You are on page 1of 24

Journal of Natural Gas Science and Engineering 21 (2014) 850e873

Contents lists available at ScienceDirect

Journal of Natural Gas Science and Engineering


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

Invited review

A comprehensive review of solid particle erosion modeling for oil and


gas wells and pipelines applications
Mazdak Parsi a, Kamyar Najmi b, *, Fardis Najafifard c, Shokrollah Hassani d,
Brenton S. McLaury a, b, Siamack A. Shirazi a, b
a
Erosion/Corrosion Research Center, The University of Tulsa, Tulsa, OK, USA
b
Tulsa University Sand Management Project, The University of Tulsa, Tulsa, OK, USA
c
Institute for Corrosion and Multiphase Technology, Athens, OH, USA
d
BP America Inc., Houston, TX, USA

a r t i c l e i n f o a b s t r a c t

Article history: Sand is commonly produced along with production fluids (oil and gas), and this is a major problem for
Received 20 August 2014 the oil and gas industry. Sand production is a concern, since it can bring about a variety of problems.
Received in revised form Amid them, three problems stand out above all: pressure drop, pipe blockage, and erosion. The latter is a
1 October 2014
complex mechanical process in which material is removed from the pipeline due to repeated sand
Accepted 7 October 2014
Available online 30 October 2014
particle impacts. As a result, the pipeline can be eroded. Eroded pipelines may cause pipe failures which
can result in financial losses and environmental issues. Therefore, it is important to know what pa-
rameters govern the erosion phenomenon and how it can be modeled. The present work describes key
Keywords:
Erosion equation
factors influencing erosion and reviews available erosion equations. Furthermore, empirical and mech-
Sand particle erosion anistic models for erosion prediction in pipelines are discussed. These models are used by oil and gas
Multiphase flow companies to limit the maximum production flow rates and avoid excessive erosion damage. Compu-
Single-phase flow tational fluid dynamics (CFD) based erosion modeling as a comprehensive method for erosion studying is
CFD-based erosion modeling explained as well. Finally, possible limitations and gaps in knowledge concerning erosion are indicated.
The current work can be used by oil and gas companies as a comprehensive review of erosion challenges
and remedies. Of course, further studies must be undertaken in order to expand the knowledge of
erosion and find applicable models for erosion damage prediction and prevention.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction results of sand production is sand erosion. Sand erosion can cause
failure of equipment, leaks in pipelines resulting in environmental
Production of sand from oil and gas wells happens very often. disasters and potential injury to personnel. Therefore, predicting
Sand production can cause severe financial and environmental solid particle erosion rate is a helpful tool in designing and selecting
difficulties. Production of sand can result in erosion, blockage of equipment to prevent failures.
pipelines, under sand deposit corrosion, disposal and other com- Predicting solid particle erosion in gases and liquids is a chal-
plications. An approach to deal with sand can be eliminating sand lenging task. Despite all the resources that have been spent to
using sand screens and gravel packs. However, there are many investigate and study erosion, the solid particle erosion mechanism
technical and financial problems to eliminate or control sand pro- is still not fully understood. A variety of models and approaches
duction especially at downhole conditions. For example, sand have been proposed by researchers. Usually, erosion prediction
screens cannot prevent smaller particles (less than 50 mm, keeping models are divided into three categories: empirical, mechanistic
in mind that this value varies with screen dimensions) from being and CFD-based. Since erosion is complicated, most proposed
entrained with the produced fluids. These small particles can pass erosion prediction models are a combination of all these categories.
through sand screens or block a portion of the screen causing The main objective of this paper is to provide a comprehensive
higher velocities in other sections resulting in erosion. This process review of literature concerning solid particle erosion modeling. The
makes the screen openings larger allowing larger particles to pass, paper is structured as follows. Section 2 discusses important pa-
erode the sand screen and cause failure. One of the important rameters for the prediction of solid particle erosion. Section 3 is
divided into different subsections. It begins with an overview of
available erosion equations in the literature. Then, various empir-
* Corresponding author.
E-mail address: kamyar-najmi@utulsa.edu (K. Najmi). ical and mechanistic erosion prediction models are surveyed.

http://dx.doi.org/10.1016/j.jngse.2014.10.001
1875-5100/© 2014 Elsevier B.V. All rights reserved.
M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873 851

Finally, CFD-based erosion modeling is discussed. In Section 4, an and target wall material characteristics to the mass loss of the wall.
evaluation of some of the surveyed models is presented. The next The following discussion describes some of the important param-
section suggests some areas for further work. eters that influence erosion. Before erosion data is considered it
should be noted that erosion data is reported differently by various
2. Mechanism of solid particle erosion investigators. For example, erosion rates are normally mass loss of
materials or thickness loss of materials as a function of time such as
When a particle impacts a surface, it scars the surface. Shapes of kg/hr or mils per year or mm per year. Some authors report erosion
these scars depend on many parameters including surface material, data in the units of mass loss, volume loss or thickness loss per
particle size, and impact angle. Researchers studied these scars to mass of impacting particles such as g/g or mils/lb (in/1000 lb), etc.
explain the mechanism of erosion and generally agree that the The latter will be referred to as “erosion ratio” in this manuscript.
mechanism of erosion changes based on the ductility of the surface.
Finnie (1958) proposed a micro-geometry model for ductile mate- 2.1.1. Particle properties
rials. He suggested that erosion in ductile materials is the result of Particles properties such as size, density, hardness, and shape
micro-cutting. When a particle impacts a surface at a low impact have significant influence on solid particle erosion. To have a better
angle, it creates a crater. Other particle impacts make the crater understanding of the influence of particle properties on erosion,
larger and also pile up material around the crater. The piled up the effect of each particle property on erosion needs to be inves-
material is eventually removed by continued particle impacts. The tigated separately.
micro-geometry model under predicts erosion magnitude from the
particles which impact the surface at higher angles compared to
experimental data. Later, Finnie (1960) modified the model to 2.1.1.1. Particle shape. It has been observed that particle shape has a
address this weakness. Based on the erosion micro-geometry model, significant effect on the magnitude of erosion. Salik et al. (1981)
work hardening of the metal surface should decrease the erosion showed that it can change the erosion magnitude by an order of
rate. However, Levy (1995) showed that initial erosion is lower than magnitude. Levy and Chik (1983) observed the same behavior and
erosion from previously eroded surfaces. Bellman and Levy (1981) reported that the sharpness of particles has a huge influence on the
proposed a macroscopic erosion mechanism. They suggested that magnitude of erosion. They employed two different particle shapes,
particles hitting the surface create shallow craters and platelet-like sharp angular particles and spherical particles. The erosion results
pieces. These platelets are easy to separate from the surface by from angular particles were four times larger compared to erosion
subsequent particle impact (Fig.1). During the formation of platelets, results from round particles. It also has been reported that the
adiabatic shear heating on the surface and work-hardening under impact angle that results in maximum erosion depends on particle
the surface occur. The occurrence of these two processes helps shape and varies based on particle angularity (Hutchings et al.,
platelet formation which explains the higher erosion rate for the 1976). A particle shape factor is introduced in most of the erosion
steady-state condition compared to the initial erosion rate. ratio equations proposed by researchers, since the shape of the
Other solid particle erosion mechanisms for ductile materials particle has a pronounced influence on erosion magnitude.
are suggested by researchers and can be found in literature (Chase
et al., 1992; Hutchings, 1980; Andrews, 1981; Jahanmir, 1980). 2.1.1.2. Particle size. Particle size is another important particle
Unlike the solid particle erosion mechanism for ductile mate- property which influences erosion magnitude because larger par-
rials, there is wide acceptance of the erosion mechanism for brittle ticles have larger kinetic energies even if they strike a target with
material. It has been suggested that in brittle material, erosion is the same velocity as the smaller particles. Some erosion data is
due to crack formation (Srinivasan and Scattergood, 1988; reported as a function of particle size (Tilly, 1973) as shown in Fig. 3.
Sundararajan 1991, Kleis and Kulu, 2008). When a particle hits a This figure indicates that the erosion ratio (mass of eroded mate-
brittle surface, it creates lateral and radial cracks. Other impacts rials/mass of impacting particles) is nearly independent of particle
cause these cracks to grow. These cracks divide the surface into sizes when the particles are larger than approximately 100 mm.
smaller pieces which can be removed by other particles impacting Gandhi and Borse (2002) investigated the effect of sand size on
the surface (Fig. 2). cast iron erosion behavior for two different impact angles of 30
and 75 . The carrier fluid velocity was 3.62 m/s and sand concen-
2.1. Important parameters in predicting solid particle erosion tration was 20 wt% (Fig. 4). They observed a linear relation between
sand size and erosion rate. This behavior was reported by other
Many parameters have been found that influence erosion. Based researchers as well (Elkholy, 1983; Clark, 1991). These results are
on these parameters, researchers have proposed different erosion influenced by the fact that impact velocity of particles is not con-
ratio equations which relate particle characteristics (shape, size, stant and changes with particle size when particles are entrained in
material, density, hardness, etc.), particle impingement information liquid streams.

Fig. 1. Schematic of erosion procedure in ductile material (a): before the impact, (b): crater formation and piling material at one side of the crater, (c): separation of material from
the surface.
852 M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873

Fig. 2. Expected mechanism of erosion in brittle material: (a) growth of cone crack and median cracks (b) closure of median and creation of lateral cracks, (c) eroded crater formed
(Sooraj and Radhakrishnan, 2013).

Desale et al. (2009) examined the effect of particle size on the viscous sub-layer and can become trapped in the sub-layer. This
erosion performance of an aluminum alloy (AA 6063) for eight can cause small particles to travel next to the wall impacting over
different sizes of silica sand in the range of 37.5e655 mm. The car- and over again, but this behavior is strongly geometry dependent.
rier fluid velocity was 3 m/s, and sand concentration was 20 wt%. Erosion of small particles may also be influenced by the parti-
They chose two different impact angles of 30 and 90 (Fig. 5). They cleeparticle interactions as the number of particles increase
concluded that at a constant sand concentration (wt%), increasing significantly, for the same mass of impacting particles, as the par-
the particle size increases the erosion rate even though it reduces ticle size approaches zero.
the number of particles and impacts. Particle impact velocity and As a conclusion, usually for similar conditions and similar
kinetic energy per impact are affected by particle size (Fig. 5). Un- impact velocity, smaller sand particles cause lower erosion rates as
fortunately, when considering slurry erosion, the mass loss can be compared to larger particles because of lower kinetic energy.
influenced by other factors such as particle impact velocities However, it needs to be considered that small sharp particles can be
varying along the targets, fluid viscosity effects, and also the con- more erosive than large semi-rounded or rounded particles. These
centration of particles and particleeparticleetarget interactions. small particles are more influenced by turbulence, and they can
Desale et al. (2009) proposed a correlation between erosion rate cause more erosion in geometries with high turbulence and recir-
and particle size which has been used by several researchers. culating areas. In general, both linear and power-law relations be-
tween sand size and erosion rate have been suggested by several
investigators. This relation has been influenced unfortunately by
Erosion RatefðParticle SizeÞn (1)
several factors including target material properties, experimental
The value for “n” can be anywhere between 0.3 and 2.0 and conditions, particle impact velocity, particleeparticle interactions,
depends on differences in material properties, experimental con- fluid properties, and sand size distribution.
ditions, particle velocity and even particle size and size distribution
(Desale et al. (2009)). In most cases, “n” is considered to be one, and 2.1.1.3. Particle material. Levy and Chik (1983) studied the effect of
a linear relation is observed between sand size and erosion rate. particle composition on erosion behavior of AISI 1020 carbon steel.
Oka et al. (2005) and Oka and Yoshida (2005) introduced a Five different brittle particles with angular shape and size range of
particle size factor in their erosion ratio (not erosion rate) model 180e250 mm were used to erode the steel at two different flow angles
which was raised to the 0.19 power. of 30 and 90 and flow velocity of 80 m/s (Fig. 6a). Erosion ratio was
In general, smaller sand causes lower erosion rates because they
have smaller kinetic energy and impact force to erode the surface.
However, particle density, shape, and hardness also affect erosivity,
but in general larger sand causes more erosion damage for similar
impact speed, and shape, density, and hardness. Smaller sand
particles are also more affected by turbulence. The exchange of
momentum between fluid and particle is more efficient for smaller
particles, so they respond to fluctuation in the flow more easily.
Smaller particles also lose more momentum as they flow through

Fig. 4. Effect of sand size on erosion rate at flow velocity of 3.62 m/s, sand concen-
Fig. 3. Erosion ratio versus particle size and particle impact velocity (Tilly, 1973). tration of 20 wt% (Gandhi and Borse, 2002).
M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873 853

 n
Ht
Erosion Ratef (2)
Hp

where Ht is the target material hardness and Hp is erodent particle


hardness. n is an empirical constant. Later, Shipway and Hutchings
(1996) also examined the effect of particle and target wall materials
on erosion behavior. They examined different materials with
different hardness. They have concluded that increasing the ratio of
erodent to target hardness toward unity rapidly increases the
erosion rate and reduces the velocity exponent of erosion rate.
Density of erodent particle is another important factor in
determining erosivity. Particles with higher density have higher
kinetic energy and create more impact force which cause higher
erosion rates. Steel grit has higher density as compared to Alumina
(Al2O3), SiC and SiO2 and consequently causes higher erosion rate
(Table 1). SiC, SiO2 and Al2O3 have almost the same density and
cause almost the same erosion rate for carbon steel target material
when particles shape and size are similar (Table 1).
In general, composition of particles determines their density
Fig. 5. Effect of sand size on erosion rate, number of striking particles and kinetic and hardness which are important factors in erosivity of erodent
energy of impact on AA 6063 target material in the presence of 20 wt% silica sand at
particles. Shape and size of particles are also important parameters
velocity of 3 m/s and two different impact angles of 30 and 90 (Desale et al., 2009;
Lynn et al., 1991). for erosion. For particles with similar size and shape, higher density
and hardness usually cause higher erosion rates (depending on
hardness of the target material). Higher density increases the ki-
low when soft erodents such as calcite (CaCO3) and apatite (Ca5(PO4)3)
netic energy and impact force and increases the erosion rate.
were used to erode the steel. Apatite has higher hardness compared to
Higher hardness (up about 700 HV) increases the erosion rate.
calcite and causes slightly more erosion than calcite (Table 1).
Barite, calcium carbonate, iron carbonate, and silica are some of
However, when Vickers hardness of particle reaches about
the solid particles occasionally produced during oil and gas produc-
700 HV, increases in hardness of erodent particles do not consid-
tion. Barite has a Mohs hardness number of 3 which is similar to
erably increase the erosion rate. This can be due to the fact that
calcite. Therefore, it has a similarly low erosivity in terms of hardness
particles are hard enough not to shatter when they strike the target
effect for the same size, shape and density of particle. Iron carbonate
steel. It has been observed that there is a correlation between the
has a Mohs hardness number of 3e4 which is similar to calcite and
hardness of target wall, hardness of particles and erosion damage.
barite, and it is expected to have a similarly low erosivity. Barite and
SiO2 is as erosive as SiC for AISI 1020 carbon steel target material
iron carbonate are expected to shatter and cause small amounts of
even though it has a hardness four times lower on the Vickers scale.
erosion when they impact a carbon steel target. On the other hand,
However for harder target materials such as WC-12Co, SiC and SiO2
silica and silicate have higher hardness and therefore higher erosivity.
particles have different erosivities (Babu et al., 2011) (Fig. 6b).
For harder target materials such as WC-12Co coating, silicate causes
When a soft particle impacts a wall, it may shatter into smaller
considerably more erosion than silica because of its higher hardness
particles with lower mass and kinetic energy and consequently
even though they have similar erosivity for a carbon steel target.
lower erosivity. SEM analysis of a steel surface after an erosion
attack also showed that break up and adherence of soft particles to ^ ^ ^ ^ ^
E Barite zE Calcium Carbonate zE Iron carbonate < E Silica zE Silicate
the surface reduces the kinetic energy and also covers the target
surface with a layer of fragment particles which also reduces the ðValid for carbon steel targetÞ
erosion rate (Levy and Chik, 1983).Therefore, harder particles can
^ ^ ^ ^ ^
Barite zE Calcium Carbonate zE Iron carbonate < E Silica < E Silicate
be more erosive because they shatter less when they impact the E
target wall. However, beyond a critical hardness value where par- (3)
ðValid for WC  12Co coating targetÞ
ticles no longer shatter, a further increase in hardness does not
increase its erosivity. ^ denotes erosivity of the particle material.
where E
Wada and Watanabe (1987) proposed the following correlation
can be used to explain the relationship between erosion rate and
2.1.2. Fluid properties
ratio of target to particle hardness
Carrier fluid may have the largest influence on erosion ratio
because carrier fluid can affect impact velocity of particles. In
practical applications involving solid particle erosion, particles are
carried to the wall by a carrier fluid. The characteristics of the
Table 1 carrier fluid such as viscosity and density affect particle behavior
Erodent particles and erosion ratio of AISI 1020 steel (Levy and Chik, 1983). and as a result influence erosion pattern and magnitude. The
Particle Density Mohs Vickers Erosion ratio (g/g) amount of carrier fluid influence on erosion depends on geometry
composition (g/cm3) hardness hardness and flow pattern. For example in a direct impingement geometry,
a ¼ 30 a ¼ 90
the carrier fluid moves towards the wall. Therefore, the carrier fluid
CaCO3 3 115 0.03
e e
picks up particles and guides them towards the wall. In other ge-
Ca5(PO4)3 e 5 300 0.5 0.3
SiO2 2.7 7 700 3.0 1.6 ometries like straight pipe, the carrier fluid moves parallel to the
Al2O3 4.0 9 1900 2.6 1.4 wall. In this geometry, turbulent fluctuations are responsible for
SiC 3.2 >9 3000 3.3 1.9 changing particles' paths and move them toward the wall. Many
Steel grit 7.9 e e 5.3 e researchers studied the effect of fluid properties on particle
Steel shot 7.9 1.4
behavior (Hinze, 1972; Humphrey, 1990). Also fluid properties
e e e
854 M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873

directly affect the local particle concentration. This means that even on erosion varies based on the surface material. Fig. 7 shows cor-
if the overall particle concentration is low, the local particle con- relations between particle impact angle and erosion rate for ductile
centration can be high due to the flow pattern and can affect the and brittle materials. Erosion rate trends for brittle materials are
erosion magnitude and erosion pattern. Higher local particle con- different from ductile materials. For ductile materials, higher
centration means more particleeparticle interactions, and the erosion rates occur at lower impact angles. This is due to more
shielding effect observed by Chen (2004) can occur. Particle wall efficient formation and cutting of platelets by particles at lower
interaction is directly influenced by the fluid boundary layer. Clark angles. On the other hand, maximum erosion for brittle material
and Burmeister (1992) proposed a squeeze film model that includes occurs at near normal impact angle because the dominant cause of
the effect of liquid boundary layer on particle behavior. erosion in brittle materials is cracking. Most materials used in the
oil and gas industry have characteristics of both ductile and brittle
2.1.3. Target wall properties materials. As a result, a variety of angle functions have been pro-
Despite much experimental data, the correlation between target posed by researchers. Most of the proposed angle functions are
material and solid particle erosion rate is still not clear. It has been empirical and are only valid for limited conditions. Since most of
believed that having materials with higher hardness results in the angle functions are empirical, it is essential to choose an angle
higher erosion resistance (Finnie et al., 1967). However, Levy and function based on carrier fluid as well as particle and target wall
Hickey (1982) showed that a material with higher hardness can properties.
result in a higher erosion rate as compared to materials with lower
hardness. Based on his observations, he proposed that ductility 2.1.6. Temperature effect
allows the surface to distribute particle impact kinetic energy by Different hypotheses are proposed to explain the role of tem-
plastic deformation which can result in lower erosion rate. perature on the erosion mechanism. Smeltzer et al. (1970) observed
Toughness of the target material may be a better indicator of that erosion rate decreases by increasing temperature. Later, Levy
erosion performance, since increasing hardness may reduce the (1979) suggested that metal ductility increases by increasing tem-
ductility and consequently increase erosion rate caused by the perature. Therefore when particles impact the wall, more of their
brittle mechanism. Increasing toughness indicates increasing kinetic energy is absorbed by plastic deformation. However, the
hardness of material without reduction of ductility. It would be exact effect of temperature on erosion is not certain but it is
more reasonable to develop a correlation for material toughness considered as having limited influence.
and its erosion performance. Foley and Levy (1983) observed that if
the local strength of the target material is less than a specific value, 2.1.7. Particleeparticle interaction
particles are able to remove material with plastic deformation. In most models, researchers neglect particleeparticle in-
teractions, but in slurry erosion it has an important effect on
2.1.4. Particle impact speed erosion magnitude. The effect of particleeparticle interactions on
Erosion rate has direct relation with particle impact velocity. erosion has been observed by many researchers (Andrews and
Horsfield, 1983; Liebhard and Levy, 1991; Shipway and Hutchings,
Erosion RatefðVL Þn (4) 1994; Chen, 2004). Researchers proposed different mechanisms
which explain the relation between erosion rate and sand con-
where VL is the particle impact velocity and n is a constant. centration. It has been suggested that at high sand concentrations,
Different researchers have proposed different values for n. Finnie when particles rebound from the wall, they hit particles that move
(1958) suggested n is equal to 2 based on theoretical work. Later, toward the wall and slow them down. This phenomenon is named
Laitone (1979a, 1979b) suggested that due to the change in car- “shielding” (Brown et al., 1981; Deng et al., 2005). Therefore,
rier fluid velocity close to the experimental surface, the n constant depending on the fluid and geometry conditions, higher sand
can raise to 4. Others observed that n can vary from 0.3 to 4.5 concentrations may result in lower erosion rate.
(Smeltzer et al., 1970; Burnett et al., 1994). More recently, it has
been suggested that n is not constant and depends on the hardness 3. Solid particle erosion modeling
of the eroded material (Oka et al., 2005, Oka and Yoshida (2005).
Typical values used for n varies from 1.6 to 2.6. 3.1. Erosion equations

2.1.5. Particle impact angle Due to industrial importance, there have been many studies
Many researchers have observed that erosion rate is also a regarding solid particle erosion available in the literature. While
function of particle impact angle. The effect of particle impact angle some of these studies investigated the phenomena experimentally,

Fig. 6. Effect of erodent hardness on erosion behavior of (a) AISI 1020 steel (Levy and Chik, 1983) (b) WC-12Co (Babu et al., 2011).
M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873 855

Fig. 7. Erosion rate of different materials versus the particle impact angle; particle impact velocity is 153 m/s, Silicon Carbide Grit (Sheldon, 1970).

there are other studies which tried to describe solid particle erosion compared to the depth of cut, iii) constant plastic flow is reached
mathematically by proposing models. On the theoretical side, pre- upon impact of particles. This model predicts no erosion would
vious studies have tried to describe characteristics of solid particle occur for normal impingement of particles. Finnie (1960) also
erosion by proposing various scenarios and mechanisms. Maybe compared the predicted volume removal for a single abrasive grain
that is why there are a large number of physical parameters that have with experimental data by many grains and stated that the single
been incorporated in solid particle erosion modeling. Meng and grain analysis can be used to predict erosion caused by many grains
Ludema (1995) published a review paper regarding friction and for low angle erosion.
wear. In that study, they focused more specifically on erosion Bitter (1963a, 1963b) proposed repeated deformation and cut-
equations and selected 28 equations including 33 parameters based ting as two mechanisms for erosion and developed two models to
on the criteria which were defined in their paper. In this study, in predict erosion rate causes by these two mechanisms for ductile
addition to some of the models mentioned by Meng and Ludema and brittle materials. In the first part of his study, he developed an
(1995), we also discuss some other models proposed after their equation, based on balance of energy for a plasticeelastic collision,
study which are more commonly used in the oil and gas industry. to predict erosion caused by the deformation mechanism:
The same nomenclature as proposed by Meng and Ludema (1995) is
 2
adopted in this study to make it easier for the readers to follow both 1 M Vp sin a  K
studies. Table 2 shows different physical parameters used in the εVP ¼ (6.a)
2 d
erosion equations reviewed in the current study.
Finnie (1960) proposed one of the earliest erosion equations He compared the model prediction with experimental data and
that rather fundamentally investigated the erosion of solid particles reported that there is good agreement between the model pre-
on a material surface. In his study, he stated that erosion on surface diction and the erosion data for brittle substances at different
of a material struck by solid particles depends on the motion of the impact angles.
particles and the material properties. He divided material into two In the second part of his study, he considered two scenarios for
categories: ductile where erosion is caused by plastic deformation impacting particles including particles that still have a horizontal
and brittle where intersection of cracks is the main reason of velocity component when they leave the body surface and particles
erosion. Based on some assumptions, he proposed two equations to with zero horizontal velocity component during the collision. He
predict volume of removed material by a single abrasive grain for proposed two models for the cutting mechanism based on these
low and high particle striking angles as follows: scenarios as:
 2 8  2 9
  2MVp Vp sin a  K < C Vp sin a  K =
mp Vp2 6 k εVC1 ¼  1=2 Vp cos a    c
εVP ¼ sinð2aÞ  sin2 ðaÞ for tan a  (5.a) Vp sin a : 1
Vp sin a 2 ;
Pjk k 6
for a  ap0
2  (6.b)
m p Vp k cos2 a k
εVP ¼ for tan a  (5.b)
Pjk 6 6 h  3=2 i
1M Vp2 cos2 a  K1 Vp sin a  K
For developing these equations, the following assumptions were 2
εVC2 ¼ for a  ap0
made: i) the ratio of the depth of contact to the depth of cut is c
constant, ii) the width of particle cutting face is uniform and is large (6.c)
856 M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873

Table 2
Physical parameters used in erosion models.

Physical property Equation number

5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

Particle density ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Target hardness ✓
Moment of inertia ✓
Roundness ✓ ✓ ✓ ✓
Grain mass ✓ ✓ ✓ ✓
Particle size ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Particle velocity ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Rebound velocity ✓
Target density ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Target hardness ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Flow stress ✓ ✓ ✓ ✓
Young modulus ✓ ✓ ✓
Fracture toughness ✓ ✓ ✓ ✓
Critical strain ✓ ✓ ✓ ✓
Depth of deformation ✓ ✓
Incremental strain per impact ✓
Thermal conductivity ✓
Melting temperature ✓ ✓
Enthalpy of melting ✓
Cutting energy ✓ ✓ ✓
Deformation energy ✓ ✓ ✓
Erosion resistance ✓
Heat capacity ✓ ✓
Grain molecular weight ✓
Impact angle ✓ ✓ ✓ ✓ ✓ 90 90 ✓ ✓ ✓ 90 ✓ ✓ ✓ ✓
Impact angle max wear ✓ ✓
KE transfer from particle to target ✓
Temperature ✓ ✓
pressure ✓ ✓
Strain hardening
Friction coefficient ✓
Critical friction coefficient ✓
Number of impacts ✓
Poisson coefficient ✓
Critical Poisson coefficient ✓

Total erosion rate is the summation of erosion caused by these these parameters can be implemented in a model (some of them
two mechanisms (repeated deformation and cutting). are not controllable or measurable). He solved the equation of
motion of an abrasive grain and proposed the following equation
εVT ¼ εVP þ εVC (6.d) for erosion rate prediction:

Neilson and Gilchrist (1968) proposed two models for erosion  


prediction of particles at small and large angles of attack based on
cMVp2 x_0
εV ¼    cos2 a  t (8.a)
previous work by Finnie (1960) and Bitter (1963a, b). It was sug- 4P 1 þ mp r 2 I Vp
gested that the normal component of kinetic energy of impacting
particles causes deformation wear while the parallel component of 2Vp
kinetic energy causes cutting wear. He introduced deformation and x_0t ¼ Vp cos a  sin a (8.b)
P
cutting wear factors and suggested that total erosion on a surface is
equal to the summation of erosion due to deformation and cutting Finnie (1972) discussed two possibilities that cease the cutting
mechanisms. In the following equations, the first term and the mechanism. The first possibility is that cutting terminates when the
second term on the right hand side of the equations represent horizontal velocity component of particle tip is zero. The other
erosion caused by cutting and deformation, respectively. possibility is that the particle leaves the surface while the tip is still
moving horizontally. Based on these two possibilities, he further
h i
1M  2 simplified the above equation to
2 Vp2 cos2 a  Vr2 1 M Vp sin a  K
εV ¼ þ for a < ap0 cMVp2
c 2 d εV ¼   cos2 a for x_0t ¼ 0 (8.c)
(7.a) 4P 1 þ mp r 2 I

 2  
1 MV 2 cos2 a 1 M Vp sin a  K cMVp2 2 sin2 a
εV ¼ 2
p
þ for a > ap0 (7.b) εV ¼    sinð2aÞ  for yt ¼ 0 (8.d)
c 2 d 4P 1 þ mp r 2 I P

He finally mentioned that further study is needed to define This model has low accuracy as contact angle approaches 90 .
deformation and cutting wear factors. In another study, Sheldon and Kanhere (1972) developed an
Finnie (1972) modified his previous model (Finnie, 1960) to erosion model based on indentation hardness characteristics of
predict erosion for ductile metals. In that study, the parameters that materials for single particles. He used an empirical correlation be-
may affect ductile erosion were listed and it was stated that not all tween applied load and diameter of indentation proposed by Meyer
M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873 857

(1908). It was assumed that this correlation is valid for erosion at εm ¼ 5:82  1010 Vp2:9 (12)
low velocities and normal impact angle. By equating maximum
value of kinetic energy to the work expended during indentation Evans et al. (1978) studied the impact damage in brittle mate-
and using some algebraic simplifications and experimental data, he rials in the elasticeplastic response regime. They analyzed radial
proposed the simple following model to predict material removal: cracks (surface extension), lateral cracks (penetration) and fracture
thresholds using a simplified discussion of impact dynamic features
3=2 and fracture mechanics in order to characterize the fracture caused
d3p Vp3 rp
εVT ¼ 3=2
(9) by the impact. They also proposed an erosion model by finding a
HV relation between radial crack formation and fracture toughness of
By comparing this model to experimental data it was claimed the target materials:
that the velocity exponent of 3 obtained in his model, matches
19=6 11=3 1=4
more closely to experimental data rather than an exponent of 2 Vp r rp
εVP f 4=3 1=4
(13)
obtained by considering kinetic energy. Kc Ht
Tilly (1973) suggested a two-stage mechanism for erosion of
ductile materials. The first stage of the erosion mechanism occurs A semi-empirical model to predict erosion rate at different
when particles impact the target surface and cut chips from it. The impact angles and velocities was developed by Tabakoff et al.
second stage of erosion is caused by particles impacting a target (1979). They assumed that the erosion process can be character-
and they break up into small fragments around the primary scar ized by two mechanisms at small and large impingement angles. It
caused by the first stage. The extent of fragmentation is a function was pointed out that the proposed model is applicable to small,
of particle size and velocity, and based on that he introduced a intermediate and large impact angles as well as a combination of
particle size and velocity threshold below which no erosion occurs. them. In their model, the effect of particle tangential restitution
For the primary erosion (stage one), he proposed a model based on coefficient as a parameter that affects the erosion rate was taken
an energy balance between the energy required to produce erosion, into account:
energy to produce elastic deformation and initial kinetic energy as     
follows: εm ¼ K1 f ðaÞVp2 cos2 a 1  e2t þ f Vp ; n (14.a)
!2 "  3=2 #2
Vp d0 V0 et ¼ 1  0:0016Vp sin a (14.b)
ε1 ¼ b
ε1 1 (10.a)
Vref dp Vp

  2
Tilly (1973) assumed that secondary erosion only happens when p a
f ðaÞ ¼ 1 þ K4 K2 sin (14.c)
particles break up, and the extent of it is proportional to the extent 2 am
of fragmentation and particle initial kinetic energy.
  4
!2 f Vp ; n ¼ K3 Vp sinðaÞ (14.d)
Vp
ε2 ¼ b
ε2 Fd;V (10.b)
Vref
1; a  3am
K4 ¼ (14.e)
0; a > 3am
W W
Fd;V ¼ 0 (10.c) Another mechanistic model to predict erosion rate for the
W0
erosion of metals impacted by spherical particles at normal angle
Therefore, total erosion is a summation of erosion in stages one was developed by Hutchings (1981). He proposed the following
and two: model by formulating an energy balance equation between
impacting particles and the metal surface:
εV ¼ ε1 þ ε2 (10.d)
art P 0:5 Vp3
!2 "  3=2 #2 !2 εm ¼ 0:033 (15)
Vp d0 V0 Vp U2c Ht1:5
εV ¼ b
ε1 1 þb
ε2 Fd;V (10.e)
Vref dp Vp Vref The main assumption of this model is that particles are not
deformed or fractured, and elastic effects are negligible. It should be
Jennings et al. (1976) experimentally reported that melting of
mentioned that term a=U2c can be measured independently. It was
the target surface is a major mechanism of ductile material erosion.
also concluded that although dynamic hardness and ductility are
Based on this hypothesis, he employed dimensional analysis to
included in the above equation, more investigation is needed to
develop a mathematical model considering the factors that resist
obtain their values and incorporate them in Equation (15).
and drive erosion phenomena:
Sundrarajan and Shewmon (1983) proposed an analytical
5=2
KT G1=3 expression for the erosion rate based on a localization concept. The
εV ¼ 1=3
(11) localization concept suggests that there is strain at which plastic
R rt kTm DHm
deformation on the target surface localizes causing lip formation
Hutchings et al. (1976) performed a series of experiments using called critical strain which in turn leads to material removal from
steel spheres impacting mild steels and reported that their exper- the target surface or erosion. In the first step, they developed an
imental test results successfully simulated erosion of sand particles expression for critical impact number by conducting algebraic
on metals. They investigated the dependence of crater dimension to calculations. There are four undetermined parameters in their
impact velocity and angle. A high speed photography technique erosion equation including critical strain (εc), mean strain incre-
was also employed to study the energy balance during an impact. A ment (Dεm), deforming volume and an exponent. A constitutive
correlation to predict erosion rate based on experimental data was equation for flow stress as a function of critical strain was proposed.
proposed as They further simplified the constitutive equation using
858 M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873

experimental data and finally proposed an equation for critical the erosion rate. In this equation, maximum impact force is corre-
strain as a function of target specific heat and melting point. The lated to particle properties as
Tabor (1951) relation was used to obtain an equation for mean
strain increment. They also assumed that the plastic deforming 2=9 2=3 22=9 2=3 1=3 8=3
rt rp Et Vp dp rt rp Et Vp dp
volume is proportional to the crater volume and proposed the εm ¼ ð1  f ÞK1 5=9 4=3
þ fK2 1=3
(18)
following erosion equation, Ht Kc Ht Kc2
Sundararajan (1991) proposed a model to predict the erosion
0:085Vp2:5 r0:25
p r1b
t aðz þ 1Þ5a ½1  ðz þ 1Þ=ðz þ 2ÞFðzÞ rate for ductile materials for all impact angles and different particle
εm ¼ h ib
shape. The basic idea behind localization of plastic deformation is
6:06b ð1  CTc Þ1:25 nc Cp Tm
0:75 ð1  436=T Þ0:75 ðK H Þ1:25b
m 1 t
that there is a critical strain (εc) that causes lips to be formed on the
(16.a) surface rather than fracture. Therefore, erosion happens when this
critical strain equals the critical strain for material removal. He
suggested that for normal impact angle, the plastic work in the
0:25nc1a1 S ðz þ 1ÞS ðK1 Ht Þa1 S
a¼h i a1 S (16.b) plastic zone is equal to the incident energy absorbed in plastic
0:75 ð1  436=T Þ0:75
6:06rt Cp Tm deformation based on a balance of energy in the plastic zone. He
m
proposed the following equation for dimensionless erosion under
normal impact by implementing the constitutive equation for
b ¼ ð1 þ 5aÞa1 (16.c) plastic deformation:
They also proposed an alternative form of the erosion model by    
using the critical number of impacts equation suggested by 2nh C
εm ¼ FðtÞVp2 sin2 1  e2 (19.a)
Hutchings (1981): nh Cp

1j
3:6  103 Vp3 r0:5p rt aðz þ 1Þ
6a
5=8
εm ¼ h ij 1:9Hp
e¼ (19.b)
6:06j ð1  CTc Þ1:5 nc Cp Tm
0:75 ð1  436=T Þ0:75 ðK H Þ1:5b
m 1 t
1=2 1=8
Ee rp Vp
1=4

(16.d)
FðtÞz0:02  0:05 (19.c)
j ¼ ð2 þ 6aÞa1 (16.e)
He utilized the energy absorption relations developed by Brach
They compared their model with experimental data, Mamoun (1988) into his previous localized plastic deformation model in
(1975) and the Hutchings (1981) model (fatigue model) and re- order to predict erosion rate for oblique impact. Based on a balance
ported that erosion rate predictions by the localization model are of energy, it was suggested that the amount of dissipated energy in
more accurate as compared to fatigue models. A simplified form of the shear zone caused by incident kinetic energy of an impacting
their erosion model was also proposed as particle is equal to plastic work which is required to strain lip
volume from zero to the critical strain value (required to cause
6:5  103 Vp2:5 r2:5
p
erosion). He used the dissipated energy equation proposed by
εm z (16.f) Brach (1988) and obtained the following equation for erosion rate
0:75 H 0:25
Cp Tm t
under oblique impact conditions:
Venougopal Reddy and Sundararajan (1986) conducted experi-
( )  
ments on two ductile materials at a constant velocity of 40 m/s and Vp2 m m
three impact angles of 30, 60 and 90 . They reported that lip for- εm ¼ Cðnh þ 1Þ 2n   2 cos2 a
2 h nh Cp ð1 þ lÞ mc mc
mation and fracture are the main erosion mechanisms based on
scanning electron microscope (SEM) image analyses. It was shown (19.d)
that in contrast with previous studies, their results show maximum
Finally, it was suggested that the overall erosion rate is the
erosion at the normal impact angle. They also used a localization
summation of erosion rate at normal (Equation 19.a) and oblique
model to show that erosion rate is proportional to
(Equation 19.d) impact angles as:

L3 DUm  .
 
εVP f (17)  m
Uc εm ¼ 2nh CVp2 sin2 aFðtÞ nh Cp 1 þ ðnh þ 1Þ
mc
Johansson et al. (1987) developed a statistical model to predict      (19.e)
m
erosion rate of brittle single-crystal materials. He stated that the 2 = 4ð1 þ lÞtan2 aFðtÞ  e2
mc
material removal caused by impact of solid particles on the surface
of such materials is controlled by brittle fracture. Taking that into An empirical correlation to predict erosion rate for carbon steel
account, he recognized two different types of spalling mechanisms: with dry or wet surface was proposed at the University of Tulsa,
lateral spalling and median spalling. When a particle impacts a Erosion/Corrosion Research Center (E/CRC) in 1994 by Ahlert
surface, the spalled volume depends on spalled area and depth. (1994). The model is based on a series of direct impact experi-
Therefore, he assumed that the spalled areas and spalling depth is ments for different particle shapes and impact angles for calcu-
proportional to the plastic zone depth for lateral spalling and pro- lating the erosion ratio:
portional to median crack extension for median spalling. The
average spalled volume is equal to the summation of fraction of εm ¼ 2:17  107  ðBHÞ0:59 Fs Vp2:41 FðaÞ (20.a)
mass loss due to median spalling and lateral spalling. He developed
a correlation between median crack size and plastic zone depth According to Zhang et al. (2007) particle impact angle function is
with material properties and proposed the following equation for defined as:
M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873 859

X
5 properties of target material and particles. He proposed an erosion
FðaÞ ¼ Ai ai (20.b) parameter based on energy loss during erosion incorporating me-
i¼1 chanical properties of the alloys as well as the evolution of these
properties during deformations.
where the values of Ai and Fs and are suggested empirically.
A study of erosion in choke valves, which are widely used in oil 82 0 10 139
>
>
5
ð Þ ð Þ >
>
and gas industry, was performed by Haugen et al. (1995). They used >
>41  @ 3:06H 4
t A@
1m 2
t
þ Ep A5>
1m2
p
>
>
> 1 E >
>
a general correlation of erosion and empirically determined the 2
mp Vp < r V
p p
2 t
=
coefficients of the equation: εm fEParameter ¼
2 > > T$LV >
>
>
> >
>
>
> >
>
εw ¼ MKp FðaÞVp p
n
(21) : ;

They examined 28 different material types, and for each they (24)
varied impact speed and angle of attack and obtained empirical
Oka et al. (2005) also proposed an empirical correlation similar
coefficients. Using the coefficients obtained for steel, a numerical to the E/CRC erosion equation. They modified the basic equation of
simulation of the flow with Lagrangian particle tracking of sand
dependence of erosion to impact velocity at the normal angle to
particles was also performed. Good agreement between the pro- take into account the effect of target material hardness, particle
posed model and numerical simulation was reported in this study.
diameter and particle properties:
Chen et al. (1997) proposed a model for erosion at normal im-
pacts based on a residual tensile stress mechanism. They assumed  k  k 3
that residual tensile stresses cause material removal from the crater ε90 ¼ Kp ðHV Þk1 Vp 2 dp  106 (25.a)
region and developed an equation for the ratio of removed volume
to indentation volume. In the next step, based on the assumed where k1 and k3 are empirical exponent factors and k2 is a function
mechanism for erosion, they used the Johnson and Cook (1985) of material hardness and particle properties. Kp is an independent
fracture model to define the critical strain for erosion. They also factor denoting particle properties such as shape or hardness. They
used random walk theory along with an empirical correlation to generalized the above equation for all impact angles by introducing
estimate the mean number of impacts needed to cause material an impact angle function as
removal. Finally, inserting all the obtained equations into the
Hutchings (1981) erosion model, the following equation was pro- f ðaÞ ¼ ðsin aÞn1 ð1 þ HV ð1  sin aÞÞn2 (25.b)
posed to estimate erosion rate:
εV ðaÞ ¼ f ðaÞε90 (25.c)
rt r0:5
p Vp
3
εm ¼ 0:064 2 2 2 where n1 and n2 are determined by material hardness or other
p3=2 ½D1 þ D2 expðD3 s* Þ ½1 þ D4 ln_ε*  ½1 þ D5 T * 
impact conditions. It was stated that the first term on the right hand
(22) side of the impact angle function represents repeated plastic
Chen et al. (1998) suggested a computational model to predict deformation and the second term represents cutting action. Good
erosion rate using a general impact friction model. In fact, based on agreement between the proposed model with aluminum, copper,
previous experimental studies of friction coefficient, they proposed carbon steel and stainless steel erosion data was reported in their
a mathematical model to calculate friction coefficient called “gen- study.
eral impact friction model”. They modified the suggested governing In another study, Oka and Yoshida (2005) considered other
equations of oblique impacts based on Hutchings (1981) study and mechanical properties apart from hardness such as work hardening
combined them with the general impact friction model. The and load relaxation ratio and proposed a correlation to predict
modified equations were numerically solved to calculate velocity erosion rate at normal impact:
components and the critical impact angle. Having those parame-
ters, they proposed the following equations to calculate deforming, ε90 ¼ Kp ðk4 HV Þbk1 0  k4 HV  1; b  0 (26.a)
cutting and total erosion rates:
They further modified Equation (26.a) for different impact an-
P gles, speed, particle size and types of material as follow:
Ty mp Vpn DVpn
εD ¼ ¼ (23.a)
mp d mp d  k2  k3
Vp dp
P ε90 ¼ Kp ðk4 HV Þk1 b (26.b)
Tx mp Vpt DVpt v0 d0
εC ¼ ¼ (23.b)
mp c mp c
εV ðaÞ ¼ f ðaÞε90 (26.c)
Tt ¼ Tx þ Ty (23.c) where K, k1 and k3 are constants determined by the particle
properties and k2 is a function of material hardness and particle
εV ¼ εC þ εD (23.d)
properties. v0 and d0 are the standard impact velocity and particle
Levin et al. (1999) proposed a model for solid particle erosion of diameter which are used in their experiments.
ductile alloys. He stated that an accurate model to represent the A phenomenological model for erosion was proposed by Huang
mechanism of material removal, should take into account the effect et al. (2008). They pointed out that particle impact on a target
of mechanical properties of the target and erodent, as well as the produces normal and tangential directions which cause deforma-
work hardening effect during the erosion process. As a result, it was tion. They considered the normal component of the force being
stated that the portion of initial kinetic energy causing plastic responsible for “deformation damage removal” and tangential
deformation of the target material is a function of the ratio of component for “cutting removal”. Considering the normal force,
particle rebound velocity to particle impact velocity called coeffi- they found equations for maximum width and depth of indentation
cient of restitution. This coefficient is also a function of mechanical using the equation of motion in the normal direction. Having those
860 M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873

parameters, an equation to calculate elementary indentation vol- reduction of production rate is one of the management methods
ume was suggested. They also derived an equation to calculate the used to avoid erosion damage. In this case, the maximum produc-
average strain introduced into the surface by normal impact. On the tion velocity is restricted to a threshold value called erosional ve-
other hand, a cutting profile to calculate cutting volume caused by locity (the velocity above which excessive erosion may occur). As a
the tangential force as a function of maximum cutting width, depth result, sand production rate and impact velocities decrease likewise
and length was suggested. They used the Coffin (1954) and Manson the erosion rate. Obviously, decreasing the production rate involves
(1953) equation along with a critical strain equation to calculate the financial consequences.
deformation damage removal: Oil and gas companies seek to use simple methods for predict-
ing erosion rate. A literature survey reveals that there are suggested
  empirical and mechanistic models for erosion prediction. The
1   2þ2b0
1

mp r4b0
p V p sin a following sub-sections explain the available empirical and mech-
εVD ¼ C1   (27.a) anistic models in the literature.
1
1þ4b0
1=b0
εc pn 3.2. Empirical erosion prediction in pipelines
For real cutting removal by a particle, they assumed it is pro-
portional to the cutting volume but inversely proportional to the One of the earliest empirical equations for estimating the
material ductility: erosional velocity was suggested by American Petroleum Institute
Recommended Practice 14E (API RP 14E, 1981):
   
sÞ sÞ
1þ3ð1n 2þ3ð1n 3ð1ns Þ C1
Ve ¼ pffiffiffiffiffiffi
4 2
C2 mp Vp ðcos aÞ2 ðsin aÞ 2
rm
(29)
εVC ¼ ð1ns Þ 3ð1ns Þ
(27.b)
dp 4
εi0 pt pn 4

where Ve is the erosional velocity in ft/s, C1 is an empirical constant,


They finally suggested the following equation to predict total and rm is the fluid mixture density in lb/ft3. For clean service
volume loss by summing the deformation damage removal and (solids-free and non-corrosive), API suggested using an empirical
cutting removal as constant value of 100 and 125 for continuous and intermittent
services, respectively. When solids and/or corrosive conditions are
εVT ¼ εVC þ εVD (27.c) present, API RP 14E recommends the reduction of C factor, but it
does not provide any guidelines how to apply that.
It was emphasized that the proposed erosion takes into account
Interestingly, there is no clear evidence of the basis of the API RP
the effect of particle mass, size, shape and speed as well as impact
14E equation. The basis of the equation was attributed to different
angle and target material properties. A simplified version of
reasons other than sand erosion. Consequently, various values for
Equation (27.b) was also proposed for situations where solid
the C factor were suggested by different investigators, and they
transport impact angle is small and cutting wear is the dominant
showed that the equation is too conservative for clean service
erosion mechanism as
(Salama and Venkatesh, 1983; Heidersbach, 1985; Gipson, 1989;
εVT Deffenbaugh and Buckingham, 1989; Smart, 1990, 1991; Coker,
εV ¼ zC2 r0:1875
p d0:5
p Vp
2:375
ðcos aÞ2 ðsin aÞ0:375 (27.d) 1990; Patton, 1993).
mp
When it comes to non-clean service, there are several limita-
Nsoesie et al. (2014) using experimental data modified the tions associated with using the API RP 14E equation. Many impor-
previous model of Sheldon and Kanhere (1972) which is based on tant factors such as solid particle size and shape, sand production
indentation hardness theory. They experimentally investigated the rate, multi-phase characteristics are not considered in the equation.
erosion rate for five Stellite alloys at two different impingement Furthermore, the equation predicts higher values of erosional ve-
angles and two velocities. Their experimental measurements locity as fluid mixture density decreases which is not physical. Drag
showed a higher rate of erosion for Stellite alloys in comparison to force exerted on a particle decreases by reducing the fluid density.
Sheldon and Kanhere (1972) model predictions. It was reported This causes the particle to impact at a higher velocity thus more
that since the Sheldon and Kanhere (1972) model was developed erosion. Several investigators concluded that the API RP 14E
based on experimental data of aluminum and aluminum is more equation is not valid for non-clean services such as liquid droplet
ductile than Stellite alloys, it under predicts critical velocity for impact, and efforts were devoted to develop an alternative
Stellite alloys. Based on these observations, they introduced an approach for erosion prediction.
order correction factor in the model. They also further modified the Salama and Venkatesh (1983) developed the following model
model to take into account the effect of impingement angle. The for elbows using the experimental data of Rabinowicz (1979) for
impingement angle correction factor includes two empirical con- erosion ratio of ductile metals:
stants called shifting coefficient (A) and shifting exponent (B)
W_ p Vf2
which are target material dependent. The final form of their ER ¼ 1:86  105 (30)
equation is P D2

   _ p is the sand
B 3 where ER is the erosion rate in mils per year (mpy), W
C1 d3p Vp A sinða=2Þ1=3
3=2
rp flow rate in bbl/month, Vf is the fluid flow velocity in ft/s, D is the
εV ¼ (28) pipe diameter in inches, and P is the material hardness in psi.
3=2
HV Salama and Venkatesh (1983) also showed that the erosion rate in
tees is about half of the erosion rate in elbows. Equation (30) can be
While some of these models have been validated with solid also written in the following form:
particle erosion data, many of these models did not consider sand
erosion. It is noted from the above, that none of these models W_ pV 2
f
address effects of produced fluids. The presence of sand particles in ER ¼ Sk (31)
D2
production fluids is the primary cause of erosion damage and the
M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873 861

Imposing in lb/day, Sk QUOTE acquires values of 0.038 and 0.019 where rc is the bend radius of curvature. The unit of the parameters
for short radius elbows and tees, respectively. Since this model was is similar to the units in Bourgoyne's (1989) equation. C1 is equal
developed based on erosion data in air-sand flow, it predicts to 4.9619 and 5.4355 for cast and seamless materials,
erosion rate more accurately for gas flows (or systems). respectively.
Bourgoyne (1989) measured erosion rate in gasesolid, liquid- Salama (1998, 2000) incorporated the particle diameter and the
esolid and mistesolid flows in a diverter. He suggested an equation fluid mixture density in the equation of Salama and Venkatesh
for the prediction of wall thickness loss in gas continuous flows (dry (1983) to account for multi-phase flow, as follows:
gas flow or mist flow):
1 W_ p V 2 dp
m
ER ¼ (36)
rp W_p  V 2
Sm D2 rm
SG
ER ¼ Fe (32)
rt Apipe 100ag where ER, W _ p and D are in mm/year, kg/day and mm, respec-
tively. dp is the particle diameter in microns, Vm is the mixture
where ER is the erosion rate in m/s, Fe is the specific erosion factor, velocity in m/s, and rm is the fluid mixture density in kg/m3. Sm is a
rP and rt are the densities of particle and wall in kg/m3, respectively, geometry dependent constant. Calibrated against the data of
W_ p is the sand flow rate in m3/s, Apipe is the cross-sectional area in
Weiner and Tolle (1976), Tolle and Greenwood (1977), Bourgoyne
m2, VSG is the superficial gas velocity in m/s, and ag is the gas vol- (1989), Kvernvold and Sandberg (1993), and Birchenough et al.
ume fraction. The specific erosion factor depends on geometry type, (1995), Sm acquires a value of 5.5 for elbows, 33 for seamless and
wall material, and flow conditions (dry gas flow or mist flow). cast ells, 68 for gaseliquid flow in plugged tees, and 1379 for gas
Bourgoyne (1989) also suggested another equation for liquid flow in plugged tees. The value of Sm for the latter case seems to be
continuous flow: very large, and the author stated that further validation is needed
for this case (gas flows in plugged tees). Since the test results did
rp W_p  V 2
not show a major difference between erosion in bends with rc of 1.5
SL
ER ¼ Fe (33)
rt Apipe 100HL and 5, Salama (2000) did not consider the effect of elbow radius of
curvature in the equation.
where VSL is the superficial liquid velocity, and HL is the liquid hold- Empirical equations are easy to use, but there are always un-
up. Bourgoyne's (1989) equations were developed based on the certainties for extrapolating them, since erosion mechanisms may
experimental data which were obtained at high flow rates. His change in different conditions and do not follow the original trend.
experiments were also obtained for a high sand flow rates that may Some researchers developed mechanistic models for predicting
be produced in diverter systems and their applications to oil and erosion that are more comprehensive. These models are developed
gas production under low concentration is questionable. based on physics and mechanisms of erosion and include different
Svedeman and Arnold (1993) studied the applicability of the parameters that can affect the erosion phenomenon. The following
Bourgoyne (1989) equations at lower flow velocities. It was found sub-sections discuss erosion mechanistic modeling.
that this model over predicts the available experimental data with
an average value of 25%.
3.3. Mechanistic models for predicting erosion in single-phase flows
Different wear mechanisms have various controlling parameters.
Considering this fact, Svedeman and Arnold (1993) recommended
In a direct impingement geometry, before impacting the ge-
that the criteria for sizing multi-phase flow lines should be divided
ometry wall, particles enter into a region in the vicinity of the wall
into four groups namely: clean service, erosive service, corrosive
in which particles are decelerated by the fluid. This region is called
service, and erosiveecorrosive service. They rearranged
the “stagnation region” (Fig. 8). To account for this particle decel-
Bourgoyne's (1989) equation and suggested an erosional velocity
eration, Shirazi et al. (1995) postulated a mechanistic procedure for
equation for erosive service based on tolerable erosion rate of 5 mpy:
erosion prediction in single-phase flow in elbows and tees. Their
procedure assumes that in order for particles to imping the target
D
Ve ¼ Ks qffiffiffiffiffiffiffiffi (34) wall particles must penetrate through a region similar to the
W _p stagnation region of a direct impingement geometry (Fig. 9). To
implement the effect of geometry size and type on erosion, they
where Ve is the erosional velocity in ft/s, D is the pipe diameter in introduced the “equivalent stagnation length” concept. Since this
inch, and W _ p is the sand flow rate in ft3/day. Ks is 1.34 and 7.04 for model was the foundation of many mechanistic models developed
long radius elbows and plugged tees, respectively. Note that the subsequently, it is explained in detail.
erosional velocity is the mixture velocity (summation of superficial In the model proposed by Shirazi et al. (1995), the impact ve-
gas and liquid velocities). locity of a representative particle moving in the stagnation region
Svedeman and Arnold (1994) postulated different values for Sk of length L is calculated. The particle impact velocity is then used to
in the equation proposed by Salama and Venkatesh (1983). The estimate erosion rate. It is worth mentioning that the particle
values of Sk were 0.017 and 6  104 for long radius elbows and impact velocity is generally not the same as the fluid bulk velocity.
plugged tees, respectively. As noted earlier, the particle impact velocity depends on factors
Jordan (1998) considered that the rate of material volume loss such as fluid density and viscosity, and particle density.
should vary with the square of the particle velocity and linearly To calculate the particle impact velocity, the fluid velocity
with the volumetric rate at which particles imping on target wall. profile should be known in the stagnation region. It is assumed
Jordan (1998) employed Bourgoyne's (1989) data and proposed the that the fluid velocity (Vf) decreases linearly from the flow char-
following equation: acteristic velocity (Vchar) at x ¼ 0 to zero at the target wall (x ¼ L) as
follows:
 2 !1:8885
2  
C1 2:349 _ 0:9535 1 x
ER ¼ 10 VSG Wp 1 1þ (35) Vf ¼ Vchar 1  (37)
2rc Lstag
862 M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873

where ER is the ratio of the mass of removed target material (due to


the particle impact) to the mass of particle. The erosion equation
was developed based on erosion data for low-carbon steels.
In Equation (37), Lstag is the stagnation region length which can
be obtained using a semi-empirical equation for elbows:

Lstag  
¼ 1  1:27tan1 1:01D1:89 þ D0:129 (42)
Lref

where Lref is 1.18 inches. For tees, Lstag can be attained from the
following equation:

Lstag  
¼ 1:35  1:32tan1 1:63D2:96 þ D0:247 (43)
Lref

where Lref is 1.06 inches. In these equations, D is the pipe diameter in


inch. These semi-empirical formulas were developed from experi-
mental data and numerical simulations. For smaller pipe diameters,
Fig. 8. Schematic of stagnation region in direct impingement geometry Zhang et al. the stagnation length is a strong function of pipe diameter. For large
(2010). pipe sizes (larger than about 6 inches in diameter), however, the
stagnation length dependency on the pipe diameter is weak. This
Particle impact velocity (VL) is evaluated employing a simple 1-D point aside, as pipe size increases, stagnation length increases and as
particle tracking approach along the stagnation length. In this a result particles must penetrate through a region with more fluid (in
approach, the particle equation of motion is numerically solved. comparison to smaller pipe) to impact the wall.
Drag force is the only force considered in the equation as shown in The Shirazi et al. (1995) model can be summarized in four steps:
the following equation,
1 Depending on the geometry type, the stagnation length is ob-
   p d2 tained from either Equation (42) or Equation (43). This length is
dVp   p
mp Vp ¼ 0:5rf Vf  Vp Vf  Vp CD (38) needed for the calculation of fluid velocity profile (Equation
dX 4
(37)).
where mp is the mass of particle, Vp is the particle velocity, Vf is the 2 The flow velocity profile is determined from Equation (37). In
fluid velocity at the particle location, rf is the fluid density, and CD is the equation, the flow characteristic velocity (Vchar) is set equal
the drag coefficient. The term on the right hand side of the equation to the fluid average velocity.
is the drag force. The drag coefficient applied was a simplified 3 Particle impact velocity is computed by solving the particle
relation: equation of motion (Equation (38)). Fluid velocity at the particle
location is obtained from the previous step. Note that the par-
24 ticle equation of motion needs the particle velocity (Vp) at X ¼ 0
CD ¼ þ 0:5 (39) as an initial boundary condition. It is assumed that the initial
Rep
particle velocity is equal to the flow characteristic velocity
where (Vchar). When the particle distance to the wall is equal to the
radius of the particle, the particle tracking is stopped and the
  particle velocity at this location is considered as the particle
 
rf Vf  Vp dp impact velocity.
Rep ¼ (40)
mf 4 Finally, the particle impact velocity is substituted into Equation
(41) to obtain the erosion ratio.
mf is the fluid viscosity and Rep is the so-called particle Reynolds
number. The calculated particle impact velocity (VL) is then Shirazi et al. (1995) compared their model predictions with the
substituted into an empirical erosion equation as follows: experimental data of Bourgoyne (1989) and Clark (1991). The
model prediction showed good agreement with the experimental
ER ¼ 1:73  106 VL1:623 (41) data.
The advantage of this model is that it accounts for many key
flow, fluid and particle parameters influencing erosion rate such as
flow velocity, fluid density and viscosity, particle shape, particle
size and density, and geometry size and shape. The model accounts
for the mentioned parameters through the particle equation of
motion and changing the stagnation length.
As noted earlier, in low-density fluid flow containing heavy
particles such as gas-particle flow, the efficiency of momentum
exchange between the carrier phase (fluid) and the dispersed phase
(particles) is very low. Hence, the carrier phase can not alter the
heavy particles path and consequently, heavy particles can cross
the flow streamlines and impinge the wall. In this instance, the
particles impinge the target wall at an impact angle of 90 and with
an impact velocity similar to the bulk velocity of the carrier phase.
Fig. 9. Schematic of stagnation region in tee and elbow. This erosion mechanism is called direct impingement mechanism.
M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873 863

The procedure developed by Shirazi et al. (1995) suffers from for elbow radius, W _ p is the sand production rate in kg/s, VL is the
two main limitations. Firstly, the procedure assumes that the par- characteristic particle impact velocity in m/s, D is the pipe diameter
ticle trajectory is a straight line (direct impingement mechanism), in inches, and Dref is the reference pipe diameter equal to 1 inch.
which is not the case for light particles in large-density fluid flow The equation was developed based upon experimental data
such as liquid-particle flow. For this case, the particles tend to collected at E/CRC and by Weiner and Tolle (1976), and Tolle and
follow flow streamlines. Secondly, the procedure does not incor- Greenwood (1977). In the above equation, Fr/D can be obtained
porate the effect of turbulence on the particles trajectory. from the following equation developed by Wang et al. (1996):
Zhang et al. (2010), found that for some cases, the tangential ! !
particle impact velocity component and the turbulent fluctuation r0:4
f
m0:65
f  
velocity component are important in erosion calculations. These Fr=D ¼ exp  0:1 þ 0:015r0:25
f þ 0:12 rc  1:5
d0:3
p
are important factors for erosion prediction in addition to the
normal particle impact velocity component. They suggested a 2-D (45)
procedure to address the limitations of the 1-D model of Shirazi
et al. (1995). where all parameters have SI units. Equation (45) was validated
In the first step of the Zhang et al. (2010) procedure, the 2-D flow against the experimental data of Bikbiaev et al. (1972), Tolle and
field information is achieved by interpolating between some pre- Greenwood (1977), Eyler (1987), and Bourgoyne (1989).
saved CFD simulation data. The CFD simulations are conducted Similar to the model proposed by Shirazi et al. (1995), the par-
beforehand for a range of Reynolds numbers. The CFD pre-saved ticle impact velocity (VL) is obtained by solving the particle equa-
data include fluid velocity components, pressure, turbulent ki- tion of motion (Equation (37)). However, since the particle moves in
netic energy and its dissipation rate, and Reynolds stresses. The multi-phase flow, the fluid mixture properties are used. Mixture
interpolation is carried out based on the Reynolds number of the properties such as mixture viscosity and density can be obtained
case of interest. from the following equations:
In the second step, the interpolated information supplied in the
VSG VSL
first step is used to compute the particle impact information such as mm ¼ m þ m (46)
impact location, speed, and angle. In the 2-D procedure, many VSL þ VSG G VSL þ VSG L
particles are tracked unlike the 1-D model in which only one
representative particle is tracked. Since the turbulent fluctuation VSG VSL
rm ¼ r þ r (47)
velocity components are temporally and spatially random, tracking VSL þ VSG G VSL þ VSG L
many particles provides statistically independent results from the
Furthermore, the following ad-hoc equations are used to
number of the tracked particles. The “eddy interaction model” of
calculate the flow characteristic velocity (Vchar):
Gosman and Ioannides (1983) is used to simulate the values of
instantaneous fluctuating velocities at the particle location during
Vchar ¼ lnL VSL þ ð1  lL Þn VSG (48.a)
the particle tracking. This model has been used and modified by
many researchers to account for different phenomena such as  0:11
turbulence anisotropy (Parsi et al., 2010a) and particle inertia effect VSL
lL ¼ (48.b)
(Parsi et al., 2010b). VSL þ VSG
In the third step, the E/CRC erosion equation (Equation (20))
uses the particle impact information to calculate the erosion rate
 
V
caused by the corresponding particle impacts. n ¼ 1  exp  0:25 SG (48.c)
VSL
It is worth mentioning that in the single-phase flow model, the
3.4. Mechanistic models for predicting erosion in multi-phase flows flow characteristic velocity (Vchar) is set equal to the average flow
velocity; whereas in the multi-phase flow model, it is obtained
Erosion modeling in multi-phase flow is more difficult than from Equation (48.a).
erosion modeling in single-phase flow because of the complexity of This mechanistic model for predicting erosion in multi-phase
flow modeling and particle tracking. Jordan (1998) developed a flow can be outlined as follows:
model for erosion prediction in multi-phase flow based on the
single-phase erosion model developed by Shirazi et al. (1995). 1 The stagnation length is determined from either Equation (42)
Jordan (1998) split the multi-phase flow into its components based or Equation (43).
on an “effective diameter” for each phase and separately conducted 2 Flow velocity profile is determined from Eqns. (37) and (48), and
the erosion rate calculation for each phase. The total erosion is the the stagnation length calculated in step 1.
summation of erosion rates in all phases. 3 Equations (37)e40, 46 and 47 are used to calculate the particle
McLaury et al. (1999) and McLaury and Shirazi (2000) also impact velocity. The particle velocity (Vp) at x ¼ 0 is set equal to
proposed a method for erosion prediction in multi-phase flow Vf at x ¼ 0.
based on the single-phase erosion model developed by Shirazi et al. 4 Particle impact velocity is substituted into Equation (44) to
(1995). In this model, the maximum erosion rate of steel materials obtain the erosion ratio.
in multi-phase flows is calculated using the following equation.

_ p V 1:73 Det Norske Veritas (DNV) developed a comprehensive guideline


W
ER ¼ FM FS FP Fr=D  L2 (44) on erosion assessment based on CFD results and experimental data.
D DNV RP 0501 (1996) provides procedure for calculation of erosion
Dref in geometries such as straight pipes, elbows, blind tees, welded
joints, and reducers.
where ER is the maximum erosion rate in m/s, FM is an empirical The empirical and mechanistic models for erosion prediction in
constant accounting for material properties, FS is the particle shape multi-phase flow presented to this point relate the erosion rate to
factor, FP is a penetration factor in m/kg, Fr/D is the factor accounting the superficial velocities of phases and the fluid mixture properties.
864 M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873

These models do not consider the multi-phase flow pattern in the


erosion calculations. Vieira (2014) and Parsi (2014) showed that for
the same flow velocities, different multi-phase flow patterns can
lead to dissimilar erosion rates which are different by a factor of 10.
Thus, a robust erosion prediction mechanistic model should
incorporate the effect of flow pattern on the erosion rate.
Here, erosion prediction models are categorized based on the
flow pattern for which they were developed. It should be noted that
except the Chen et al. (2006a) model, all of these models were
developed based on the McLaury and Shirazi (2000) mechanistic
model. The Chen et al. (2006a) procedure assumes multi-phase
flow as a homogenous single-phase flow based on mechanistic
analysis. Then, in the procedure, a CFD-based erosion modeling is
performed for the representative single-phase flow.
The following models are different in the way that they account
for the flow characteristic velocity (Vchar) and consequently the
particle initial velocity (Vp at x ¼ 0). The particle initial velocity
plays an important role in the erosion calculation, due to its strong
influence on particle impact velocity. Thus, it is crucial to predict
the representative particle initial velocity in multi-phase flow
accurately. These models estimate the flow characteristic velocity Fig. 10. Main flow patterns in horizontal pipe.
based on the physics of multi-phase flow. Before discussing the
mechanistic models, which were developed based on multi-phase
flow pattern, some of the characteristics of different flow regimes
Different mechanistic models for predicting erosion based on
are briefly introduced.
the flow pattern of interest are discussed below.
In multiphase flow, the interfacial forces between the phases
distribute them into different flow patterns. The change of phase
properties and superficial velocities results in the change of the 3.4.1. Bubble flow
flow pattern. Flow orientation and pipe inclination angle are other Mazumder et al. (2005) suggested using the flow mixture ve-
factors that influence the flow pattern. The main flow patterns locity as the flow characteristic velocity (V0) in the McLaury and
observed in horizontal multi-phase flow are dispersed bubble, Shirazi (2000) model (see Equation (37)):
stratified, slug and annular flows (Fig. 10). In vertical flow, the main
Vchar ¼ Vm ¼ VSL þ VSG (49)
flow patterns observed are bubble, slug, churn and annular flows
(Fig. 11). Assuming bubbly flow as a homogenous flow, Chen et al.
Bubble flow is defined as a flow of the continuous liquid phase in (2006a) simplified bubbly flow to a single-phase flow with
which small bubbles are approximately uniformly distributed. It mixture properties. Then, they carried out CFD erosion simulations
exists in relatively large pipe diameters and takes place at low gas for the representative single-phase flow. They obtained the mixture
and high liquid flow rates. viscosity from Equation (46), and the mixture density from the
Annular flow occurs at low liquid and very high gas flow rates. following equation instead of Equation (47):
This flow is characterized by a gas core with entrained droplets and
a continuous thin liquid film around the pipe wall. Details about the rm ¼ ð1  HL ÞrG þ HL rL (50)
effect of liquid viscosity and gas and liquid flow rates on horizontal
and vertical annular flow can be found in the works of Vieira et al. where HL is the in-situ liquid holdup. They reported very good
(2014a) and Vieira et al. (2014b), respectively. In particulate annular agreement between the procedure prediction and the experi-
flow, particles can be present both in the liquid film and the gas mental data of Salama (2000) for erosion in bubbly flow.
core.
Slug flow can be considered as an alternating flow of gas pockets
(Taylor bubble) and liquid slugs. There is a downward flow of the
liquid film around the Taylor bubble in vertical upward slug flow.
Slug flow occurs over a wide range of gas and liquid flow rates. At
high gas flow rates, the Taylor bubble penetrates through the liquid
slug and forms an aerated slug which is called “pseudo slug”. Un-
fortunately, there is limited published material on slug flow at
those high flow rates. Discussion about the characteristics of
pseudo slug flow and the effect of different parameters on it can be
found in Kokal and Stanislav (1989) and Kesana et al. (2013a,b).
Churn flow occurs in vertical and near-vertical flow and is
bounded by the slug and annular flow regimes. In churn flow, there
are no clear boundaries between the phases. There are liquid lumps
that move up and down. Due to its complexity, there is limited
published material on churn flow most of which study the churn
flow regime at low flow rates. Discussions about periodic structures
in churn flow (Sekoguchi and Mori, 1997) and the effect of viscosity
on its structure at high flow rates (Parsi et al., 2014) can be found in
the literature. Fig. 11. Main flow patterns in vertical pipe.
M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873 865

3.4.2. Annular flow mass of sand in the liquid film


1E ¼ (60)
Mazumder et al. (2005) presented a mechanistic model that total mass of sand
considers erosion caused by particles in the liquid film as well as by
The total erosion rate is the summation of erosion rate caused by
particles that are in the gas core. Two sets of particle tracking cal-
the particles in both regions. More details about this model can be
culations are carried out in this model: for a representative single
found in Mazumder et al. (2003); Mazumder (2004, 2007). Slightly
particle in the liquid film and also for a representative single par-
modified versions of this model can be found in McLaury et al.
ticle in the gas core. The characteristic flow velocity for the particle
(2006, 2010, 2011).
in the liquid film (Vchar-film) is different from the characteristic flow
Before the development of the discussed model, Mazumder
velocity for the particle in the gas core(Vchar-gas core). These char-
et al. (2004) developed a preliminary mechanistic model in
acteristic velocities are calculated as follow:
which the characteristic flow velocity was determined as follows:
Vcharfilm ¼ Vfilm (51) Vchar ¼ ð1  EÞVfilm þ EVDroplet (61)

Vchargas core ¼ VDroplet (52) The Chen et al. (2006a) procedure assumes that erosion is
chiefly caused by the particles in the gas core and erosion caused by
particles in the liquid film is negligible. Furthermore, the gas/liquid
where Vfilm and VDroplet are the liquid film and droplets velocity,
droplets flow can be treated as a homogenous flow. Hence, the gas
respectively. The liquid film velocity is calculated from the
core region can be considered as a single-phase flow with mixture
following equation:
properties. Similar to bubbly flow, Eqns. (46), (47) and (50) are
  applicable for the gas core. However, in these equations, VSL should
1  E D2
Vfilm ¼ VSL   (53) be substituted with the velocity of liquid droplets in the annular gas
4dfilm D  dfilm core. They used the Ansari et al. (1994) mechanistic model to
evaluate the velocity of liquid droplets.
where dfilm is the liquid film thickness and can be determined using Chen et al. (2006a) found that the thin liquid film can protect the
the Paz and Shoham (1994) approach. E is the liquid droplet wall from the impacting particles. Consequently, as a particle rea-
entrainment fraction and can be obtained from the Ishi and ches the gaseliquid film interface, the particle tracking should be
Mishima (1989) correlation: continued performed across the liquid film.
 
E ¼ tanh 7:25  107 We1:25 Re0:25
L (54) 3.4.3. Slug flow
The Mazumder et al. (2004, 2005) model assumes that particles
where We and ReL are the modified Weber and liquid Reynolds within the liquid slug cause erosion in slug flow, and erosion due to
numbers, respectively. They can be obtained from the following the particles in the liquid film is negligible. Furthermore, it is
equations: assumed that particles are uniformly distributed in the liquid
phase. The characteristic flow velocity is calculated as:
2 D 
rG VSG rL  rG 1=3
We ¼ (55) Vchar ¼ HLLS  VLLS (62)
s rG
rL VSL D where HLLS and VLLS are the liquid holdup in the slug body and the
ReL ¼ (56) liquid velocity of the slug body. HLLS is calculated from the Gomez
mL
et al. (2000) correlation:
where s is the surface tension in N/m.  
The droplet velocity in the gas core is less than the gas core  0:45qþ2:48106
r L Vm D
mL
velocity, due to the slippage between the phases. According to Fore HLLS ¼ e (63)
and Dukler (1994), the slippage ratio is 0.8. As a result, droplet
velocities in the gas core can be determined as follows:
where q is the inclination angle in radian.
VDroplet ¼ 0:8VG_Core (57) VLLS is determined from Taitel and Barnea (1990) model as
follows:
where VG_Core is the gas core velocity and can be determined from
Vm  VGLS ð1  HLLS Þ
the following equation: VLLS ¼ (64)
!2 HLLS
D
VG_Core ¼ VSG (58)
D  2dfilm where VGLS is the velocity of gas in the slug body.
The Chen et al. (2006a) procedure simplifies slug flow to a
The erosion rate due to the particles in each region should be representative single-phase flow. The single-phase flow has the
weighted by the sand mass fraction of each region. Santos (2002) mixture properties of either the slug unit or the slug body. When
obtained sand distributions in horizontal and vertical annular slug unit is used as the representative flow, the mixture properties
flow shows that showed the liquid droplet entrainment fraction are obtained from Eqns. (46) and (50). For the other case, the
and the mass ratio of particles in the gas core are nearly identical. mixture properties are determined as follows:
Accordingly, the fraction of sand entrained in the gas core can be
written as: rm ¼ ð1  HLLS ÞrG þ HLLS rL (65)

mass of sand in the gas core mm ¼ ð1  HLLS ÞmG þ HLLS mL (66)


E¼ (59)
total mass of sand
Mixture velocity (Equation (49)) is used as the velocity of the
and the sand fraction in the liquid film can be determined as: representative flow in both cases. Employing the mixture
866 M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873

properties for the representative single-phase flow, CFD-based the liquid phase. Mazumder et al. (2005) proposed that the char-
erosion simulation is used to predict the erosion rate. The pre- acteristic velocity for churn flow is equal to the mixture velocity
dicted erosion rate, then, is weighted by the sand mass ratio in the (Equation (63)). Although there are many data on erosion in slug
liquid slug body that can be calculated from following equation: and annular flows, there is a lack of studies on erosion in churn
flow. Therefore, the hypothesis of Mazumder et al. (2005) requires
LS HLLS Vm further validation.
m_ ¼ (67)
LS HLLS Vm þ LF HLF Vfilm
3.5. CFD-based erosion modeling
where LS and LF are the slug body length and liquid film length,
respectively. Theses parameters are obtained from Zhang et al. As mentioned earlier, there are many factors that influence
(2003a,b). erosion such as material properties, sand size, geometry, flow ve-
A part of the complexity of erosion in slug flow is due to the locity, operating pressure, turbulence, liquid hold-up and particle-
transient behavior of this flow regime. When a slug body passes multiphase flow interactions. These factors interact with each
through a bend, the front of the slug body has a gas region in its other, and thus the experimental study of their impact on erosion
path before hitting the bend wall. The back of slug body, however, requires a great deal of endurance. CFD is a powerful tool that can
should push the front of slug body which primarily has liquid. As a be used to study the effect of different parameters on erosion rate,
result, the impact velocity of particles at the front of the liquid slug predict the maximum erosion rate, and find areas likely to be
body is different from the impact velocity of particles at the back of susceptible to erosion, even in complex geometries in which setting
the liquid slug body. McLaury et al. (2010) concluded that using a up an experimental study is difficult.
single value for a representative particle impact is not suitable and Many researchers used CFD to simulate erosion in bends (Wang
instead, a range of impact velocities should be used for erosion et al., 1996; Edwards et al., 1998, 2001; Wang and Shirazi, 2003; Fan
calculations. This range of impact velocities depends on the initial et al., 2001, 2002, 2004; Chen et al., 2004, 2006b; Zhang et al., 2007;
location of particles in the slug body. Njobuenwu et al., 2012; Njobuenwu, and Fairweather, 2012; Pereira
The McLaury et al. (2010) model divides the liquid slug body to et al., 2014; Lu and Agrawal, 2013), sudden contraction/expansion
two segments: slug front (front of liquid slug body) and slug tail (McLaury, 1993, 1996; Habib et al., 2004, 2007, 2008; Li et al., 2009;
(remaining portion of liquid slug body). The slug front has a length Najafifard, 2014), Valve (Mazur et al., 2004), and complex geome-
equal to the pipe diameter and is divided into 500 sections. For each tries (Atkinson et al., 2007; Ferng, 2008; Gandhi et al., 2012; Song
section, a representative particle impact velocity and the resulting et al., 2013).
erosion are determined. The 500 resulting erosion rates are aver- CFD-based erosion modeling consists of three steps: flow
aged to obtain the erosion rate for the slug front. For the slug tail, modeling, particle tracking, and relating particle impact informa-
one representative particle impact velocity and consequently one tion to erosion damage. Each step is dependent on the previous one.
erosion rate is calculated. Therefore, any non-physical result in any of the three steps affects
Zhang et al. (2011) simplified slug flow to a representative the final erosion results.
single-phase flow, and for determining the erosion rate they used Flow modeling is the start for the CFD-based erosion prediction.
the 2-D approach of Zhang et al. (2010) for erosion calculation in Usually industrial CFD software (CFX, FLUENT, STAR-CCMþ) are used
single-phase flows. This method is a combination of CFD and to simulate the flow field. To obtain flow field data (pressure, ve-
mechanistic modeling. locity components, turbulent kinetic energy), NaviereStokes
A mechanistic model for predicting erosion in pseudo slug flow in equations need to be solved. Since calculating the time resolved
horizontal pipes was developed by Kesana et al. (2014a). They found transport quantities (momentum, energy) is computationally
that in slug flow at high superficial gas velocities, liquid droplets and expensive, usually for engineering applications the time-averaged
consequently particles are entrained in the gas region. The previous forms of the NaviereStokes equations are used to obtain those
slug flow erosion models, however, assume that erosion is caused quantities. However, time-averaging filters small scale turbulent
only by the particles which are present in the slug body. Their model fluctuations. As a result, this process introduces unknown variables
accounts for erosion due to the particles entrained in the gas region. called Reynolds stresses and consequently closure equations are
Moreover, performing sand sampling experiments, Kesana et al. needed. There are different turbulence models (k  ε, k  u, and
(2014b) showed that increasing superficial gas velocity increases RSM) offering closure equations. Each of these models has advan-
the mass fraction of particles impacting the pipe wall. This obser- tages and disadvantages. Since they provide flow field predictions
vation is in contrast with Equation (67). They used an approach at various conditions and geometries with different accuracy, it is
based on pressure drop to incorporate the effect of superficial gas important to choose an appropriate turbulence model. Choosing a
velocity on sand mass fraction in the slug body. The sand mass ratio suitable turbulence model requires an understanding of flow
in the slug body is shown in Equation (68): characteristics in the geometry of interest. Choosing the turbulence
model is not the only preparation for CFD modeling. A CFD mesh
DPSB sensitivity study also needs to be performed to assure that the flow
m_ ¼ (68)
DPSU field prediction is independent of the mesh. Considering that the
results from flow modeling are used to predict erosion, it is
where DPSB is the pressure drop in the slug body, and DPSU is the essential to accurately model the flow field. Therefore, it is impor-
pressure drop in the entire slug unit. They compared the perfor- tant to validate flow field prediction before predicting particle
mance of their model with the experimental data of Gundameedi trajectories (Timothy, 1999; Timothy and McLaury, 2002).
(2008), Rodriguez (2008), Throneberry (2010), Kesana et al. The second step in predicting erosion rate using CFD is deter-
(2013c), and Kesana et al. (2013d) and reported good agreement mining particle motion. A Lagrangian or Eulerian approach can be
between predicted and measured erosion data. used to simulate particle motion. In the case of flow with high
particle concentration, the Lagrangian approach becomes
3.4.4. Churn flow computationally expensive. Therefore, the Eulerian approach can
In churn flow, similar to the erosion models for other flow be the alternative way to calculate erosion. For the Eulerian
patterns, it is assumed that particles are uniformly distributed in approach, particles are considered as a continuous phase.
M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873 867

However, using Eulerian method may be problematic in predicting fluctuations on particles (Hinze, 1972, Phillips, 1980; Milojevic,
particle behavior close to the wall. Close to the wall, particle 1990; Lu et al., 1992; Mehrotra et al., 1998; Klose et al., 2001;
motion consists of impacting and rebounding, and the Eulerian Marchiolo and Soldati, 2002). One of the most famous methods is
method only gives mean value of particle motions in each control the particleeeddy interaction model of Gosman and Ioannides
volume. This can cause inaccurate prediction of particle impact (1983).
velocity which subsequently can influence erosion prediction (Lee The particleeeddy interaction model predicts the effect of tur-
et al., 2002). bulent flow fluctuations on particle trajectories. The model assumes
In the Lagrangian approach, particles are considered as a that as particles travel through turbulent flow they pass through a
dispersed phase interacting with the fluid which is the continuous succession of turbulent structures which are called eddies. The
phase. The motion of each single particle is calculated by solving model has two steps. The first step is to determine the turbulent
the particle equation of motion which includes forces acting on the flow fluctuation velocity components at the particle location, which
particle. The following equation shows the particle equation of is needed for Equation (69). As mentioned earlier, solving the time-
motion commonly used in CFD: averaged NaviereStokes equations does not provide these fluctu-
ations. The second step is to calculate the time during which the
dVp particle interacts with each specific eddy.
¼ FD þ FV þ Fp þ FG (69)
dt Each of these eddies has different stochastic turbulent charac-
teristics which are defined by Gaussian distributed random velocity
In Equation (69), the right hand side represents the forces acting
fluctuations (V ), time scale (te) and length scale (le). The turbulent
on the particle, and the left hand side is the particle inertia. f
Drag force (FD) is the most important force acting per unit par- velocity fluctuation value, which is assumed to be constant during
ticle mass: the life time of the eddy and follows a Gaussian distribution, is
calculated from the following equation:
18mf CD Rep   rffiffiffiffiffiffi
FD ¼ Vf  Vp (70)
rp d2p 24  ¼ x
V
2
Vf (74)
f

Virtual mass is another force that affects particle trajectories. rffiffiffiffiffiffi


2
Virtual mass is a force that represents the volume of fluid which is Where, x is a Gaussian random number, Vf is the local root-
displaced by an accelerating or a decelerating particle. This force mean square (RMS) value of the velocity fluctuation and V  is the
f
cannot be neglected when the ratio of fluid density to particle velocity fluctuation. It is important to calculate separate random
density is high (solid particle-liquid flow). This force can be written numbers for different velocity components; else it results in non-
as follows: physical particle trajectory predictions.
The local RMS value of the velocity fluctuation can be calculated
1 rf d  
using the local value of turbulent kinetic energy (TKE):
FV ¼ Vf  Vp (71)
2 rp dt rffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffi
2 2TKE
When a particle passes through areas with high pressure Vf ¼ (75)
gradient, the force due to the pressure gradient can have important 3
impact on the particle trajectory. The following equation describes
Using the calculated value of velocity fluctuation, the local
the pressure gradient force:
instantaneous velocity can be calculated by summing the mean and
! fluctuating velocity components.
rf dVf The second step of the particleeeddy interaction model is to
Fp ¼ V pi (72)
rp dxi calculate the interaction time of a particle and a specific eddy. Ac-
cording to this model, when a particle enters a turbulent eddy, two
Other forces which can affect particle trajectories are gravity and different particle behaviors may be observed. First, the particle has
buoyancy: higher inertia as compare to the eddy. In this case, the particle
  crosses the eddy. The time needed for the particle to cross the eddy
rp  rf g
is called “eddy crossing” time. Second, the particle is trapped inside
FG ¼ (73)
rp the eddy and moves with it until the eddy life-time expires. The
particle eddy interaction time is the minimum of eddy crossing
Saffman lift force can also be included in Equation (69). It is a
time and eddy life time.
force resulting from shear in the flow. Particle trajectories can be
predicted by solving Equation (69). Tint ¼ min$〈tlife ; tcross 〉 (76)
In turbulent flow, the fluid velocity has two components: The eddy life time is calculated as follows:
mean and random fluctuation. The random velocity fluctuations
3
can affect particle trajectories. In turbulent particulate flow, 3
Cm4 TKE 2
ε
particle dispersion is ascribed to the fluctuation velocity com- tlife ¼ rffiffiffiffiffiffi (77)
2
ponents. Turbulence enhances the particles dispersion, and it is Vf
the dominant particle dispersion mechanism for regions not very
close to the wall (Bahoosh Kazerooni et al., 2010; Parsi et al.,
where Cm ¼ 0.09 is a turbulence model constant, and ε is the tur-
2012). The trajectories of small particles are more affected by
bulent dissipation rate.
turbulence than the trajectories of large particles (Parsi et al.,
The time for a particle to cross the eddy is
2013).
To have a more physical and accurate prediction of particle 0 1
behavior, the effect of turbulent fluctuation velocity components on le
tcross ¼ t Ln@1:0    A
 (78)
particle trajectories should be considered. Different researchers tVf  Vp 
proposed a variety of methods to calculate the effect of turbulent
868 M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873

where t as impact speed and impact angle of each particle in each CFD mesh
next to the wall are saved. The third step in predicting erosion rate
4 rp d is to feed the impact information into an erosion equation. Infor-
t¼  p  (79)
3 rf C V  V  mation from each impact is used in an erosion equation to calculate
D f p
surface mass loss that each particle causes. The overall erosion rate
is the summation of all mass losses that all particles cause. Fig. 12
is the particle relaxation time.
summarizes the three steps of the CFD-based erosion simulation.
When the interaction time reaches its end, a new eddy with new
To have an accurate erosion distribution on the geometry sur-
characteristics is sampled, and the new particleeeddy interaction
face, a statistically significant number of particle trajectories needs
time is calculated.
to be calculated. The number of particles needed for erosion eval-
When a particle impacts a wall, it loses some of its energy. In
uation depends on the geometry and flow pattern. For example, in a
different directions, different restitution coefficients are used to
direct impingement geometry, the carrier fluid moves and carries
account for this energy loss. Restitution coefficients show the ratio
particles towards the wall. Therefore, the necessary number of
of the particle velocity component after the impact to the particle
particle trajectories that need to be calculated is less than that in a
velocity component before the impact in the corresponding direc-
sudden contraction/expansion geometry in which the carrier fluid
tion. The restitution coefficients en and er, show the change in
moves in a parallel direction to walls in the straight sections.
particle momentum through the impact in normal and tangential
directions to the wall, respectively.
4. Evaluation of models versus experimental and field data
Vpn2
en ¼ (80.a)
Vpn1 Erosion rate prediction using E/CRC model and the software
developed based on many studies is called Sand Production Pipe
Vpt2 Saver (SPPS) is compared with 640 data points at different flow
et ¼ (80.b)
Vpt1 geometries, materials types, pipe diameters, particle diameters,
particle rates, water rates, oil rates, gas rates, and flow regimes
Vpn and Vpt are the particle velocity normal and tangential (Fig. 13). SPPS has been widely used in the oil and gas industry for
components, respectively. Subscript 1 is prior to impact and erosion rate prediction. This software contains all erosion models
Subscript 2 is after the impact. A coefficient of restitution of 1 in- developed by E/CRC and is under continuous development and
dicates that there is no momentum loss during the impact (elastic improvement. Erosion rate predictions show good agreement with
rebound), and a coefficient of restitution equal to 0 means that the experimental data. However, there are always opportunities for
particle loses all its momentum after the impact. Physically, this improvement especially for the conditions that erosion rate is un-
means that the particle sticks to the wall after impact. der predicted, since consequences of under prediction of erosion in
Incorporating a particleewall rebound model is essential in the oil and gas industry can be catastrophic. On the other hand, over
erosion simulations, since a particle can impact a wall multiple prediction of erosion rate to some extent helps to improve safety,
times. Researchers proposed different correlations for the co- but it is important to mention that over prediction of erosion rate
efficients of restitution based on the particle characteristics and the means lower production rate which has an important financial
impact conditions. Grant and Tabakoff (1975) and Forder et al. impact on oil and gas production. Therefore, developing more ac-
(1998) conducted experiments that reveal particle impact angle curate erosion models has always been supported by the oil and gas
affects the restitution coefficients. Forder et al. (1998) suggested the industry.
following correlations for the restitution coefficients: According to TÜV NEL report (Barton, 2003), there are a number
of erosion management techniques. Reduction of flow velocity and
en ¼ 1  0:4159a þ 0:59942  0:292a3 (81.a) sand production rate results in a decrease in erosion rate. Therefore,
reducing the production rate can be a way for managing erosion.
et ¼ 1  2:12a þ 3:0775a2  1:1a3 (81.b)
Devices such as sand screen, gravel pack, and hydro-cyclone can be
employed to separate sand from the flow and consequently reduce
where a is the impact angle in radians. They used AISI 413 carbon in
erosion rate. In designing of pipelines, some issues should be
their experiments. Grant and Tabakoff (1975) used a statistical
considered to minimize erosion. For example, abrupt changes
approach to develop correlations for these coefficients. They sug-
(elbow, plugged tee) must be avoided as changes in flow direction
gested the mean values of the coefficients and their corresponding
can be resulted in severe erosion. Pipe coating can provide a level of
standard deviations using 200 mm sand and Aluminum 2024 (target
protection against erosion.
wall) in their experiments:

en ¼ 0:993  1:76a þ 1:56a2  0:49a3 (82.a)

et ¼ 0:998  1:66a þ 2:11a2  0:67a3 (82.b)

sn ¼ 0:0005 þ 0:62a  0:535a2 þ 0:089a3 (82.c)

st ¼ 2:15a  5:02a2 þ 4:05a3  1:085a4 (82.d)


Chen et al. (2006b) showed the significance of incorporating a
stochastic particle rebound model for erosion prediction in plugged
tees.
Erosion occurs when particles crosses fluid streamlines and hit Fig. 12. Three steps of CFD-based erosion simulation: flow modeling, particle tracking
walls. When particles impact the wall, the impact information such and erosion calculation.
M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873 869

 Currently available erosion models are too conservative which


affects the efficiency of oil and gas production. Developing more
accurate erosion models will definitely be helpful for optimizing
production and reducing risk.

Nomenclature

a, b exponent constants
a1 constants
A shifting coefficient
Ai impact angle empirical constant
Apipe pipe cross-sectional area
0
b target material property constant
B shifting exponent
BH Brinell hardness of target
c Fraction of particles cutting in an idealized manner
C flow stress temperature dependence constant
Cp heat capacity of target
C1,2,3,… constants
Fig. 13. Erosion rate prediction (E/CRC models) versus experimental data at different CD drag coefficient
flow geometries, materials types, pipe diameters, particle diameters, particle rates,
Cm turbulence model constant
water rates, oil rates, gas rates, and flow regimes.
CFD computational fluid dynamics
dp particle diameter
5. Conclusions and scope for future work
D pipe diameter
Dref reference pipe diameter equal to 1 in
In this paper, the effects of different parameters on solid particle
d0 threshold particle diameter below which no erosion
erosion were discussed in detail. Several erosion equations devel-
occurs
oped by different authors were presented. Different empirical and
d0 standard particle diameter
mechanistic erosion prediction models for pipelines were surveyed.
D1,2,… fracture constants
CFD-based erosion prediction as a comprehensive method for
e coefficient of restitution
erosion prediction was explained in detail.
en restitution coefficients in normal direction
From the reviewed investigations it is concluded that erosion is
et restitution coefficients in tangential direction
affected by many factors. Therefore, there is scope for further im-
E liquid droplet entrainment fraction in annular flow gas
provements on erosion modeling and prediction as listed below:
core
Ep elastic modulus of particle
 On the erosion equation side of the problem, utilizing such a
ER erosion rate
large number of physical properties (Meng and Ludema (1995))
Et elastic modulus of target
shows that there are still aspects of the erosion phenomenon
Ee effective elastic modulus
which are unknown. The way that the particle size effect has
f fraction of volume loss caused by median spalling
been incorporated in the proposed models available in the
mechanism
literature shows that there is still room to further investigate the
F(t) number of impact constant
effect of particle size on erosion of materials. Small particle
F(z) function varying with z
erosion modeling is also considered as an important challenge
F(a) particle impact angle function
for erosion modeling which needs to be addressed in the future.
Fd,v fragmentation for test condition
 Considering the effect of target properties, it seems that the
FS particle shape factor
transition between brittle to ductile properties of the target
FD drag force
material requires further investigation.
FV virtual mass force
 Research is needed to determine the effect of high sand con-
FP pressure gradient force
centration on erosion. High sand concentration erosion can
Fe specific erosion factor
happen during reinjection of drill cuttings on offshore
FP penetration factor
platforms.
FM empirical constant accounting for material hardness
 Erosion rate prediction for complex geometries such as choke
Fr/D factor accounting for elbow radius
valves is also an area which needs more attention. Current
G gram molecular weight of target
models are limited to elbows, tees, and a few other simple ge-
g gravity acceleration
ometries. Developing erosion models for complex geometries is
Hp hardness of particle
an important and complex challenge for erosion modeling.
Ht hardness of target
 More experimental data are needed for erosion rate in different
HL liquid hold-up
multi-phase flow regimes. E/CRC is in the process of investi-
HLLS liquid holdup in slug body
gating erosion patterns in multi-phase churn flow in a standard
HV Vickers hardness of target
elbow. Future work will concentrate on developing a mecha-
i empirical exponent
nistic model for erosion predictions in churn flow.
I moment of inertia of single particle
 The effect of fluid viscosity on erosion rate is not fully under-
j exponent constant
stood. Further experimental studies are needed to determine
k thermal conductivity of target
fluid viscosity effect on erosion phenomenon.
k1,2,3… empirical constant
870 M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873

K velocity component normal to surface below which no VLLS liquid velocity of slug body
erosion takes place in certain hard materials Vm mixture velocity
Kc fracture toughness Vp particle velocity
Kp physical characteristics constant Vpn normal velocity component of particle
Ks fitting erosion constant Vpt tangential velocity component of particle
KT kinetic energy transferred from impacting particle to Vr residual parallel component of particle velocity at small
target per unit mass of particles angles of attack
K1,2,3,… proportionality constants Vref standard test velocity of particle
le eddy length scale VSG superficial gas velocity
L depth of deformation VSL superficial liquid velocity
LV plastic zone volume V0 threshold velocity below which distortion is entirely
Lstag stagnation region length elastic and no damage occurs
Lref reference length V0 standard impact velocity
LS slug body length W proportion of sample (by weight) within specified size
LF liquid film length range after testing
mp mass of single particle W0 proportion of sample (by weight) within specified size
M total mass of impinging particle range before testing
m_ sand mass ratio in slug body W _p sand flow rate
mils mili-inch We modified Weber number
nh strain hardening exponent xt horizontal coordinate
nc instantaneous strain hardening coefficient at strain Wc x_0t horizontal velocity of tip of particle when cutting ceases
np physical characteristics constant yt vertical coordinate
ns shape constant z exponent determining steepness of strain gradient
n1,2,… impact condition constant associated with each impact
p pressure
p mean stress
pn normal pressure component Greek letters
pt tangential pressure component a particle impact angle
P constant plastic flow stress ag gas volume fraction
r radius of particle am maximum erosion impact angle
R roundness of particle ap0 impact angle at which horizontal velocity component has
rc bend radius of curvature just become zero when particle leaves body
Rep particle Reynolds number d deformation wear factor, the amount of energy needed to
ReL liquid Reynolds numbers remove unit volume of material
S empirical constant in the range 0e0.5 dfilm liquid film thickness in annular flow
Sk constant DHm enthalpy of melting of target
Sm geometry dependent constant εc critical strain
SEM scanning electron microscope ε0 material ductility
T toughness ε_ * dimensionless strain rate
TC crater temperature εC erosion rate due to cutting (mass loss/impact mass)
Tm melting temperature εD erosion rate due to deformation (mass loss/impact mass)
Tt total work done by abrasive εm erosion rate (mass loss/impact mass)
Tx tangential work done by abrasive εV erosion rate (volume loss/impact mass)
Ty normal work done by abrasive εVC volume of material removed by cutting mechanism
T* dimensionless temperature εVC1,2 volumes of material removed by cutting mechanism 1
t time and 2
te eddy time scale εVP volume of material removed by single abrasive grain of
TKE turbulent kinetic energy mass
Tint particle eddy interaction time εVD volume of material removed by deformation mechanism
tlife eddy life time εVT total volume removal
tcross particle eddy cross time εw weight loss of target material
Vchar flow characteristic velocity ε1, ε2 primary erosion rate
Vchar-film characteristic flow velocity for particle in the liquid film of bε1 ; b
ε2 maximum erosion rates for velocity Vref
annular flow ε90 erosion rate (volume loss/impact mass)
Vchar-gas core characteristic flow velocity for particle in the gas core q pipe inclination angle
of annular flow k ratio of vertical to horizontal force component on particle
VDroplet droplet velocity in annular flow gas core l distance between center of mass of particle and the
Ve erosional velocity contact point
Vf fluid flow velocity m friction coefficient
Vfilm liquid film velocity mC critical friction coefficient
Vf0 fluctuation velocity mf fluid viscosity
VG_Core gas core velocity in annular flow mG gas viscosity
VGLS gas velocity in slug body mL liquid viscosity
VL particle impact velocity mm fluid mixture viscosity
M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873 871

mp Poisson coefficient of particle Chen, X., 2004. Application of Computational Fluid Dynamic (CFD) to Single Phase
and Multiphase Flow Simulation and Erosion Prediction (Ph.D. dissertation).
mt Poisson coefficient of target
Department of Mechanical Engineering, The University of Tulsa.
rf fluid density Chen, D., Sarumi, M., Al-Hassani, S., Gan, S., Yin, Z., 1997. A model for erosion at
rG density of gas normal impact. Wear 205 (1e2), 32e39.
rL density of liquid Chen, D., Sarumi, M., Al-Hassani, S., 1998. Computational mean particle erosion
model. Wear 214 (1), 64e73.
rm fluid mixture density Chen, X.H., McLaury, B.S., Shirazi, S.A., 2004. Application and experimental valida-
rp density of particle tion of a computational fluid dynamics (CFD)-based erosion prediction model in
rt density of target wall elbows and plugged tees. Comput. Fluids 33 (10), 1251e1272.
Chen, X., McLaury, B.S., Shirazi, S.A., 2006a. A comprehensive procedure to estimate
s* pressure-stress ratio erosion in elbows for gas/liquid/sand multiphase flow. ASME J. Energy Resour.
c cutting wear factor, the quantity of energy needed to Technol. 128, 70e78.
scratch out unit volume from a surface Chen, X.H., McLaury, B.S., Shirazi, S.A., 2006b. Numerical and experimental inves-
tigation of the relative erosion severity between plugged tees and elbows in
j ratio of depth of contact to depth of cut dilute gas/solid two-phase flow. Wear 261 (7e8), 715e729.
Uc critical strain for onset of lip formation Clark, H. Mcl, 1991. on the impact rate and impact energy of particles in a slurry pot
DUm mean strain increment induced by each impact erosion tester. Wear 146, 165e183.
Clark, H., Burmeister, L., 1992. The influence of the squeeze film on particle impact
ε turbulent dissipation rate velocities in erosion. Int. J. Impact Eng. 12 (No. 3), 415e426.
t particle relaxation time Coffin, L., 1954. A study of the effect of cyclic thermal stresses on a ductile metals.
sn standard deviation of restitution coefficients in normal Trans. ASME 76, 931e950.
Coker, A.K., 1990. Understand two-phase flow in process piping. Chem. Eng. Prog. 86
direction
(11), 60e65.
st standard deviation of restitution coefficients in tangential Deffenbaugh, D.M., Buckingham, J.C., 1989. A Study of the Erosional/Corrosional
direction Velocity Criterion for Sizing Multi-phase Flow Lines. Southwest Research
ag gas volume fraction Institute Final Report, Project No. 04-2433, prepared for the Minerals and
Management Service, U.S. Department of the Interior.
DPSB pressure drop in slug body Deng, T., Chaudhry, A.R., Patel, M., Hutchings, I., Bradley, M.S.A., 2005. Effect of particle
DPSU pressure drop in slug unit concentration on erosion rate of mild steel bends in a pneumatic conveyor. Wear
x Gaussian random number 258 (1e4), 480e487. http://dx.doi.org/10.1016/j.wear.2004.08.001.
Det Norske Vertitas, 1996. Recommended Practice RP 0501: Erosive Wear in Piping
Systems. Revision 1999.
Desale, G.R., Gandhi, B.K., Jain, S.C., 2009. Particle size effects on the slurry erosion
of aluminum alloy (AA 6063). Wear 266 (11e12), 1066e1071.
References Edwards, J.K., McLaury, B.S., Shirazi, S.A., 1998. Supplementing a CFD Code with
Erosion Prediction Capabilities. American Society of M0065chanical Engineers
Ahlert, K., 1994. Effect of Particle Impingement Angle and Surface Wetting on Solid (ASME) Fluids Engineering Division Summer Conference, Paper FEDSM
Particle Erosion of AISI 1018 Steel (M.Sc. thesis). Department of Mechanical 98e5229 Washington, DC, June 21e25.
Engineering, The University of Tulsa, Tulsa. Edwards, J.K., McLaury, B.S., Shirazi, S.A., 2001. Modeling solid particle erosion in
Andrews, D., 1981. An analysis of solid particle erosion mechanisms. J. Phys. D Appl. elbows and plugged tees. J. Energy Resour. Technol. 123 (4), 277e284.
Phys. 14, 1979e1991. Elkholy, A., 1983. Prediction of abrasion wear for slurry pump materials. Wear 84
Andrews, D.R., Horsfield, N., 1983. Particle collisions in the vicinity of an eroding (1), 39e49.
surface. J. Phys. D Appl. Phys. 16 (4), 525e538. http://dx.doi.org/10.1088/0022- Evans, A.G., Gulden, M.E., Rosenblatts, M., 1978. Impact damage in brittle materials
3727/16/4/014. in the elastic-plastic response regime. Proc. R. Soc. Lond. Ser. A (Math. Phys. Sci.)
Ansari, A.M., Sylvester, N.D., Sarica, C., Shoham, O., Brill, J.P., 1994. A comprehensive 361 (1706), 343e365. http://dx.doi.org/10.1098/rspa.1978.0106.
mechanistic model for upward two-phase flow in wellbores. SPE Prod. Facil. Eyler, R.L., 1987. Design and Analysis of a Pneumatic Flow Loop (M.S. thesis). West
20630, 143e152. Virginia University, WV, U.S.A.
API, 1981. API RP 14E Recommended Practice for Design and Installation of Offshore Fan, J., Yao, J., Zhang, X., Cen, K., 2001. Experimental and numerical investigation of a
Production Platform Piping Systems, third ed. American Petroleum Institute, new method for protecting bends from erosion in gas-particle flows. Wear 251
Washington, DC, p. 22. (1e12), 853e860.
Atkinson, M., Stepanov, E.V., Goulet, D.P., Sherikar, S.V., Hunter, J., 2007. High Fan, J.R., Yao, J., Cen, K.F., 2002. Anti-erosion in a 90 degrees bend by particle
pressure testing sand erosion in 3D flow channels and correlation with CFD. impaction. AIChE J. 48 (7), 1401e1412.
Wear 263, 270e277. Fan, J.R., Luo, K., Zhang, X.Y., Cen, K.C., 2004. Large eddy simulation of the anti-
Babu, P.S., Basu, B., Sundararajan, G., 2011. The influence of erodent hardness on the erosion characteristics of the ribbed-bend in gasesolid flows. J. Eng. Gas Tur-
erosion behavior of detonation sprayed WC-12Co coatings. Wear 270 (11e12), bines Power 126 (3), 672e679.
903e913. http://dx.doi.org/10.1016/j.wear.2011.02.019. Ferng, Y.M., 2008. Predicting local distributions of erosionecorrosion wear sites for
Bahoosh Kazerooni, R., Noghrehabadi, A.R., Bahrainian, S.S., Parsi, M., 2010. V2f the piping in the nuclear power plant using CFD models. Ann. Nucl. Energy 35,
study of nano and micro-particles transportation in turbulent boundary layer in 304e313.
conjunction with Eddy Interaction Model. In: Proc. The WSEAS 8th Conference Finnie, I., 1958. The Mechanism of Erosion of Ductile Metals, 3rd U.S. Nat. Congress
on Fluid Mechanics, Taipei, Taiwan. of Applied Mechanics, New York, ASME, pp. 527e532.
Barton, N.A., 2003. Erosion in Elbows in Hydrocarbon Production Systems: Review Finnie, I., 1960. Erosion of surfaces by solid particles. Wear 3 (2), 87e103.
Document. TÜV NEL Limited RESEARCH REPORT 115 for the Health and Safety Finnie, I., 1972. Some observations on the erosion of ductile metals. Wear 19 (1),
Executive. 81e90.
Bellman, R., Levy, A., 1981. Erosion mechanism in ductile metals. Wear 70, 1e28. Finnie, I., Wolak, J., Kabil, Y., 1967. Erosion of metal by solid particles. J. Mater. 2,
Bikbiaev, K.A., Krasnov, V.I., Maksimenko, M.I., Berezin, V.L., Zhilinskii, I.B., 1972. 682e700.
Main factors affecting gas abrasive wear of elbows in pneumatic conveying Foley, T., Levy, A., 1983. The erosion of heat treated steels. Wear 91, 45e64.
pipes. Chem. Pet. Eng. 8, 465e466. Forder, A., Thew, M., Harrison, D., 1998. A numerical investigation of solid particle
Birchenough, P.M., Dawson, S.G.B., Lockett, T.J., McCarthy, P., 1995. Critical Flow erosion experienced within oilfield control valves. Wear 216 (2), 184e193.
Rates Working Party. Report No. AEA-TSD-0348. AEA Technology, UK. Fore, L.B., Dukler, L.E., 1994. The distribution of drop size and velocity in gas-liquid
Bitter, J., 1963a. A study of erosion phenomena part I. Wear 6 (1), 5e21. annular flow. Int. J. Multiph. Flow 21 (2), 137e149.
Bitter, J., 1963b. A study of erosion phenomena: part II. Wear 6 (3), 169e190. Gandhi, B.K., Borse, S.V., 2002. Effect of particle size and size distribution on esti-
Bourgoyne, A.T., 1989. Experimental study of erosion in diverter systems due to mating erosion wear of cast iron in sand-water slurries. Indian J. Eng. Mater. Sci.
sand production. In: Proc., SPE/IADC Drilling Conference, New Orleans, LA, SPE/ 9, 480e486.
IADC 18716. Gandhi, M.B., Vuthaluru, R., Vuthaluru, H., French, D., Shah, K., 2012. CFD based
Brach, R., 1988. Impact dynamics with applications to solid particle erosion. Int. J. prediction of erosion rate in large scale wall-fired boiler. Appl. Therm. Eng. 42,
Impact Eng. 7 (1). 90e100.
Brown, R., Jun, E.J., Edington, J.W., 1981. Erosion of a-Fe by spherical glass particles. Gipson, F., 1989. Petroleum Production Engineering, Pits and Pieces. Manual of
Wear 70 (3), 347e363. http://dx.doi.org/10.1016/0043-1648(81)90355-0. Southwest Petroleum Short Course, Texas Tech University, April 17e20.
Burnett, A., De Silva, S., Reed, A., 1994. Comparison between sand blast and cen- Gomez, L.E., Shoham, O., Taitel, Y., 2000. Prediction of slug-liquid holdup horizontal
tripetal effect accelerator type erosion testers. In: Proceeding of the 8th Inter- to upward vertical flow. Int. J. Multiph. Flow 263, 517e521.
national Conference on Erosion by Liquid and Solid Impact, pp. 168e178. Gosman, A.D., Ioannides, E., 1983. Aspects of computer simulation of liquid-fuelled
Chase, D., Rybicki, E., Shadley, J., 1992. A model for the effect of velocity on erosion combustors. J. Energy 7, 482e490.
of N80 steel tubing due to the normal impingement of solid particle. J. Energy Grant, G., Tabakoff, W., 1975. Erosion prediction in turbomachinery resulting from
Resour. Technol. 114, 54e64. environmental solid particles. J. Aircr. 12 (5), 471e478.
872 M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873

Gundameedi, V.M., 2008. Performance of Electrical Resistance Probes and Acoustic Levy, A., 1979. The role of plasticity in erosion. In: Proceeding of the 5th Interna-
Monitors in Slug Flow (M.S. thesis). Department of Mechanical Engineering, The tional Conference on Erosion by Liquid and Solid Impact, pp. 39-41e39-10.
University of Tulsa, Tulsa, Oklahoma, USA. Levy, A., 1995. Solid Particle Erosion and Erosion-Corrosion of Materials. ASM In-
Habib, M.A., Badr, H.M., Ben-Mansour, R., Said, S.A.M., 2004. Numerical calculations ternational, Material Park, Ohio.
of erosion in an abrupt pipe contraction of different contraction ratios. Int. J. Levy, A., Chik, P., 1983. The effect of erodent composition and shape on the erosion
Numer. Methods Fluids 46 (1), 19e35. of steel. Wear 89, 151e162.
Habib, M.A., Badr, H.M., Ben-Mansour, R., Kabir, M.E., 2007. Erosion rate correlations Levy, A., Hickey, G., 1982. Surface degradation of metals in simulated synthetic fuels
of a pipe protruded in an abrupt pipe contraction. Int. J. Impact Eng. 34 (8), plant environments. In: NACE Corrosion/82, International Corrosion Forum, p. 154.
1350e1369. Li, G., Wang, Y., He, R., Cao, X., Lin, C., Meng, T., 2009. Numerical simulation of
Habib, M.A., Ben-Mansour, R., Badr, H.M., Kabir, M.E., 2008. Erosion and penetration predicting and reducing solid particle erosion of solideliquid two-phase flow in
rates of a pipe protruded in a sudden contraction. Comput. Fluids 37 (2), a choke. Pet. Sci. 6 (1), 91e97.
146e160. Liebhard, M., Levy, A., 1991. The effect of erodent particle characteristics on the
Haugen, K., Kvernvold, O., Ronald, A., Sandberg, R., 1995. Sand erosion of wear-resistant erosion of metals. Wear 151, 151e162.
materials: erosion in choke valves. Wear 186e187 (Part I), 179e188. Lu, Y., Agrawal, M., 2013. A computational-fluid-dynamics-based Eulerian-granular
Heidersbach, R., 1985. Velocity limits for erosion-corrosion. In: Proc., 17th Offshore approach for characterization of sand erosion in multiphase-flow systems. SPE
Technology Conference, Houston, Texas, May 6e9, Paper OTC 4974. J. 19 (4), 586e597.
Hinze, J., 1972. Turbulent fluid and particle interaction. Prog. Heat Mass Transf. 6, Lu, Q.Q., Fontaine, J.R., Aubertin, G., 1992. Particle motion in two-dimensional
433e452. confined turbulent flows. Aerosol Sci. Technol. 17 (3), 169e185.
Huang, C., Chiovelli, S., Minev, P., Nandakumar, K., 2008. A comprehensive Lynn, S., Wong, K.K., Clark, H.M., 1991. On the particle size effect in slurry erosion.
phenomenological model for erosion of materials in jet flow. Powder Technol. Wear 149 (1e2), 55e71. http://dx.doi.org/10.1016/0043-1648(91)90364-Z.
187 (3), 273e279. Mamoun, M., 1975. Analytical Models for the Erosive-Corrosive Wear Process.
Humphrey, J., 1990. Fundamental of fluid motion in erosion by solid particle impact. Argonne National Laboratory.
Int. J. Heat Fluid Flow 11, 170e195. Manson, S., 1953. Behavior of Materials Under Conditions of Thermal Stress, Heat
Hutchings, I., 1980. Some comments on the theoretical treatment of erosive particle Transfer Symposium. University of Michigan Engineering Research Institute,
impacts. In: Proceeding of the 5th International Conference on Erosion by Michigan.
Liquid and Solid Impact, pp. 36e41. Marchiolo, C., Soldati, A., 2002. Mechanisms for particle transfer and segregation in
Hutchings, I., 1981. A model for the erosion of metals by spherical particles at a turbulent boundary layer. J. Fluid Mech. 468, 283e315.
normal incidence. Wear 70 (3), 269e281. Mazumder, Q.H., 2004. Development and Validation of a Mechanistic Model to
Hutchings, I., Winter, R., Field, J., 1976. Solid particle erosion of metals: the removal Predict Erosion in Single-phase and Multiphase Flows (Ph.D. thesis). Depart-
of surface material by spherical projectiles. Proc. R. Soc. Lond. Ser. A Math. Phys. ment of Mechanical Engineering, The University of Tulsa.
Sci. 348, 370e392. Mazumder, Q.H., 2007. Prediction of erosion due to solid particle impact in single-
Ishi, M., Mishima, K., 1989. Droplet entrainment correlation in annular two-phase phase and multiphase flows. J. Press. Vessel Technol. 129, 576e582.
flow. Int. J. Heat Mass Transf. 32 (10), 1835e1846. Mazumder, Q.H., Santos, G., Shirazi, S.A., McLaury, B.S., 2003. Effect of Sand Dis-
Jahanmir, S., 1980. The mechanics of subsurface damage in solid particle erosion. tribution on Erosion in Annular Three Phase Flow, 8th International Symposium
Wear 61, 309e338. on Liquid-Solid Flow, ASME Summer Meeting, Honolulu, HI, Paper No.
Jennings, W., Head, W., Manning Jr., C., 1976. A mechanistic model for the prediction FEDSM2003e45498.
of ductile erosion. Wear 40 (1), 93e112. Mazumder, Q.H., Shirazi, S.A., McLaury, B.S., 2004. A Mechanistic Model to Predict
Johansson, S., Ericson, F., Schweitz, J., 1987. Solid particle erosion d a statistical Sand Erosion in Multiphase Flow in Elbows Downstream of Vertical Pipes.
method for evaluation of strength properties of semiconducting materials. Corrosion, 2004, Paper No. 04662, NACE International Annual Conference, New
Wear 115 (1e2), 107e120. Orleans, Louisiana.
Johnson, G., Cook, W., 1985. Fracture characteristics of three metals subjected to Mazumder, Q., Shirazi, S., McLaury, B., Rybicki, E., Shadley, J., 2005. Development
various strains, strain rates, temperatures and pressures. Eng. Fract. Mech. 21 and validation of a mechanistic model to predict solid particle erosion in
(1), 31e48. multiphase flow. Wear 259, 203e207.
Jordan, K., 1998. Erosion in Multiphase Production of Oil & Gas. Corrosion 98, Paper Mazur, Z., Campos-Amezcua, R., Urquiza-Beltran, G., Garcia-Gutierrez, A., 2004.
No. 58, NACE International Annual Conference, San Antonio. Numerical 3D simulation of the erosion due to solid particle impact in the main
Kesana, N.R., Vieira, R.E., Schleicher, E., McLaury, B.S., Shirazi, S.A., Hampel, U., stop valve of a steam turbine. Appl. Therm. Eng. 24, 1877e1891.
2013a. Experimental investigation of slug characteristics through a standard McLaury, B.S., 1993. A Model to Predict Solid Particle Erosion in Oilfield Geometries
pipe bend. In: Proc. of the ASME 2013 International Mechanical Engineering (M.Sc. thesis). The University of Tulsa, Tulsa, OK.
Congress & Exposition, IMECE2013, November 15-21, 2013, San Diego, Califor- McLaury, B.S., 1996. Predicting Solid Particle Erosion Resulting from Turbulent
nia, USA. Fluctuations in Oilfield Geometries (Ph.D. thesis). The University of Tulsa, Tulsa,
Kesana, N.R., Vieira, R.E., Schleicher, E., McLaury, B.S., Shirazi, S.A., Hampel, U., OK.
2013b. Experimental study of slug characteristics: implications to sand erosion. McLaury, B.S., Shirazi, S.A., 2000. An alternative method to API RP 14E for predicting
In: Proceedings of: the ASME 2013 Fluids Engineering Division Summer solids erosion in multiphase flow. ASME J. Energy Resour. Technol. 122, 115e122.
Meeting FEDSM2013 July 7-11, 2013, Incline Village, Nevada, USA. McLaury, B.S., Shirazi, S.A., Shadley, J.R., Rybicki, E.F., 1999. How Operating and
Kesana, N.R., Grubb, S.A., McLaury, B.S., Shirazi, S.A., 2013c. Ultrasonic measurement Environmental Conditions Affect Erosion. Corrosion 99, Paper No. 34, NACE
of multiphase flow erosion patterns in a standard elbow. ASME J. Energy Resour. International Annual Conference, San Antonio, TX.
Technol. 135 (3), 032905. McLaury, B.S., Shirazi, S.A., Mazumder, Q.H., Viswanathan, V., 2006. Effect of Up-
Kesana, N.R., Throneberry, J.M., McLaury, B.S., Shirazi, S.A., Rybicki, E.F., 2013d. Effect stream Pipe Orientation on Erosion in Bends for Annular Flow. Source: NACE,
of particle size and liquid viscosity on erosion in annular and slug flow. J. Energy CORROSION 2006, Paper No. 065721-0657219, 2006.
Resour. Technol. doi:136(1):012901-012901-10. McLaury, B.S., Shirazi, S.A., Rybicki, E., 2010. Sand Erosion in Multiphase Flow For
Kesana, N.R., Vieira, R.E., Parsi, M., McLaury, B.S., Shirazi, S.A., 2014a. A mechanistic Slug and Annular Flow Regimes. Corrosion 2010, Paper No. 10377, NACE Inter-
model for predicting erosion in pseudo slug flow. In: Proc. of: “The NACE 2014 national Annual Conference, San Antonio, TX.
Conference & Expo, June 8e11, 2014, San Antonio, TX. McLaury, B.S., Shirazi, S.A., Viswanathan, V., Mazumder, Q.H., Santos, G., 2011. Dis-
Kesana, N.R., Vieira, R.E., McLaury, B.S., Shirazi, S.A., 2014b. Experimental study of tribution of sand particles in horizontal and vertical annular multiphase flow in
sand particle concentration profiles in straight and pipe elbow for horizontal pipes and the effects on sand erosion. J. Energy Resour. Technol. doi:133(2):
multiphase flows. J. Energy Resour. Technol. doi:136(3):033001-033001-11. 023001-023001-10.
Kleis, I., Kulu, P., 2008. Solid Particle Erosion Occurrence, Prediction and Control. Mehrotra, V., Silcox, G.D., Smith, P.J., 1998. Numerical simulation of turbulent par-
Springer-Verlag London Limited. Library of Congress Control Number: ticle dispersion using a Monte Carlo approach. In: Proc. of. ASME Fluids Engi-
2007937988. neering Division Summer Meeting, Washington.
Klose, G., Rembold, B., Koch, R., Wittig, S., 2001. Comparison of state-of-the-art Meng, H., Ludema, K., 1995. Wear models and predictive equations: their form and
droplet turbulence interaction models for jet engine combustor conditions. content. Wear 181e183 (Part 2), 443e457.
Int. J. Heat Fluid Flow 22, 343e349. Meyer, E., 1908. Contribution to the knowledge of hardness and hardness testing.
Kokal, S.L., Stanislav, J.F., 1989. An experimental study of two-phase flows in slightly Z. Ver. Dtsch. Ing. 52, 645e654.
inclined pipes-І. Flow patterns. Chem. Eng. Sci. 44 (3), 665e679. Milojevic, D., 1990. Lagrangian Stochastic-Deterministic (LAD) prediction of particle
Kvernvold, O., Sandberg, R., 1993. Production Rate Limits in Two-phase Flow Sys- dispersion in turbulence. Syst. Charact. 7, 181e190.
tems: Erosion in Piping Systems for Production of Oil and Gas. Technical Report Najafifard, F., 2014. Predicting Near Wall Particle Behavior with Application to
No. 93-3252. Det Norske Veritas (DNV), Norway. Erosion Simulation (Ph.D. dissertation). Department of Mechanical Engineering,
Laitone, J., 1979a. Aerodynamic effect in the erosion process. Wear 56, 239e246. The University of Tulsa.
Laitone, J., 1979b. Erosion prediction near the stagnation point resulting from Neilson, J., Gilchrist, A., 1968. Erosion by a stream of solid particles. Wear 11 (2),
aerodynamically entrained solid particles. J. Aircr. 16, 809e814. 111e122.
Lee, B.E., Tu, J.Y., Fletcher, C.A.J., 2002. On numerical modeling of particleewall Njobuenwu, D.O., Fairweather, M., 2012. Modelling of pipe bend erosion by dilute
impaction in relation to erosion prediction: Eulerian versus Lagrangian method. particle suspensions. Comput. Chem. Eng. 42, 235e247.
Wear 252, 179e188. Njobuenwu, D.O., Fairweather, M., Yao, J., 2012. Prediction of turbulent gasesolid
Levin, B., Vecchio, K., Marder, A., 1999. Modeling solid-particle erosion of ductile flow in a duct with a 90 bend using an EulerianeLagrangian approach. AIChE J.
alloys. Metall. Mater. Trans. A 30A, 1763e1774. 58 (1), 14e30.
M. Parsi et al. / Journal of Natural Gas Science and Engineering 21 (2014) 850e873 873

Nsoesie, S., Liu, R., Chen, K., Yao, M., 2014. Analytical modeling of solid-particle Song, X.J., Park, J.H., Kim, S.G., Park, Y.C., 2013. Performance comparison and erosion
erosion of Stellite alloys in combination with experimental investigation. prediction of jet pumps by using a numerical method. Math. Comput. Model. 57,
Wear 309 (1e2), 26e232. 245e253.
Oka, Y., Okamura, K., Yoshida, T., 2005. Practical estimation of erosion damage Sooraj, V.,S., Radhakrishnan, V., 2013. Elastic impact of abrasives for controlled
caused by solid particle impact: part 1: effects of impact parameters on a erosion in fine finishing of surfaces. ASME J. Manuf. Sci. Eng. 135
predictive equation. Wear 259 (1e6), 95e101. (051019e051021).
Oka, Y., Yoshida, T., 2005. Practical estimation of erosion damage caused by solid Srinivasan, S., Scattergood, R.O., 1988. Effect of erodent hardness on erosion of
particle impact: part 2: mechanical properties of materials directly associated brittle materials. Wear 128 (2), 139e152.
with erosion damage. Wear 259 (1e6), 102e109. Sundararajan, G., 1991. A comprehensive model for the solid particle erosion of
Parsi, M., 2014. Sand Erosion in Vertical Slug/Churn Flow. Advisory Board Meeting, ductile materials. Wear 149 (1e2), 111e127.
Erosion/Corrosion Research Center, The University of Tulsa, May 2014. Sundrarajan, G., Shewmon, P., 1983. A new model for the erosion of metals at
Parsi, M., Noghrehabadi, A.R., Bahreinian, S.S., 2010a. Numerical study of particle normal incidence. Wear 84 (2), 237e258.
inertia effect using 3-Eddy Interaction Model. In: Proc. The WSEAS 8th Con- Svedeman, S.J., Arnold, K.E., 1993. Criteria for Sizing Multiphase Flow Lines for
ference on Fluid Mechanics, 20-22 August, 2010 a, Taipei, Taiwan. Erosive-Corrosive Service. Proceedings of 68th SPE Annual Fall Technical Con-
Parsi, M., Noghrehabadi, A.R., Bahreinian, S.S., 2010b. Numerical study of turbulence ference, Houston, Paper No. SPE-26569.
anisotropy effect on particle deposition rate using DNS data. In: Proc. The WSEAS Svedeman, S.J., Arnold, K.E., 1994. Criteria for sizing multi-phase flow lines for
8th Conference on Fluid Mechanics, 20-22 August, 2010 b, Taipei, Taiwan. erosive/corrosive services. SPE Prod. Facil. 9 (1), 74e80.
Parsi, M., Mahdavimanesh, M., Noghrehabadi, A.R., 2012. Surveying aerosol deposi- Tabakoff, W., Kotwal, R., Hamed, A., 1979. Erosion study of different materials
tion in a turbulent duct flow. In: Proceeding 20th Annual International Conference affected by coal ash particles. Wear 52 (1), 161e173.
on Mechanical Engineering-ISME2012, 16e18, May, 2012, Shiraz, Iran, 4 Pages. Tabor, D., 1951. The Hardness of Metals. Clarendon Press, Oxford.
Parsi, M., Mahdavimanesh, M., Noghrehabadi, A.R., Ahmadi, G., 2013. Particle Taitel, Y., Barnea, D., 1990. Two-phase Slug Flow Academic. New York.
Deposition in a Turbulent Channel Flow. ASME 2013 Fluids Engineering Division Throneberry, J.M., 2010. Solid Particle Erosion in Slug Flow (M.S. thesis). Depart-
Summer Meeting FEDSM2013, 5. ment of Mechanical Engineering, The University of Tulsa, Tulsa, Oklahoma, USA.
Parsi, M., Vieira, R.E., Torres, C.F., Kesana, N.R., McLaury, B.S., Shirazi, S.A., Tilly, G., 1973. A two stage mechanism of ductile erosion. Wear 23 (1), 87e96.
Hampel, U., Schleicher, E., 2014. Characterizing Slug/Churn Flow Using Wire Timothy, L.B., 1999. Laser Doppler Velocimeter Measurements for Validation of
Mesh Sensor. The ASME 2014 Fluids Engineering Division Summer Meeting Turbulence Modeling in Choke Geometries (Master of Science thesis). The
FEDSM2014 Chicago, US. University of Tulsa, Tulsa, OK.
Patton, C.C., 1993. Are We out of the Iron Age yet?. Corrosion 93, Paper No. 56, NACE. Timothy, L.B., McLaury, M.S., 2002. Laser Doppler velocimeter measurements to
Paz, R.J., Shoham, O., 1994. Film Thickness Distribution for Annular Flow in Direc- characterize turbulence in a constriction with sharp and rounded inlets. Exp.
tional Wells e Horizontal to Vertical. SPE Annual Technical Conference and Fluids 32 (4), 472e480.
Exhibition, SPE 28541. Tolle, G.C., Greenwood, D.R., 1977. Design of Fittings to Reduce Wear Caused by Sand
Pereira, G.C., De Souza, F.J., De Moro Martins, D.A., 2014. Numerical prediction of the Erosion. API OSAPER Project No. 6, American Petroleum Inst., Texas A&M
erosion due to particles in elbows. Powder Technol. 261, 105e117. Research Foundation.
Phillips, M., 1980. A force balance model for particle entrainment into a fluid Venougopal Reddy, A., Sundararajan, G., 1986. Erosion behaviour of ductile mate-
stream. J. Phys. D Appl. Phys. 13, 221e233. Printed in Great Britain. rials with a spherical non-friable erodent. Wear 111 (3), 313e323.
Rabinowicz, E., 1979. The Wear Equation for Erosion of Metals by Abrasive Particles. Vieira, R., 2014. Erosion Prediction under Multiphase Annular and Low-liquid Flow
Proc. 5th International Conference on Erosion by Liquid and Solid Impact, Paper Conditions. Advisory Board Meeting, Erosion/Corrosion Research Center, The
No 38, Cambridge, England. University of Tulsa, May 2014.
Rodriguez, J.C., 2008. Effects of Liquid Viscosity and Sand Size on Erosion in Slug Vieira, R.E., Kesana, N.R., Torres, C.F., McLaury, B.S., Shirazi, S.A., Schleicher, E.,
Multiphase Flow (M.S. thesis). Department of Mechanical Engineering, The Hampel, U., 2014a. Experimental investigation of horizontal gas-liquid stratified
University of Tulsa, Tulsa, Oklahoma, USA. and annular flow using wire-mesh sensor. ASME J. Fluids Eng. 136 (12), 121301-
Salama, M.M., 1998. An Alternative to API 14E Erosional Velocity Limits for Sand 1e121301-16.
Laden Fluids. Offshore Technology Conference, Houston, OTC 8898, 1998. Vieira, R.E., Parsi, M., Torres, C.F., McLaury, B.S., Shirazi, S.A., Hampel, U., Schleicher, E.,
Salama, M.M., 2000. An alternative to API 14E erosional velocity limits for sand- 2014b. Experimental study of vertical gas-liquid pipe flow for annular and liquid
laden fluids. J. Energy Resour. Technol. 122, 71e77. loading conditions using dual wire-mesh sensors. In: Proc. ASME 2014 Fluids
Salama, M.M., Venkatesh, E.S., 1983. Evaluation of API RP 14E Erosional Velocity Engineering Division Summer Meeting FEDSM 2014 Chicago, US.
Limitations for Offshore Gas Wells. Proc., 15th Offshore Technology Conference, Wada, S., Watanabe, N., 1987. Solid particle erosion of brittle materials (Part 3), the
Paper OTC 4485. interaction with material properties of target and that of impingement on
Salik, J., Buck;ey, D., Brainard, W., 1981. The effect of mechanical surface and heat erosive wear mechanism. Yogyo-Kyyokai-Shi 95 (10), 573e578.
treatments on erosion resistance of 6061 aluminum alloy. Wear 65, 351e358. Wang, J.R., Shirazi, S.A., 2003. A CFD based correlation for erosion factor for long-
Santos, G., 2002. Effect of Sand Distribution on Erosion and Correlation Between radius elbows and bends. J. Energy Resour. Technol. 125 (1), 26e34.
Acoustic Sand Monitor and Erosion Test in Annular Multiphase Flow (M.S. Wang, J., Shirazi, S., Shadley, J., Rybicki, E., 1996. Application of flow modeling and
thesis). Department of Mechanical Engineering, The University of Tulsa. particle tracking to predict sand erosion rates in elbows. ASME Fluids Eng. Div.
Sekoguchi, K., Mori, K., 1997. New development of experimental study on interfacial 236, 725e734.
structure in gaseliquid two-phase flow. In: Giot, M., Mayinger, F., Celata, G.-P. Weiner, P.D., Tolle, G.C., 1976. Detection and Prevention of Sand Erosion of Pro-
(Eds.), Experimental Heat Transfer Fluid Mechanics and Thermodynamics, vol. duction Equipment. API OSAPER Project No. 2, American Petroleum Inst., Texas
2. Edizione ETS, pp. 177e1188. A&M Research Foundation.
Sheldon, G., 1970. Similarities and differences in the erosion behavior of materials. Zhang, H.Q., Wang, Q., Sairca, C., Brill, J.P., 2003a. Unified model for gas-liquid pipe
J. Basic Eng. Trans. ASME 619e626. flow via slug dynamics e part 1: model development. J. Energy Resour. Technol.
Sheldon, G., Kanhere, Ashok, 1972. An investigation of impingement erosion using 125 (4), 266e273.
single particles. Wear 21 (1), 195e209. http://dx.doi.org/10.1016/0043-1648(72) Zhang, H.Q., Wang, Q., Sairca, C., Brill, J.P., 2003b. Unified model for gas-liquid pipe
90257-8. flow via slug dynamics e part 2: model validation. J. Energy Resour. Technol.
Shipway, P., Hutchings, I., 1994. A method of optimizing the particle flux in erosion 125 (4), 274e283.
testing with a gas blast apparatus. Wear 174, 169e174. Zhang, Y., Reuterfors, E.P., McLaury, B.S., Shirazi, S.A., Rybicki, E.F., 2007. “Compar-
Shipway, P., Hutchings, I., 1996. The role of particle properties in the erosion of ison of computed and measured particle velocities and erosion in water and air
brittle materials. Wear 193 (1), 105e113. http://dx.doi.org/10.1016/0043- flows”. Wear 263, 330e338.
1648(95)06694-2. Zhang, Y., McLaury, B.S., Shirazi, S., Rybicki, E.F., 2010. A Two-Dimensional Mechanistic
Shirazi, S.A., Shadley, J.R., McLaury, B.S., Rybicki, E.F., 1995. A procedure to predict Model for Sand Erosion Prediction Including Particle Impact Characteristics.
particle erosion in elbows and tees. J. Press. Technol. 117, 45e52. Corrosion 2010, Paper No. 10378, NACE, March 14e18, San Antonio, Texas, USA.
Smart, J.S., 1990. A Review of Erosion Corrosion in Oil and Gas Production. Corrosion Zhang, Y., McLaury, B.S., Shirazi, S.A., Rybicki, E.F., 2011. Predicting Sand Erosion in
90, Paper 10, NACE. Slug Flows Using a Two-Dimensional Mechanistic Model. Corrosion 2011, Paper
Smart, J.S., 1991. The Meaning of the API RP 14E Formula for Erosion/Corrosion in Oil No. 11243, NACE International Annual Conference, Houston, TX, USA, March 13-
and Gas Production. Corrosion 91, Paper 468, NACE. 17, 2011.
Smeltzer, C., Gulden, M., Compton, W., 1970. Mechanisms of metal removal by
impacting dust particles. J. Basic Eng. Trans. ASME 639e654.

You might also like