Vous êtes sur la page 1sur 22

# Computers & Fluids 34 (2005) 721742

www.elsevier.com/locate/compuid

## Numerical investigation of erosion threshold velocity in

a pipe with sudden contraction
H.M. Badr *, M.A. Habib, R. Ben-Mansour, S.A.M. Said
Mechanical Engineering Department, King Fahd University of Petroleum and Minerals,
Box # 322, Dhahran 31261, Saudi Arabia
Received 14 April 2003; received in revised form 31 January 2004; accepted 27 May 2004
Available online 27 October 2004

Abstract
This paper deals with erosion prediction in a pipe with sudden contraction for the special case of twophase (liquid and solid) turbulent ow with low particle concentration. The pipe axis was considered vertical and the ow was either in direction of gravity (downow) or against it (upow). The mathematical
models for the calculations of the uid velocity eld and the motion of the solid particles have been established and an erosion model was used to predict the erosion rate. The uid velocity (continuous phase)
model was based on the time-averaged governing equations of 3-D turbulent ow and the particle-tracking
model (discrete phase) was based on the solution of the governing equation of each particle motion taking
into consideration the eect of particle rebound behavior. The eects of ow velocity and particle size were
investigated for one contraction geometry considering water ow in a steel pipe. The results showed the
strong dependence of erosion on both particle size and ow velocity but with little dependence on the direction of ow. The eect of ow direction was found to be signicant only for large particle size and moderate
ow velocity. The erosion critical area was found to be the inner surface of the tube sheet (connecting the
two pipes) in the region close to the small pipe. The results also indicated the presence of a threshold velocity below which erosion is insignicant for all particle sizes.

Corresponding author. Tel.: +966 3 860 2543; fax: +966 3 860 2949.

doi:10.1016/j.compuid.2004.05.010

722

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

Nomenclature
A
b
CD
Cl
C1
C2
C 2
d
Dp
E
F
Gk
g
k
mp
Np
p
Rep
s
Uj
u
uj
up
Vi
Vt
up
xj
t
Greek
a
e
l
q
rk
re

surface area
constant dened in Eq. (13)
drag coecient
constant dened in Eq. (4)
constant dened in Eq. (6)
constant dened in Eq. (8)
constant dened in Eq. (6)
diameter
Solid particle diameter
Erosion rate, mg/g
force
generation of turbulent kinetic energy
gravitational acceleration
turbulent kinetic energy
mass of individual particle
number of particles
pressure
particle Reynolds number
Sand ow
average velocity component
uid velocity vector
uctuating velocity component
particle velocity
Flow inlet velocity
Threshold erosional velocity
particle velocity
space coordinate
time
letters
impact angle
dissipation rate of turbulent kinetic energy
dynamic viscosity
density
eective Prandtl number for k
eective Prandtl number for e

Superscripts

time rate

time average

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

723

Subscripts
D
drag
f
uid
sl
Saman lift
lc
local
m
target material
p
particle
pg
vm
virtual mass

1. Introduction
Erosion is one of the important problems in various gas and liquid ow passages such as ow in
pipes and pipe ttings (valves, bends, elbows, ow meters, . . . etc.), ow in pumps, turbines, compressors and many others. Erosion may cause equipment malfunctioning (vibration, leakage,
excessive energy losses, . . . etc.) and may also lead to complete failure of machine components.
Accurate prediction of the rate of erosion in a specic application is one of the very complicated
problems since it requires detailed investigation of the solid particle motion before and after impact. The diculty arises mainly from the fact that most ows occurring in industrial processes
are turbulent which makes the particle trajectory and impact characteristics dicult to predict
taking into consideration all uid forces acting on the particle. The following literature review
is limited to previous work done on erosion in pipes and pipe ttings.
Previous erosion studies can be classied under three categories; experimental investigations,
erosion model developments, and numerical simulations. Tilly [1] presented a thorough analysis
of the various parameters aecting erosion, including particle properties, impact parameters, particle concentration, material temperature, and tensile stress. He also reviewed the dierent mechanisms of erosion, which were categorized into brittle and ductile behaviors. Ru and Wiederhorn
[2] presented another review of the solid particle erosion phenomena considering single and multiple particle models on erosion of metals and ceramics. The signicant parameters for eroding
particles and material characteristics were also presented. Humphrey [3] reported a more comprehensive review of the fundamentals of uid motion and erosion by solid particles. The review includes a discussion of the experimental techniques and the various fundamental considerations
relating to the motion of solid particles. An assessment of the uid mechanics phenomena that
can signicantly inuence erosion of material surfaces by impinging particles was also presented.
Because of its direct relevance to gas and oil industries, erosion of pipes and pipe ttings attracted
many researchers. Several experimental studies were conducted with the main objective being to
determine the rate of erosion in such ow passages and its relation with the other parameters involved in the process. Among these studies are the works by Rochester and Brunton [4], True
and Weiner [5], Glaeser and Dow [6], Roco et al. [7], Venkatesh [8], and Shook et al. [9]. Soderberg
et al. [10,11] and Hutchings [12,13] reported the advantages and disadvantages of such experiments.
The recent experimental study by McLaury et al. [14] on the rate of erosion inside elbows and
straight pipes provided correlations between the penetration rate and the ow velocity at dierent

724

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

values of the elbow diameter and sand rate and size. Edwards et al. [15] reported the eect of the
bend angle on the normalized penetration rate. The objective of most of these experimental studies
was to provide data for establishing a relationship between the amount of erosion and the physical
characteristics of the materials involved, as well as the particle velocity and angle of impact. Blanchard et al. [16] carried out an experimental study of erosion in an elbow by solid particles entrained
in water. The elbow was examined in a closed test loop. Electroplating the elbow surface and photographing after an elapsed period of time were carried out to show the wear pattern. The theoretical model developed by Rabinowicz [17] was used to calculate the volume of material removed.
The results indicated that the sand particle trajectories appeared to be governed by the secondary
ows and that there was no simple liquid velocity prole that can be used to calculate the particle
trajectories in order to make an accurate prediction of the location of the point of maximum wear.
Several erosion models/correlations were developed by many researchers to provide a quick answer to design engineers in the absence of a comprehensive practical approach for erosion prediction. One of the early erosion prediction correlations is that developed by Finnie [18] expressing
the rate of erosion in terms of particle mass and impact velocity. In that correlation, the rate of
erosion was proportional to the impact velocity squared. In a recent study, Nesic [19] found that
Finnies model overpredicts the erosion rate and presented another formula for the erosion rate in
terms of a critical velocity rather than the impact velocity. One of the early erosion models was
that suggested by Bitter [20,21]. In that model, the erosion was assumed to occur in two main
mechanisms; the rst was caused by repeated deformation during collisions that eventually results
in the breaking loose of a piece of material while the second was caused by the cutting action of
the free-moving particles. Comparisons between the obtained correlations and the test results
showed a good agreement. It was concluded that cutting wear prevails in places where the impact
angles are small (such as in risers and straight pipes) and it is sucient to use hard material in such
places to reduce erosion. Tilly [1] suggested another two-stage mechanism for explaining dierent
aspects of the erosion process for ductile materials. In the rst stage, the particles indent the target
surface, causing chips to be removed and some material to be gouged and extruded to form vulnerable hillocks around the scar. The second stage was the one in which the particles break up on
impact causing fragments to be projected radially to produce a secondary damage. A correlation
was presented relating erosion to the energy required to remove a unit mass and the particle velocity and size. The calculated values of erosion were compared with the experimental data for different particle sizes and a reasonable agreement was found, however, the validity of the work was
limited to ductile materials and could not be generalized to include other materials. Other erosion
models were suggested by Laitone [22], Salama and Venkatesh [23], Bourgoyne [24], Chase et al.
[25], McLaury [26], Svedeman and Arnold [27], and Jordan [28].
Recently, Shirazi and McLaury [29] presented a model for predicting multiphase erosion in
elbows. The model was developed based on extensive empirical information gathered from many
sources, and it accounts for the physical variables aecting erosion, including uid properties,
sand production rate and size, and the uid-stream composition. An important dierent feature
of this model was the use of the characteristic impact velocity of the particles. The method used
for obtaining this characteristic velocity for an elbow was an extension of a previous method
introduced by the same authors for the case of a single-phase ow. The results from the model
were compared with previous experimental results for elbows and were found to have a better
agreement with eld failure data.

725

## The use of computational methods in erosion prediction constitutes a combination of ow

modeling, Lagrangian particle-tracking, and the use of erosion correlations. The ow model is
used to determine the ow eld for a given geometry while the particle-tracking model is used
to determine the particle trajectories for solid particles released in the ow. The particle impingement information extracted from the trajectories is used along with the empirical erosion equations to predict the erosion rates. This model, which is sometimes called the Lagrangian
approach, requires expertise in uid dynamic modeling and a large amount of computational
work. Boulet et al. [30] conducted numerical solutions for the turbulent ow of an airsolid suspension in a heated vertical pipe using EulerianEulerian and EulerianLagrangian formulations.
The main task was to assess the accuracy of these two formulations, taking the experimental data
reported by Tsuji et al. [31] and Jepson et al. [32] as a base for comparison. The rst part of the
pipe contains developing ow with no heat transfer. In the second part of the pipe, the dynamically fully developed ow was heated using a heated section of the pipe (constant heat ux). The
simulation was carried out for dierent values of mass loading with particles of 500 lm diameter.
The comparisons with experimental data for the dynamic features of the ow showed the same
accuracy level for both formulations, especially for dilute ows. However, the accuracy was found
to decrease signicantly in both formulations as more particles were injected in the ow.
Lagrangian models were developed by many researchers such as Lu et al. [33], Wang et al. [34],
Keating and Nesic [35] and Wallace et al. [36] who used combinations of computational uid
dynamics and dierent Lagrangian particle-tracking models to predict the particle movement
through complex geometries. Dierent computational uid dynamic packages such as PHOENICS [35,37] and CFX-code [38] were used to predict the uid ow eld. Wang et al. [34] developed a computational model for predicting the rate of erosive wear in a 90 elbow for the two
cases of sand in air and sand in water. The ow eld was rst obtained and then the particle trajectory and impacting characteristics were then determined by solving the equation of particle motion taking into consideration all the forces including drag, buoyancy, and virtual mass eects
with the assumption of a uniform distribution of the solid particles at the starting section. The
penetration rate was obtained using a semi-empirical relation that was previously developed by
Ahlert [39]. A comparison between the predicted penetration rates and the available experimental
data showed a good agreement.
In a recent study by Edwards et al. [15], an erosion prediction procedure was developed and
veried based on a CFD code combining ow eld analysis and particle-tracking for obtaining
particle impingement data. The erosion rate was then computed using the empirical relations of
Ahlert [39] and applied to predict erosion in a pipe bend tting made of carbon steel. The
CFD code utilized a nite-volume multiblock approach for solving NavierStokes equations
based on a user-dened computational model that was described by Patankar [40]. The authors
used the Lagrangian particle-tracking algorithm of the CFD code for the prediction of individual
trajectories of the dispersed phase through the ow eld.
Based on the above literature search and to the best of the authors knowledge, most of the published work on erosion in pipes focused on straight pipes and pipe ttings such as bends and
elbows. Apart from the work of Nesic [19], Postlethwaite and Nesic [41] and Blatt et al. [42],
the erosion process occurring in a pipe with sudden contraction or sudden enlargement was
not considered in any previous study. Postlethwaite and Nesic [41] and Blatt et al. [42] provided
experimental data for erosion inside a pipe with sudden contraction and sudden enlargement. The

726

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

present research work aims at studying the eect of uid ow parameters on the rate of erosion in
a pipe contraction under conditions simulating the actual working conditions. The calculations of
the ow pattern and solid particle motion inside the pipe contraction were performed and the
available data in the literature were used for estimating the rate of erosion. The computational
procedure is validated against the results of Postlethwaite and Nesic [41].

## 2. Problem statement and solution methodology

The ow domain consists of a straight pipe of diameter, D = 200 mm, connected to a smaller
pipe of diameter, d, as shown in Fig. 1 with diameter ratio, d/D = 0.5. The pipe centerline is always
vertical while the direction of uid ow is either vertically upward or vertically downward. Both
pipes are made of carbon steel and both are long enough to justify the assumption of fully developed ow at the entrance and exit sections of the ow domain. The uid considered in this study is
water at 20 C with low particle concentration such that the eect of particle motion on the uid
ow eld is negligibly small.
In general, the rate of erosion in tubes depends upon many parameters such as the properties of
the impacting particles, the properties of the tube material, and the other parameters of the impact
process [1,2,43]. In this study, the main parameters aecting erosion are the ow velocity and particle size and concentration. In order to predict the rate of erosion, the ow eld characteristics
and the details of the particle impact process in addition to the erosion rate correlations are required. The Lagrangian particle-tracking method is used to model the erosion process and is normally carried out using the following steps [36]:
(a) Predict the ow velocity eld in the domain of interest.
(b) Calculate the trajectories of solid particles entrained in the uid using Lagrangian particletracking calculations and then extract the particle impact data.
(c) Predict the erosive wear using one of the available semi-empirical correlations.

Fig. 1. Flow passage geometry for the two cases of upow and downow.

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

727

This approach represents a one-way ow-to-particle coupling method that can be used when
low volume of particles is simulated. Two computational models were developed; the rst is
the continuous phase model (dealing with the prediction of the ow velocity eld) and the second
is the particle-tracking model (dealing with the prediction of particle motion). A brief discussion
of the two models is presented in the following sections.
2.1. The continuous phase model
A combination of computational uid dynamics and Lagrangian particle tracking are normally
used to predict the particle movement through complex geometries [34,35,15,36]. To predict the
ow pattern of the continuous ow phase, the conservation equations for mass and momentum
are solved. Additional transport equations for the turbulence model are also solved since the ow
is turbulent. The time-averaged governing equations of 3-D turbulent ow can be found in many
Refs. [44,45] and can be presented as follows.
2.1.1. The continuity and momentum equations
The steady state time-averaged conservation equations of mass and momentum can be written
as
o
qU j 0
oxj



o
op
o
oU i
o
qU i U j 

l
qui uj

oxj
oxi oxj
oxj
oxj

where p is the static pressure and the stress tensor qui uj is given by
 

oU i oU j
2
qui uj leff

 qkdij
3
oxj
oxj

where dij is the Kronecker delta and le = lt + l is the eective viscosity. The turbulent viscosity,
lt, is calculated using the high-Reynolds number form as
k2
4
e
with Cl = 0.0845, k and e are the kinetic energy of turbulence and its dissipation rate. These are
obtained by solving their conservation equations as given below.
lt qC l

## 2.1.2. Conservation equations for the turbulence model

The conservation equations of the turbulence model (Reynolds [46] and Shih et al. [47]) are
given as follows:


o
o leff ok
Gk  qe
qU j k
5
oxj
oxj rk oxi

728

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742



o
o leff oe
e
e2
qU j e
C 1 Gk  C 2 q
oxj
oxi re oxi
k
k

where Gk represents the generation of turbulent kinetic energy due to the mean velocity gradients
and is given by
oU j
7
Gk qui uj
oxi
The quantities rk and re are the eective Prandtl numbers for k and e, respectively and C 2 is given
by Shih et al. [47] as
C 2 C 2 C 3

where C3 is a function of the term k/e and, therefore, the model is responsive to the eects of rapid
strain and streamline curvature and is suitable for the present calculations. The model constants
C1 and C2 have the values; C1 = 1.42 and C2 = 1.68.
The wall functions establish the link between the eld variables at the near-wall cells and the
corresponding quantities at the wall. These are based on the assumptions introduced by Launder
and Spalding [48] and have been most widely used for industrial ow modeling. The details of the
wall functions are provided by the law-of-the-wall for the mean velocity as given by Habib et al.
[44].
2.1.3. Boundary conditions
The velocity distribution is considered fully developed q
at
the inlet
section. Kinetic energy and its

dissipation rate are assigned through a specied value of k=U equal to 0.1 and a length scale, L,
equal to the diameter of the inlet pipe. The boundary condition applied at the exit section is that
of fully developed ow. At the wall boundaries, all velocity components are set to zero in accordance with the no-slip and impermeability conditions. Kinetic energy of turbulence and its dissipation rate are determined from the equations of the turbulence model.
2.1.4. Solution procedure
The solution domain was rst divided into a large number of nite volumes (at least 30,000
nite volumes). All meshes are of structured type and have ne meshing at the contraction
down to a minimum rectangular cell size of 0.1 mm (in the radial direction) 0.15 mm (in the
axial direction). This ne mesh is necessary to capture the steep velocity gradients close to
the contraction. The conservation equations are integrated over every nite volume to yield
the details of the velocity eld. The equations are solved simultaneously using the solution procedure described by Patankar [40]. Convergence is considered when the maximum of the summation of the residuals of all the elements for U, V, W and pressure correction equations is less
than 0.1%. The grid independence tests were performed by increasing the number of control
volumes from 7000 to 43,750 in two steps; 700030,000 and 30,00043,750. The results of rening the grids are shown in Fig. 2a and b for the axial and radial velocity proles. The inuence
of rening the grid on the continuous-phase velocity eld is very negligible and indicates that
more mesh renement will result in negligible changes in the results of the computational
model.

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

729

20.0
18.0
16.0
Axial velocity, m/s

14.0
12.0

10.0

## Mesh 1, 7000 volumes

Mesh 3, 43750 volumes

8.0
6.0
4.0
2.0

-0.30

-0.20

(a)

0.0
0.00

-0.10

0.10

0.20

0.30

0.025

0.020

## Mesh 1, 7000 volumes

Mesh 2, 30000 volumes
Mesh 3, 43750 volumes

0.015

0.010

0.005

0.000
-2.0

(b)

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

## Axial velocity, m/s

Fig. 2. Grid independence tests. (a) Comparison of axial velocity along the tube axis for dierent mesh sizes.
(b) Comparison of axial velocity along the radius at 1 mm upstream of the contraction for dierent mesh sizes.

2.2. Particle-tracking
The particle-tracking calculations aim to determine the particle trajectory from the moment it
enters the ow domain until it leaves the small tube. Of special interest is the particle velocity
(magnitude and direction) before every impact either on the pipes walls or anywhere on the tube
sheet. Such impact velocity is not only important for the calculation of solid surface erosion but
also important in the determination of the particle trajectory during its subsequent course of motion following impact. One of the main assumptions in this study is that the solid particles are not
interacting with each other (the particles do not collide and the motion of any particle is not inuenced by the presence or motion of neighboring particles). Moreover, the inuence of particle motion on the uid ow eld is considered very small and can be neglected. These two assumptions
are based on the condition of fairly dilute particle concentration. The same assumptions were

730

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

made by Lu et al. [33], Shirazi et al. [49], Edwards et al. [15], Keating and Nesic [35] and Wallace
et al. [36] in the solution of similar problems of low particle concentration (<23% by weight).
Taking the main hydrodynamic forces into consideration, the particle equation of motion can
be written as:
dup
F D u  up gqp  q=qp F vm F pg F sl
dt

where FD(uup) is the drag force per unit particle mass and F D 3C D lRep =4qp D2p , gqp  q=qp is
the buoyancy force term, Fvm is the virtual mass term (force required to accelerate the uid surrounding the particle), Fpg is the pressure gradient term and Fsl is the Saman lift force [50]. The
Magnus lift force (resulting from particle rotation) and the Basset history force (the force accounting for the ow eld unsteadiness) have been neglected. The particle Reynolds number, Rep, and
the drag coecient, CD, are obtained from
Rep

qDp jup  uj
l

C D a1

a2
a3

Rep R2ep

10
11

where the as are constants given by Morsi and Alexander [51] for smooth spherical particles over
several ranges of Re. Another equation that is frequently used for CD [52] is given by
CD

24
b3 Rep
1 b1 Rbep2
b4 Rep
Rep

12

where b1, b2, b3 and b4 are constants that depend on the particle shape.
Because of the low particle concentration assumed in the present study, the particle motion is
considered non-interacting and the dominant force in Eq. (9) is the drag force [15]. Some of the
other forces given in Eq. (9) are of small order of magnitude and can be neglected in this study.
The rst of these is the virtual mass term that takes care of the force required to accelerate the
uid surrounding the particle. This term can be expressed as
F vm

1 q d
u  up
2 qp dt

13

Although the virtual mass force, Fvm, is only important when q > qp which is not the case in the
present study, the virtual mass force was considered in the present calculations. The second force
is that due to pressure gradient, Fpg, that arises from the inuence of the pressure gradient in the
ow which acts on every volume element of the owing medium and can be written as:
!
q
F pg
rp
14
qp
The above statement implies that the pressure does not vary signicantly over a distance of one
particle diameter, a condition that is normally satised for reasonably small particles. Accordingly,
the pressure gradient force is neglected in the present study not only due to the small size of the

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

731

particles but also due to the small pressure gradient prevailing in the ow eld. The other forces
include the thermophoretic force which is related to small particles suspended in a gas that has
a temperature gradient. The particles under such circumstances experience a force in the direction
opposite to that of the gradient. Brownian force [53] apply for sub-micron particles. These forces
are neglected in the present study. The Samans lift force, or lift due to shear is also neglected.
2.2.1. Particle trajectory
The particle velocity, up, is rst obtained by stepwise integration of the particle equation of motion (9) over a discrete time step. The particle trajectory is then predicted by integrating the equation
dr
15
up
dt
where r is the position vector. The above equation is integrated in each coordinate direction to
predict the trajectories of the discrete phase. During the integration, the uid phase velocity, u,
is taken as the velocity of the continuous phase at the particle position. Turbulent dispersion
of particles was modeled using a stochastic discrete-particle approach, Wallace et al. [36]. The
tracking for the particle is done with a step size of 0.1 mm to make sure that the particle-tracking
is updated in every cell in the particle path. Including all the important forces (gravity, virtual
mass and pressure gradient) the particle-tracking is expected to be as accurate as the prediction
of the uid ow eld.
2.2.2. Discrete phase boundary conditions
The boundary conditions considered when a particle strikes a boundary surface depends on the
nature of that surface and one of the following possibilities may occur:
(a) Reection via an elastic or inelastic collision
Reection is the term used to describe the particle rebound o the solid boundary with a change in
its momentum. The normal coecient of restitution denes the amount of momentum in the direction normal to the wall that is retained by the particle after colliding with the boundary [54]. The
coecient of restitution is taken as 0.9 in the present calculations for the case of reection at a wall.
(b) Escape through the boundary
The calculations of the particle trajectory are terminated at the point when it passes through an
open boundary (the exit section). When the particle encounters such boundary, it is considered
that the particle has escaped and the trajectory calculations are then terminated.
(c) Particle trapping
The trajectory calculations for some particles (normally very few particles) are terminated when
the particles get trapped in the ow eld. This is found to occur when a particle circulates in a
conned ow zone. In such a case, the trajectory calculations are terminated.

## 2.3. The erosion model

Erosion is dened as the wear that occurs when solid particles entrained in a uid stream strike
a surface. The previous experimental results [43,55] show that the erosive wear-rate exhibits a

732

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

power-law velocity dependence. The velocity exponent ranges from 1.9 to 2.5. The results also
indicate that the erosion rate is a function of the angle of impact. It is shown that the inuence
of the angle of impact depends greatly on the type of material being brittle or ductile. Prediction
of erosion in straight pipes, elbows and tees show the strong inuence of uid properties, sand size
and ow velocity on the rate of erosion [56,49,41].
There have been many attempts in the past to express the solid particle erosion by an analytical
formula that could be used to predict erosion under any condition. The complexity of the erosion
process and the number of factors involved made it dicult for obtaining a generally applicable
equation. Almost all of the formulae generated have therefore some degree of dependence on
empirical coecients provided by various experimental erosion tests. No denitive theory of erosion currently exists, however, a number of qualitative and quantitative models do exist. These
were described by Finnie [18] and Finnie et al. [57], Wang et al. [34], Keating and Nesic [35], Edwards et al. [15] and Shirazi and McLaury [29].
The empirical erosion equations suggested by Neilson and Gilchrist [58] were later used by Wallace et al. [36] to correlate the experimental erosion data in order to develop an erosion modeling
technique. Wallace et al. [36] reported the following formulae that resulted in good accuracy when
compared to the experimental data:
(
)
2
1 12 u2p cos2 a sin 2a 12 u2p sin a
a 6 45
16a

E
Np
c
r
1
E
Np

1 2
u cos2 a
2 p

1 2
u sin2 a
2 p

a > 45

16b

where c and r are the cutting wear and deformation wear coecients having the values 33316.9
and 77419.7 respectively [36]. This formula is used in the present calculations of erosion rate.
Using the particle-tracking model, the impingement data (impact speed and angle) were rst
compiled for all particles impacting the solid boundaries of the ow domain. The compiled data
were then used together with Eqs. (16a) and (16b) for computing the erosion rate at dierent locations on the tube sheet. This part required the use of FORTRAN subroutines together with the
CFD code.

## 3. Results and discussion

The rate of erosion in a pipe with sudden contraction has been investigated for the case when
the diameter ratio, d/D = 1/2. The direction of ow is considered either vertically upward (against
gravity) or vertically downward (in direction of gravity) as shown in Fig. 1. Although the direction
of the pipe axis has no eect on the ow pattern, it is expected to aect the particle trajectory
through the contribution of gravity forces on the particle motion. The upstream pipe diameter
is 200 mm and the average velocity of the approaching ow ranges from 1.0 to 10.0 m/s. The uid
considered is water at 20 C (q = 998 kg/m3 and l = 103 N s/m2) which results in ow Reynolds
number (based on the diameter of the large pipe, D) ranging from 2 105 to 2 106. The solid

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

733

particles are considered sand particles of spherical shape with diameters ranging from 10 lm to
400 lm.
In order to verify the accuracy of the computational scheme, the present results were compared
to the experimental data of Postlethwaite and Nesic [41]. The experimental data were obtained for
contraction ratio of d/D = 0.5. The large tube is 42.1 mm diameter and has an inlet velocity of
3.3 m/s. Three values of particle concentration of 2%, 5% and 10%, by volume, were considered.
The sand particle diameter was 430 lm. Short 3-mm segments were used at the inlet region of the
sudden contraction to determine the penetration rate. In the present calculations, the step height
was divided to 1000 discs, then, data were integrated to give the results over a disc of 3 mm width
having its inner diameter as the smaller tube. The results of the comparison are given in Fig. 3 in
terms of the penetration rate. The penetration rate, Pn, is calculated using the following equation
[34,49]:
pn 31:536
106

s_
qm A

Elc

17

where A is the impingement area (m2), Elc is the local erosion rate (mg/g), Np is the total number
of particles being tracked, pn is the penetration rate (mm/year), s_ is the sand rate (kg/s) and qm is
the density of target material (kg/m3). The comparison shown in Fig. 3 indicates a reasonably
good agreement.
To present the obtained data in a meaningful way, a number of investigations were carried out
with the objective to determine the critical erosion areas. These investigations covered the entire
ranges of ow velocity and particle diameter. It was found that erosion occurs mainly in the contraction section ABCD shown in Fig. 1 while being insignicant upstream and downstream of it.
Fig. 4 shows the trajectories of a number of particles released at the same time at the inlet section
of the ow eld when the ow velocity is 10 m/s and the particle diameter is 400 lm. The gure
shows that almost all particle impacts occur on the at surface ABCD while impacts on the pipe
walls are insignicant. Accordingly, erosion data will be presented only at section ABCD.
Fig. 5 shows the variation of the local erosion rate on the tube sheet (ABCD) for the upow
case. This tube sheet has the shape of an annulus with inner radius, r/R = 0.5, and outer radius,
r/R = 1.0, where R = D/2. The erosion rates are obtained for four values of the particle diameter
(10 lm, 100 lm, 200 lm, 400 lm) and for three values of ow velocity (1 m/s, 5 m/s, 10 m/s) as
shown in Fig. 5ad. The results indicate that for particles of small diameter (Dp = 10 lm), the erosion rate is negligibly small in the outer region of the annular plate (0.67 6 r/R 6 1.0) and reaches
its maximum close to the entrance of the small pipe (r/R 0.5) as shown in Fig. 5a. The highest
rate of erosion (E 8 107 mg/g) was found when the velocity of ow is maximum (10 m/s) and
decreases rapidly with the decrease of ow velocity until reaching zero value when the ow velocity reaches 1 m/s. Although the rate of erosion increases with the increase of particle diameter as
shown in Fig. 5bd, the trend is almost the same in the four cases. However, for large particle size
(Dp = 400 lm), the region of negligible erosion for all ow velocities diminishes to (0.83 6
r/R2 6 1.0) which is much smaller than that obtained in the case of small particle sizes. The other
interesting feature that is common in the four gures is the absence of erosion for all particle sizes
in the entire ow domain in the case of low ow velocity (1 m/s). Qualitatively, such behavior is in
conformity with the erosion prevention criterion established by Salama [59] in which a threshold
velocity was set by the recommended practice API RP 14E for eliminating erosion. Another

734

1000

100

## Postlethwaite and Nesic [41]

Present calculations

10
0

10

12

## Particle volumetric percentage

Fig. 3. Comparison of the calculated penetration rate and the experimental data of Postlethwaite and Nesic [41].

Fig. 4. The trajectories of a number of particle released at the same time at the inlet section showing impact on the
contraction plate for the case of downow with Vi = 10 m/s and Dp = 400 lm.

similar criterion for the threshold velocity was developed by Salama and Venkatesh [23] for erosion in elbows.

9.0E-07

735

1.2E-06
1.0E-06

7.0E-07
1m/s

6.0E-07

## Erosion rate, mg/g

8.0E-07

5m/s

5.0E-07

10m/s

4.0E-07
3.0E-07
2.0E-07
1.0E-07
0.0E+00
0.7

(a)

10m/s

4.0E-07
2.0E-07

0.5

0.9

0.7

(b)

r/R

0.9
r/R

8.0E-06

4.0E-06
3.5E-06

1m/s

3.0E-06

5m/s

2.5E-06

10m/s

7.0E-06
Erosion rate, mg/g

## Erosion rate, mg/g

5m/s

6.0E-07

0.0E+00

0.5

2.0E-06
1.5E-06
1.0E-06
5.0E-07
0.0E+00

1m/s

6.0E-06

5m/s

5.0E-06

10m/s

4.0E-06
3.0E-06
2.0E-06
1.0E-06
0.0E+00

0.5

(c)

1m/s
8.0E-07

0.6

0.7

0.8
r/R

0.9

0.5

1.0

(d)

0.7

0.9

r/R

Fig. 5. The variation of the local erosion rate on the contraction plate (ABCD) for the case of upow: (a) Dp = 10 lm,
(b) Dp = 100 lm, (c) Dp = 200 lm, (d) Dp = 400 lm.

The variation of the local erosion rate presented in Fig. 4 can be explained on the basis of the
streamline pattern plotted in Fig. 6a for the case when the ow velocity is 5 m/s. The gure shows
a recirculating ow region upstream of the contraction section and extending to the tube sheet
(ABCD). An enlarged view of that region is shown in Fig. 6b. The ow velocity in this region
is very small and the presence of solid particles, if any, in such low velocity zone will cause negligible erosion in accordance with Eqs. (16a) and (16b). The gure also shows that this recirculating ow zone occupies the area on the annular plate between r/R 0.72 and r/R = 1.0. This is
approximately the same area characterized by negligible erosion in Fig. 4ad. Moreover, the maximum erosion rate occurs in a region where the approaching ow has high velocity and large curvature. Both eects will give rise to higher particle velocity that impacts the surface of the tube
sheet close to r/R 0.5. These features are conrmed by the particle trajectories given in Fig. 4
that clearly shows the high intensity of particle impact on the tube sheet in the region close to
r/R 0.5.
The eect of particle diameter on the total rate of erosion occurring on the tube sheet (ABCD)
is presented in Fig. 7 for four dierent velocities. The strong dependence of erosion on ow velocity is very clear in the gure. It is also clear that there is a threshold velocity, Vt, below which
erosion is insignicant. The gure also shows that the rate of erosion increases exponentially with

736

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

Fig. 6. (a) The streamline pattern for the case of Vi = 5 m/s. (b) An enlarged view of the circulatory ow zone at the
contraction regioncase of Vi = 5 m/s.

2.5E-05
1m/s
Erosion rate, mg/g

2.0E-05

3m/s
5m/s
10m/s

1.5E-05

1.0E-05

5.0E-06

0.0E+00
0

50

100

150

200

250

300

350

400

Particle diameter, m

Fig. 7. Eect of particle diameter on the total rate of erosion occurring on the tube sheet for dierent inlet velocities in
the case of upow.

particle diameter. Fig. 8 represents the same data plotted in Fig. 7, however, the inlet ow velocity
is used as abscissa instead of the particle diameter. The gure emphasizes the power law growth of

737

2.5E-05
10 m
100 m

## Erosion rate, mg/g

2.0E-05

200 m
400 m
1.5E-05

1.0E-05

5.0E-06

0.0E+00
0

4
5
6
Inlet flow velocity, m/s

10

Fig. 8. Eect of inlet ow velocity on the total rate of erosion occurring on the tube sheet for dierent particle diameters
in the case of upow.

1.2E-06

8.0E-07

1m/s

7.0E-07

5m/s

6.0E-07

10m/s

5.0E-07
4.0E-07
3.0E-07
2.0E-07

5m/s

8.0E-07

4.0E-07
2.0E-07
0.0E+00

0.0E+00
0.5

0.6

0.7

(a)

0.8

0.9

0.5

1.0

0.7

(b)

r/R

0.9

r/R

8.0E-06

3.5E-06

1m/s

3.0E-06

5m/s

2.5E-06

10m/s

2.0E-06
1.5E-06
1.0E-06

## Erosion rate, mg/g

4.0E-06
Erosion rate, mg/g

10m/s

6.0E-07

1.0E-07

7.0E-06

10m/s

4.0E-06
3.0E-06
2.0E-06
1.0E-06
0.0E+00
0.5

0.9

r/R

5m/s

5.0E-06

0.0E+00
0.7

1m/s

6.0E-06

5.0E-07
0.5

(c)

1 m/s

1.0E-06
Erosion rate, mg/g

## Erosion rate, mg/g

9.0E-07

(d)

0.6

0.7

0.8

0.9

1.0

r/R

Fig. 9. The variation of the local erosion rate on the contraction plate (ABCD) for the case of downow:
(a) Dp = 10 lm, (b) Dp = 100 lm, (c) Dp = 200 lm, (d) Dp = 400 lm.

738

2.5E-05
1m/s

## Erosion rate, mg/g

2.0E-05

3m/s
5m/s
10m/s

1.5E-05

1.0E-05

5.0E-06

0.0E+00
0

50

100

150

200

250

300

350

400

Particle diameter

Fig. 10. Eect of particle diameter on the total rate of erosion occurring on the tube sheet for dierent inlet velocities in
the case of downow.

the rate of erosion with the increase of inlet ow velocity. Moreover, the threshold velocity, Vt,
can be approximately determined from Fig. 7 since the rate of erosion is insignicant for ow
velocities below 2 m/s for all particle diameters.
The erosion rates obtained for the downow case are presented in Fig. 9 for the same particle
diameters and inlet ow velocities. The results are almost the same as those obtained in the upow
case except in Fig. 9d (Dp = 400 lm) that shows higher rate of erosion ( 50% increase) at a ow
velocity of 5 m/s. It is quite expected that the eect of gravity on particle motion becomes significant at low ow velocities. However, such eect did not inuence the rate of erosion at the lowest
ow velocity (1 m/s) because such velocity is considerably below the threshold velocity, Vt. On the
other hand, the ow velocity of 5 m/s is denitely above the threshold velocity (see Fig. 8) and the
2.5E-05

10 m

2.0E-05

100 m
200 m
400 m

1.5E-05

1.0E-05

5.0E-06

0.0E+00
0

10

## Inlet flow velocity, m/s

Fig. 11. Eect of inlet ow velocity on the total rate of erosion occurring on the tube sheet for dierent particle
diameters in the case of downow.

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

739

eect of gravity becomes sensible. A quick comparison of the data presented in Figs. 5 and 9
shows that the eect of gravity on the rate of erosion is very small in the case of high inlet ow
velocity (10 m/s) for all particle sizes. This can be explained based on the fact that the relative contribution of gravity to the motion of solid particles gets smaller with the increase of ow velocity.
Fig. 10 shows the variation of the total erosion rate at the contraction section with particle size for
dierent ow velocities. Although the trends are the same as in Fig. 7 the values obtained are
slightly dierent especially in the case of moderate ow velocity (5 m/s) and large particle size
(Dp = 400 lm). The same data is presented in Fig. 11, however, the inlet ow velocity is used as
abscissa instead of the particle diameter. The gure shows that the threshold inlet velocity is
approximately the same as that in the case of upow ( 2 m/s). To clearly show the eect of ow
direction on the rate of erosion, Fig. 12a and b are plotted with inlet ow velocity as abscissa and
8.0E-06

7.0E-06
6.0E-06
5.0E-06
4.0E-06
3.0E-06
upward flow

2.0E-06

downward flow
1.0E-06
0.0E+00
0

(a)

10

12

2.50E-05

2.00E-05

1.50E-05

1.00E-05
upflow

5.00E-06

downflow

0.00E+00
0
(b)

10

12

## Inlet flow velocity, m/s

Fig. 12. Eect of ow direction and inlet velocity on the total rate of erosion for the two cases of (a) Dp = 200 lm, (b)
Dp = 400 lm.

740

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

the total rate of erosion as ordinate for the two particle sizes of 200 lm and 400 lm respectively. It
is clear from the gure that the eect of ow direction is only appreciable at the moderate velocity
of 5 m/s for the large particle size (400 lm) as explained earlier.

4. Conclusions
The eects of ow velocity and particle size on erosion in a vertical pipe with sudden contraction were investigated for the special case of two-phase (liquid and solid) turbulent ow with low
particle concentration. The ow was either in direction of gravity (downow) or against it (upow). The investigation follows the Lagrangian approach in which two mathematical models were
used for the determination of the uid velocity eld and the solid particle trajectory and an erosion model was used to predict the erosion rate. The investigation was limited to one diameter
ratio for the pipe contraction and one uid. The ow velocity in the large pipe ranged from
1 m/s to 10 m/s and the particle size ranged from 10 lm to 400 lm. In these ranges, the results
showed the strong dependence of erosion on both particle size and ow velocity but with little
dependence on the direction of ow. The eect of ow direction was found to be signicant only
for large particle size (400 lm) and moderate ow velocity (5 m/s). The erosion critical area was
found to be the inner surface of the tube sheet (connecting the two pipes) in the region close to
the small pipe inlet. The results also indicated the presence of a threshold velocity of approximately 2 m/s below which erosion is insignicant for all particle sizes.

Acknowledgement
The authors wish to acknowledge the support received from King Fahd University of Petroleum and Minerals during this study.

References
[1] Tilly GP. A two-stage mechanism of ductile erosion. Wear 1973;23:8796.
[2] Ru AW, Wiederhorn SM. Erosion by solid particle impact. In: Preece CM, editor. Treatise on materials science
and technology, vol. 16. New York: Academic Press; 1979. p. 69125.
[3] Humphrey JAC. Fundamentals of uid motion in erosion by solid particle impact. Int J Heat Fluid Flow
1990;11(3):17095.
[4] Rochester MC, Brunton JH. Inuences of physical properties of the liquid on the erosion of solids. Erosion, wear
and interfaces with corrosion ASTM STP 567. American Society of Testing and Materials; 1974. p. 12851.
[5] True ME, Weiner PD. A laboratory evaluation of sand erosion of oil and gas well production equipment. Annual
API production division meeting. Los Angeles, CA: 1976. p. I-1, I-27.
[6] Glaeser WA, Dow A. Mechanisms of erosion in slurry pipelines. In: Proceedings of the second international
conference on slurry transportation. Las Vegas, NV: March 24, 1977. p. 13640.
[7] Roco MC, Nair P, Addie GR, Dennis J. Erosion of concentrated slurries in turbulent ow. J Pipelines
1984;4:21321.
[8] Venkatesh ES. Erosion damage in oil and gas wells, SPE paper 15183. Rocky mountain regional meeting of the
society of petroleum engineers. Billings, MT: 1986.

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

741

[9] Shook CA, Mckibben M, Small M. Experimental investigation of some hydrodynamics factors aecting slurry
pipeline wall erosion. ASME paper no. 87-PVP-9, 1987.
[10] Soderberg S, Hogmark S, Swahn H. Mechanisms of material removal during erosion of a stainless steel, Paper no.
82-AM-4A-1. 37th ASLE Annual Meeting, Cincinnati, May 1013, 1982.
[11] Soderberg S, Hogmark S, Engman U, Swahn H. Erosion classication of materials using a centrifugal erosion
tester. Tribol Int 1981:33344.
[12] Hutchings IM. Monograph on the erosion of materials by solid particle impact. Materials Technology Institute of
Chemical Process Industries, Inc.; MTI publication no. 10, 1983.
[13] Hutchings IM. Introduction to the microscopy of erosion. J Microscopy 1983;130:3318.
[14] McLaury BS, Wang J, Shirazi SA, Shadley JR, Rybicki EF. Solid particle erosion in long radius elbows and
straight pipes. Society of Petroleum Engineers, Paper no. SPE 38842, 1997. p. 97786.
[15] Edwards JK, McLaury BS, Shirazi SA. Evaluation of alternative pipe bend ttings in erosive service. Proceedings
of 2000 ASME uids engineering summer meeting, June 1115, 2000. Boston, MA: Paper no. FEDSM2000-11245,
2000.
[16] Blanchard DJ, Grith P, Rabinowicz E. Erosion of a pipe bend by solid particle entrained in water. J Eng Indust
1984;106:2137.
[17] Rabinowicz E. The wear equation for erosion of metals by abrasive particles, Department of ME, MIT: 1979.
[18] Finnie I. The mechanism of erosion of ductile metals. In: Proceedings of 3rd US national congress of applied
mechanics. 1958. p. 52732.
[19] Nesic S. Computation of localized erosion-corrosion in disturbed two-phase ow, PhD thesis, University of
[20] Bitter JGA. A study of erosion phenomena, Part I. Wear 1963;6:521.
[21] Bitter JGA. A study of erosion phenomena, Part II. Wear 1963;6:16990.
[22] Laitone JA. Erosion prediction near a stagnation point resulting from aerodynamically entrained solid particles.
J Aircraft 1979;16(12):80914.
[23] Salama MM, Venkatesh ES. Evaluation of erosional velocity limitations of oshore gas wells. 15th Annual OTC.
Houston, TX: May 25, OTC no. 4485, 1983.
[24] Bourgoyne AT. Experimental study of erosion in diverter systems due to sand production, Presented at the SPE/
[25] Chase DP, Rybicki EF, Shadley JR. A model for the eect of velocity on erosion of N80 steel tubing due to the
normal impingement of solid particles. Trans ASME J Energy Resour Technol 1992;114:5464.
[26] McLaury BS. A model to predict solid particle erosion in oil eld geometries. MS thesis, The University of Tulsa,
1993.
[27] Svedeman SJ, Arnold KE. Criteria for sizing Multiphase ow lines for erosive/corrosive services. Paper presented
at the 1993 SPE conference, Houston SPE 26569, 1993.
[28] Jordan K. Erosion in multiphase production of oil and gas. Corrosion 98, Paper no. 58, NACE International
Annual Conference, San Antonio: 1998.
[29] Shirazi SA, McLaury BS. Erosion modeling of elbows in multiphase ow. In: Proceedings of 2000 ASME uids
engineering summer meeting, June 1115, 2000. Boston, MA: Paper no. FEDSM2000-11251, 2000.
[30] Boulet P, Oesterle B, Andreux R. Comparisons between EulerianEulerian and EulerianLagrangian formulation
of a gassolid suspension ow in a heated pipe. In: Proceedings of the 1999 ASME uids engineering division
summer meeting. San Francisco, CA: FEDSM99-7860, 1999.
[31] Tsuji Y, Morikawa Y, Shiomi H. LDV measurements of an airsolid two-phase ow in a vertical pipe. J Fluid
Mech 1984;139:41734.
[32] Jepson G, Poll A, Smith W. Heat transfer from gas to wall in a gas/solid transport line. Trans Inst Chem Engrs
1963;41:20711.
[33] Lu QQ, Fontaine JR, Aubertin G. A Lagrangian model for solid particles in turbulent ows. Int J Multiphase
Flow 1993;19(2):34767.
[34] Wang J, Shirazi SA, Shadley JR, Rybicki EF. Application of ow modeling and particle tracking to predict sand
erosion rates in 90-degree elbows. FED-vol. 236, 1996 ASME uids engineering division conference, vol. 1. 1996. p.
72534.

742

## H.M. Badr et al. / Computers & Fluids 34 (2005) 721742

[35] Keating A, Nesic S. Particle tracking and erosion prediction in three-dimensional bends. In: Proceedings of 2000
ASME uids engineering summer meeting, June 1115, 2000. Boston, MA: Paper no. FEDSM2000-11249, 2000.
[36] Wallace MS, Peters JS, Scanlon TJ, Dempster WM, McCulloch S, Ogilvie JB. CFD-based erosion modeling of
multi-orice choke valves. In: Proceedings of 2000 ASME uids engineering summer meeting, June 1115, 2000.
Boston, MA: Paper no. FEDSM2000-11244, 2000.
[37] Hanson R, Patel MK. Development of a model to predict the life of pneumatic conveyor bends subject to erosive
wear. In: Proceedings of 2000 ASME uids engineering summer meeting, June 1115, 2000. Boston, MA: Paper no.
FEDSM2000-11246, 2000.
[38] Forder A, Thew M, Harrison D. A numerical investigation of solid particle erosion experienced within oileld
control valves. Wear 1998;216:18493.
[39] Ahlert KR. Eects of particle impingiment angle and surface wetting on solid particle erosion on ANSI 1018 steel,
MS thesis, University of Tulsa, USA, 1994.
[40] Patankar SV. Numerical heat transfer and uid ow. 1st ed. Taylor and Francis; 1980.
[41] Postletwaite J, Nesic S. Erosion in disturbed liquid/particle pipe ow; eects of ow geometry and particle surface
roughness. Corrosion 1993;49(10):8509.
[42] Blatt W, Kohley T, Lotz U, Heitz E. The inuence of hydrodynamics on erosion-corrosion in two-phase liquid
particle ow. Corrosion 1989;45(10):793804.
[43] Davies JE, Stead RJ, Andrews CJ, Richards JR. The airborne particle erosion resistance of a Range of engineering
materials. Key Eng Mater 1991;117:4552.
[44] Habib MA, Attya AE, McEligot DM. Calculation of turbulent ow and heat transfer in channels with streamwise
periodic ow. ASME J Turbomach 1989;110:40511.
[45] Versteeg HK, Malalasekera W. An introduction to computational uid dynamics; the nite volume method. London: Longman Scientic and Technical; 1995.
[46] Reynolds WC. Fundamentals of turbulence for turbulence modeling and simulation, Lecture notes for Von
Karman institute, Agard report no. 755, 1987.
[47] Shih TH, Liou WW, Shabbir A, Zhu J. A new ke eddy-viscosity model for high Reynolds number turbulent
owsmodel development and validation. Computers and Fluids 1995;24(3):22738.
[48] Launder BE, Spalding DB. The numerical computation of turbulent ows. Comput Meth Appl Mech Eng
1974;3:26989.
[49] Shirazi SA, Shadley JR, McLaury BS, Rybicki EF. A procedure to predict solid particle erosion in elbows and tees.
J Pressure Vessel Technol 1995;117:4552.
[50] Saman PG. The lift on a small sphere in a slow shear ow. J Fluid Mech 1965;22(2):385400.
[51] Morsi SA, Alexander AJ. An investigation of particle trajectories in two-phase ow systems. J Fluid Mech
1972;52(2):193208.
[52] Haider A, Levenspiel O. Drag coecient and terminal velocity of spherical and nonspherical particles. Powder
Technol 1989;58:6370.
[53] Li A, Ahmadi G. Dispersion and deposition of spherical particles from point sources in a turbulent channel ow.
Aerosol Sci Technol 1992;16:20926.
[54] Tabako W, Wakeman T. Measured particle rebound characteristics useful for erosion prediction. ASME paper
82-GT-170, 1982.
[55] Isomoto Y, Nishimura M, Nagahashi K, Matsumura M. Impact angle dependence of erosion by solid particle
impact for metallic materials. Erosion Eng 1999;48(6):35561.
[56] McLaury BS, Shirazi SA. Predicting erosion in straight pipes. In: Proceedings of the 1998 ASME uids engineering
division, FEDSM 98-5226, June 2125. Washington, DC: 1998.
[57] Finnie I, Stevick GR, Ridgely JR. The inuence of impingement angle on the erosion of ductile metals by angular
abrasive particles. Wear 1992;52:918.
[58] Neilson JH, Gilchrist A. Erosion by a stream of solid particles. Wear 1968;11:11122.
[59] Salama MM. An alternative to API 14E erosional velocity limits for sand-laden uids. 1998 Oshore Technology
Conference, Houston, TX, May 47, 1998. p. 72131.