n
2
(27)
eration method is used in this study. Computational domain and
Constants of Mansouri et al. erosion model could be found in the generated grid on the inlet plane are shown in Fig. 4.a and b. In
Table 1. the computational domain, the length of outlet pipe is chosen to
Arabnejad (2015): By combining empirical and mechanistic be 164.465 mm, therefore the flow would be fully-developed at
methods, Arabnejad et al. [13] developed a semi-mechanistic the outlet.
erosion model which accounts for cutting and deformation me- Flow calculation, particle tracking and erosion prediction have
chanisms, and also is in agreement with experimental data. been accomplished using ANSYS FLUENT CFD software in-
corporating a unified UDF code. Table 3 lists numerical schemes
ER=ERC +ERD (28)
and models which have been implemented in the present study.
Uniform injection of particles at the inlet plane was fulfilled using
UDF code.
⎧ 2.41 sin θ ⎡ 2 K cos θ −sin θ ⎤
⎪ C F Vp ( )⎣ ( ) ( )⎦ Regarding boundary conditions, inlet boundary type was set to
⎪ 1 s 2
; θ≤tan−1( K )
ERC = ⎨ 2 K velocity inlet in which according to Table 2, average inflow velocity
⎪ Vp2.41 cos2 ( θ ) of 2.5 m/s was used as input. Outflow condition was applied to the
⎪ C1Fs ; θ > tan−1( K ) outlet. This type of boundary assumes that the flow is fully de-
⎩ 2
veloped at the outlet. At the end of the flow calculations, all re-
(30)
siduals were below 1 × 10−5, except for the continuity which was
equal to 2.27 × 10−3. Also, solution convergence was confirmed by
monitoring velocity at the centerline of the outlet pipe section.
(
ERD=C2 Fs Vp sin ( θ ) − Vtrsh )2 (31)

Where ERC and ERD are erosion ratio due to cutting and de-
formation mechanisms. The particle shape constant, Fs , acquires
values of 0.25, 0.5 and 1 for fully rounded, semi-rounded and
sharp particles, respectively. Empirical constants of Arabnejad
erosion model for different wall materials have been given in
Table 1.
A User-Defined Function code was written and implemented in
the particle tracking calculations accounting for the particle-wall Fig. 2. Schematic of the choke (Choke Design).
46 F. Darihaki et al. / Wear 372-373 (2017) 42–53
Table 2
Structural and flow parameters for choke erosion test.
Inlet and outlet pipe diameter 15.875mm
Choke diameter size 6.477mm
Inlet pipe length 53.975mm
Choke length 76.2mm
Outlet pipe length 66.675mm
Fluid density 998.2kg /m3 (Water )
Fig. 3. Wall thickness loss rate in choke geometry after 10 hours [8].
5. Results and discussion
5.1. Grid study
To check grid independency of the solution, various grids with
different number of cells in the cross-section plane and axial di-
rection were generated and examined. In this part of the study, k-ε
turbulence model and Mansouri et al. erosion model were em-
ployed for the flow calculation and erosion prediction, respec-
tively. Maximum erosion, as the most important parameter, was
considered as the criteria for grid sensitivity analysis. Note that all
erosion rates presented in graphs are the result of axial averaging
of raw erosion data every 0.5 mm.
Table 4 lists the generated grids with different number of
computational nodes in the axial direction but the same number
of cells in cross-section. Also, normalized values of the cell num-
bers have been given. These normalized values are the ratio of
number of cells in the specified section of each grid to the corre-
sponding number of cells in grid A1.
Fig. 5 shows maximum values of the predicted erosion for the
both constriction and outlet pipe sections. It can be concluded
from this figure that using grid A3, the solution has an acceptable
combination of the accuracy and computational cost. In addition,
Fig. 4. a. Computational domain, b. Generated grid on the inlet plane.
Fluid viscosity 1. 003×10−3 kg /(m. s )
Inlet velocity m
2.5
s
Particle diameter 155μm
Particle density 2650kg /m3
Solid phase concentration 2% by weight
Wall density 2700kg /m3 (6061-T6 Aluminum)
this figure suggests that the constriction section is more affected
by variation of the axial grid density. This is related to the fast
developing flow in this section.
Table 5 lists the generated grids with different number of cells
in the cross-section, but they all share the same number of axial
nodes as the grid A3. It should be noted that the grid C3 is identical
to the grid A3.
Fig. 6 shows all of the examined grids with different cross-
sectional cells. Here the predicted values of maximum erosion by
different grids are very similar in both sections. Hence, Figs. 5 and
6 can bring us to this point that the grid A3 (C3) is suitable for the
numerical calculation and has been taken for the rest of this study.
Also, locations of the maximum erosion for all of the grids were
found to be roughly at 129 mm and 194 mm from the inlet for the
choke section and the outlet pipe, respectively.
The size of the very first cell near the wall boundaries and the
maximum calculated yþ values using k-ε turbulence model are
given in Table 6. The largest size of near wall cell in the compu-
tational domain is less than particle radius. However as Zhang
et al. [6] investigated the near wall behavior of the particles, this
region becomes important when the particle size is as small as 25
μm. This is because of non-physical impacts which occur when
such small particles are present in the near wall region [6]. Man-
souri et al. [7] also showed that near wall treatment becomes
important when it comes to the small particles with 25 μm in size.
All of the predicted values for erosion in this section have been
obtained by tracking a number of 80,000 particles. To see if this
amount of particle is sufficient to have a proper representation of
actual particles and also to get enough statistical impact data,
various numbers of particles were examined for erosion calcula-
tion. Fig. 7 indicates that 80,000 particles are well sufficient for
erosion calculation.
5.2. Erosion prediction
At first, the effect of turbulence model on the predicted erosion
profile was investigated. Four different turbulence models were
employed to solve flow equations. These models include both
standard and renormalized group (RNG) variation of the k − ϵ
model, SST k − ω model and seven-equation RSM model. Table 7
lists the calculated location of reattachment point where the wall
shear stress is minimum, in the expansion downstream region
 F. Darihaki et al. / Wear 372-373 (2017) 42–53 47

Table 3
Employed numerical models for liquid-solid flow calculations.

Temporal state Steady-state Particle injection method Uniform plane injection

Pressure velocity Coupling scheme
Momentum spatial Discretization
scheme
Turbulence model
SIMPLE (Pressure-based solver) Solid-liquid coupling method One-way coupling
QUICK Forces acting on discrete phase Spherical drag, virtual mass, pres-
sure gradient
Standard and RNG k − ϵ , k − ω and RSM Turbulence transfer Stochastic tracking
using scalable wall function scheme (Turbulent dispersion)

Table 4
Details of grids with different number of axial nodes.

Grid Name Grid A1 Grid A2 Grid A3 Grid A4

Outlet pipe wall Number of cells 3,340 3,820 4,780 5,260

Normalized value 1 1.14 1.43 1.57

Choke wall Number of cells 2,500 2,860 3,580 3,940

Normalized value 1 1.14 1.43 1.58
Total hexahedral elements 511,940 585,620 732,980 806,660

Fig. 6. Variation maximum erosion with the normalized cross-section cells.

Table 6
First cell height and the maximum of Y þ values.

Section of geometry Inlet pipe Constriction region Outlet pipe

First cell height [mm] 0.06 0.043 0.05

Y þ value 15 22.5 11.5

which are located at the both sides of the reattachment point.

Fig. 5. Variation of normalized maximum erosion with the normalized axial cells. The trend of the predicted thickness loss rate at the expansion
downstream could be explained by findings of Mansouri et al. [14].
Table 5 In their experiments on a slurry jet hitting a specimen at an angle
Details of grids with different number of element in the cross-section. lower than 90°, the erosion profile has three extremums; zero
erosion at the stagnation point and two peaks on both sides of that
Grid Name Grid C1 Grid C2 Grid C3 Grid C4
point. Due to the high drag force of stagnant liquid on the parti-
(A3)
cles, the surface at stagnation or reattachment point would ex-
Inlet/outlet sec- Number of cells 805 1,040 1,640 1,991 perience lowest erosional damage, as the impacting particles have
tion plane Normalized 1 1.29 2.04 2.47 smaller momentum. This explanation is supported by Fig. 9.a and
value
b showing the location of reattachment point has experienced
considerably low impact speed and thickness loss rate. While the
Choke plane Number of cells 239 336 540 649 deviation of the flow-streamlines before the stagnation point and
Normalized 1 1.30 2.08 2.51
high velocity fluid close to the wall region after the fluid has left
value
Total hexahedral elements 358,701 463,664 732,980 888,679
the stagnation point, cause two peaks on erosion profile at both
sides of reattachment point. It seems that for the slurry flow in
sudden expansion geometry the difference between two peaks on
using the mentioned turbulence models. It shows that the mea- both sides of the reattachment point is smaller than the difference
sured erosion confined up to the recirculation region and doesn’t of the corresponding points for slurry jet impacting a specimen.
include the region of new boundary layer formation. This could be explained by the cyclic impingements of particles in
Fig. 8 shows the predicted erosion profile using these models. the recirculation region in the downstream of sudden expansion
Almost all turbulence models suggest two peaks in outlet pipe section, which increases the erosional damage of the lower peak.
48 F. Darihaki et al. / Wear 372-373 (2017) 42–53

In the sudden geometry changes like in choke, random im-

pingement mechanism is very important due to the presence of a
high turbulent kinetic energy region near contraction plane
(Fig. 10). According to Fig. 8, this suggests that except standard
k − ϵ model, all of the models are not properly capable of ac-
counting for the random impingements and as a result, these
models fail to predict any impingement in constriction section. So
the standard k − ϵ model is employed for the rest of this study.
Also, it was found out without implementing Stochastic Tracking
model, literally no impingement occurs neither in the constriction
section nor the outlet pipe.
Amongst turbulence models, Reynolds Stress Model (RSM)
shows a very high erosion rate in the downstream of the re-
attachment point. Although this turbulence model is able to ac-
count for the anisotropic turbulent disturbances, the shape of the
recirculation region suggests this model might solve the flow
unrealistically as the front of the recirculation region is not at-
tached to the wall (Fig. 11). By closely observing the velocity vec-
tors near the wall it was found that there is a considerable increase
over the k-ε model, which explains the high level of predicted
erosion. Hence, the predicted erosion rate using RSM is believed to
be incorrect.
Another point about Fig. 8 is that the predicted erosion on the
contraction plane of the choke is much greater than the experi-
Fig. 7. Maximum erosion in choke and outlet section with different number of mental data. The reason is the way erosion has been measured;
particles. the eroded mass from the wafer is averaged over its entire surface
while the erosion on the contraction face is much greater than
Table 7 other faces of the wafer. In addition, calculated erosion is averaged
Calculated reattachment point location in expansion downstream. over very small distances (0.5 mm) in the axial direction. There-
fore, the comparison of measured and predicted erosion in con-
Turbulence model standard k − ϵ RNG k − ϵ SST k − ω RSM
traction plane won’t be practicable.
Reattachment location 172.6 179.8 186.4 162.7 After choosing the proper turbulence model, various erosion
[mm] models were examined. Fig. 12 shows the Oka et al., Zhang and
Arabnejad models underestimate thickness loss rate. McLaury
erosion model has a good prediction in the choke section but
overestimates the erosion in the expansion downstream; this is
the recirculation region where multiple impacts with low im-
pingement velocity happen; or there is low impingement angle i.e.
downstream of boundary layer reattachment. Zhang [11] also has
mentioned the over-predictions with McLaury model. Fig. 12
suggests only Mansouri et al. model is capable of correct predic-
tion of erosion trend and also its rate. Hence, Mansouri et al.
erosion model is used for the later calculations.
It should be mentioned that the value of 0.53 is chosen for the
particle shape constant, assuming particles are semi-rounded.
Fig. 13 shows the effect of shear lift force on the predicted
erosion. Lift force results in higher erosion in the choke section;
however, there is little difference in the expansion downstream.
By looking at Eq. (7) one can conclude that the presence of a re-
gion with high relative velocity or high relative Reynolds number
would result in high lift force on the particle toward the wall.
Fig. 14 demonstrates such region in the choke entrance.
However, including the correction factor shows an improve-
ment over the lift force without correction (Fig. 13), it does not
Fig. 8. Predicted erosion using different turbulence models (Mansouri et al. erosion result in better prediction of either erosion trend or its rate. This
model).
might be due to the higher relative Reynolds number of the flow
than the allowable range in Eq. (8) (Fig. 14).
Generally, two different mechanisms have been suggested for From Figs. 8 and 11, it can be found out that the combination of
impingement of the particles on the wall and erosion process [8]; the standard k − ϵ model and Mansouri et al. erosion model would
direct impingement in which carrier fluid has mean velocity afford the most acceptable prediction of erosion rate. Therefore,
pointing toward the wall, and random impingement in which these models have been employed for the rest of the calculations.
there is no mean velocity components in direction of the wall and
turbulent fluctuating velocity components cause the impact. Sto- 5.3. Effect of geometry modification on thickness loss rate
chastic Tracking model is employed to calculate fluctuating velo-
cities in discrete phase calculations. This process is closely related In this section, the effect of choke configuration on the erosion
to turbulence parameters of the continuous phase. has been investigated. Two modifications including gradual
 F. Darihaki et al. / Wear 372-373 (2017) 42–53
Fig. 9. Location of reattachment point within: a. Impact speed graph, b. Thickness loss rate graph.
Fig. 10. Turbulent kinetic energy contour on the choke symmetry plane (standard
k − ϵ model).
Fig. 11. Velocity contours and vectors in expansion downstream. above: k-ε model,
below: RSM.
contraction and initial step entrance have been made to the ori-
ginal case. At the end of this section, the effect of corner rounding
has been studied.
5.3.1. Gradual contraction
A gradual contraction is employed in between inlet pipe and
49
Fig. 12. Predicted erosion damage using different erosion models.
constriction section, assuming its length equals step height, i.e.
4.669 mm. (Fig. 15).
Fig. 16 shows the predicted erosion in the gradual contraction
geometry against the sudden contraction.
Gradual contraction has caused significant reduction of erosion
in the choke section and also a considerable decrease in re-
circulation region. With this geometry excluding contraction sur-
face, maximum erosion belongs to the region downstream of the
reattachment point.
The reason for erosion reduction is the elimination of the high
turbulent kinetic energy region in the choke entrance; causing
reduction of random impingement. (Fig. 17) This leads to the ex-
istence of fewer particles near the wall in the choke section, which
provide fewer particles for impingement in recirculation region
also. (Fig. 18).
5.3.2. Initial Step entrance
Another modification to the choke entrance was putting an
initial step before the main contraction step. The height and length
of this initial step were chosen to be 0.225 and two times of the
50 F. Darihaki et al. / Wear 372-373 (2017) 42–53

Fig. 13. Effect of shear lift force on the predicted erosion (Mansouri et al. erosion Fig. 16. Effect of gradual contraction on predicted erosion (Mansouri et al. erosion
model). model).

Fig. 14. Relative Reynolds number contour on the choke symmetry plane.

Fig. 17. Turbulent kinetic energy contour for gradual contraction geometry.

Fig. 15. Choke domain with gradual contraction.

main step, respectively. Choke construction including initial step

modification is shown in Fig. 19.
Fig. 20 displays predicted erosion for the initial step modifica-
tion. It suggests that dual-step configuration lowered erosion in
both constriction section and recirculation region. However, more
thickness loss occurs in the constriction section compared to the
gradual contraction case.
In addition, it has been found that although the initial step has
been reduced the turbulent kinetic energy region in choke en-
trance, its value is higher than the gradual contraction case which
explains why erosion rate for the dual step geometry is higher
than of the gradual contraction.
Fig. 21 illustrates thickness loss rate on constriction and outlet
pipe walls. It is clear that in the modified cases, erosion in con-
striction section is reduced by a large margin, but changes in the
outlet section are less noticeable.
Fig. 18. Solid volume fraction contour in sudden expansion section. a. Sudden
contraction case, b. Gradual contraction case.
In the choke geometry, highest erosion rate belongs to the
contraction plane. This is where the stream-lines will deviate
considerably that the particles cannot follow them and direct
impingement mechanism would dominate. Fig. 22 shows the wall
thickness loss rate on the contraction plane of studied cases. While
the location of maximum erosion has remained unchanged, this
figure demonstrates that modifications also reduced erosion in the
 F. Darihaki et al. / Wear 372-373 (2017) 42–53 51

Fig. 19. Choke configuration including initial step.

Fig. 22. Erosion contour on the contraction wall of: a. Original choke geometry, b.
Gradual contraction, c. Initial step at choke entrance (Mansouri et al. erosion
model).

Table 8
Maximum predicted erosion rate for original and modified cases.
Fig. 20. Effect of step entrance modification on predicted erosion (Mansouri et al.
erosion model). Case Maximum erosion rate Maximum erosion reduc-
[mm/yr] tion percent [%]

Original 96930.0 –
Gradual contraction 11928.6 88%
Initial step entrance 27016.5 72%

rounded corner (Fig. 23).

Fig. 24 shows predicted thickness loss rate for the rounded
corner case. This figure suggests that corner rounding lowered
erosion rate to the values close to gradual contraction case.
However, the difference is that here Maximum Erosion Reduction
Percent equals 56%. So the rounding process might be still going
on. For better long-term erosion prediction of choke case, dynamic
simulation is desirable which accounts for the effects of geometry
change and flow on each other.

Fig. 21. Erosion contours for: a. Original choke geometry, b. Gradual contraction, c.
Initial step at choke entrance (Mansouri et al. erosion model).

contraction plane to a great extent.

Table 8 lists maximum erosion of three cases. It indicates there
is a significant erosion reduction with both modifications.

5.3.3. Corner rounding effect

Figs. 21, 22, and Table 8 show that erosion rate on the edge of
contraction wall is much greater than other areas. Such high local
erosion rates will cause rounding of the choke entrance corner in
the long-term. So, a new profile is designed and considered for the Fig. 23. Rounded corner profile in the contraction plane.
52 F. Darihaki et al. / Wear 372-373 (2017) 42–53
Fig. 24. Effect of corner rounding on predicted erosion.
5.4. Effect of wall material on thickness loss rate
To have a measure for erosion damage of such component in oil
and gas production system, a choke constructed out of carbon
steel with same dimensions was chosen. The material has been
considered to be AISI 1018 carbon steel which has the density and
Brinell hardness of 7870 kg/m3 and 126, respectively. Due to the
long operating time in production, rounded corner profile which
was studied in Section 5.3.3 has been applied to the contraction
plane.
Fig. 25 shows the predicted erosion rate for both carbon steel
and aluminum wall materials. It points that thickness loss rate for
the carbon steel case is one-third of the aluminum one.
Maximum erosion for both cases in the contraction plane has
been listed in Table 9.
Fig. 25 and Table 9 suggest that erosion damage of a choke
made of carbon steel is much less than the aluminum choke.
However, it should be mentioned that the predicted erosion rates
are completely beyond the acceptable limits in the oil and gas
industry. This is because of the liquid-solid flow conditions applied
in the erosion test such as the high velocity of the slurry flow
Fig. 25. Comparison of predicted thickness loss rate for aluminum and carbon steel
cases.
Table 9
Maximum predicted erosion rate for chokes with different wall materials.
Material Maximum erosion rate [mm/yr] Ratio [-]
Aluminum 42238.8 1
AISI 1018 carbon steel 13250.8 0.14
which is accompanied by high levels of turbulence kinetic energy
in some sections of the choke. Also, it should be noticed that the
dimensions of the test geometry are much smaller than what
might be found in the oil and gas production systems. In addition,
it should be considered that aluminum and carbon steel have
different applications and they act differently regarding corrosion
process. Aluminum is known to be resistant to oxidation through
the passivation phenomenon. While in environments which con-
tain hydrogen sulfide (H2S) like oil and gas production systems,
carbon steel may experience more damage due to cathodic
cracking.
6. Conclusion
The Eulerian-Lagrangian approach was utilized for two-phase
liquid-solid flow calculation and erosion prediction in a choke
geometry. A User-Defined Function (UDF) was written and em-
ployed in different parts of particle tracking approach, including
particle injection, shear lift force and its correction, particle re-
bound and erosion calculation using different models. Solutions
implied that only with the standard k − ϵ model, stochastic par-
ticle tracking scheme is able to calculate the velocity fluctuations
properly and model the random impingement mechanism. While
the other studies on particle erosion have employed lift force
without the correction factor, in this study a correction factor was
applied to calculate the lift force. However, the results showed that
it does not improve the predicted erosion, as the flow conditions
are beyond the range of its application. Comparison of the pre-
dicted erosion rates using different erosion models, suggested
Mansouri et al. erosion model estimates both erosion pattern and
magnitude more accurately. The solution showed two peaks for
erosion rate in downstream of the choke, located at both sides of
the reattachment point.
Two modifications to the choke entrance including gradual
contraction and initial step entrance were examined and the
predicted erosion for the designed profiles showed significant
reduction of erosion rate in different parts of the geometry. This
includes a roughly 90% reduction of maximum erosion rate with
gradual contraction case. It was shown that the elimination of high
turbulent kinetic energy region at choke entrance is responsible
for erosion reduction. Effect of corner rounding for the long-term
erosional damage was studied and a comparison of the predicted
erosion rates with carbon steel material which might be found in
oil and gas production systems showed that thickness loss rate is
one-third of the aluminum for the choke walls and much smaller
for the contraction plane.
Acknowledgment
The present study is part of an investigation carried out in
collaboration with National Iranian South Oil Company (NISOC).
The authors gratefully acknowledge support from NISOC.
 F. Darihaki et al. / Wear 372-373 (2017) 42–53
References
[1] J.K. Edwards, DevelOpment, Validation, and Application of A Three-dimen-
sional, CFD-based Erosion Prediction Procedure, Mechanical Engineering,
University of Tulsa, Tulsa, OK, USA, 2000.
[2] S.S. Nugroho, CFD-based Erosion Prediction Procedure for Use in Sudden Ex-
pansions, Mechanical Engineering, University of Tulsa, Tulsa, OK, USA, 2001.
[3] A. Founti Maria, S. Klipfel Athinodoros, Shear Lift Effects on the Particle Motion
in Two-phase Sudden Expansion Flows, ASME FED, Numerical Methods in
Multiphase Flows, 1994, pp. 81-92.
[4] M.Sommerfeld, Numerical Simulation of the Particle Motion in Turbulent
Flow, Second International Lecture Course on Multiphase Flow, Tokyo, Japan,
1989.
[5] C.Y. Wong, C. Solnordal, A. Swallow, J. Wu, Experimental and computational
modelling of solid particle erosion in a pipe annular cavity, Wear 303 (2013)
109–129.
[6] Y. Zhang, B.S. McLaury, S.A. Shirazi, Improvements of particle near-wall velo-
city and erosion predictions using a commercial CFD code, J. Fluids Eng. 131
(2009) 031303.
[7] A. Mansouri, H. Arabnejad, S. Karimi, S.A. Shirazi, B.S. McLaury, Improved CFD
modeling and validation of erosion damage due to fine sand particles, Wear
338 (2015) 339–350.
[8] B.S. McLaury, Predicting Solid Particle Erosion Resulting from Turbulent
Fluctuations in Oilfield Geometries, Mechanical Engineering, University of
Tulsa, Tulsa, OK, USA, 1996.
53
[9] Y. Oka, T. Yoshida, Practical estimation of erosion damage caused by solid
particle impact: part 2: mechanical properties of materials directly associated
with erosion damage, Wear 259 (2005) 102–109.
[10] Y.I. Oka, K. Okamura, T. Yoshida, Practical estimation of erosion damage caused
by solid particle impact: part 1: effects of impact parameters on a predictive
equation, Wear 259 (2005) 95–101.
[11] Y. Zhang, Application and Improvement of Computational Fluid Dynamics
(CFD) in Solid Particle Erosion Modeling, Mechanical Engineering, The Uni-
versity of Tulsa, Tulsa, OK, USA, 2006.
[12] R.E. Vieira, A. Mansouri, B.S. McLaury, S.A. Shirazi, Experimental and compu-
tational study of erosion in elbows due to sand particles in air flow, Powder
Technol. 288 (2016) 339–353.
[13] H. Arabnejad, A. Mansouri, S. Shirazi, B. McLaury, Development of mechanistic
erosion equation for solid particles, Wear 332 (2015) 1044–1050.
[14] A. Mansouri, H. Arabnejad, S. Shirazi, B. McLaury, A combined CFD/experi-
mental methodology for erosion prediction, Wear 332 (2015) 1090–1097.
[15] M. Sommerfeld, Particle Motion in Fluids, VDI-Buch: VDI Heat Atlas, Part, 11,
2010, pp. 1181-1196.
[16] R. Mei, An approximate expression for the shear lift force on a spherical
particle at finite Reynolds number, Int. J. Multiph. Flow. 18 (1992) 145–147.
[17] D.S. Dandy, H.A. Dwyer, A sphere in shear flow at finite Reynolds number:
effect of shear on particle lift, drag, and heat transfer, J. Fluid Mech. 216 (1990)
381–410.
[18] A. Forder, M. Thew, D. Harrison, A numerical investigation of solid particle
erosion experienced within oilfield control valves, Wear 216 (1998) 184–193.