You are on page 1of 11

Wear 256 (2004) 937947

Comparison of predicted and experimental erosion estimates


in slurry ducts
R.J.K. Wood
a,
, T.F. Jones
b
, J. Ganeshalingam
b
, N.J. Miles
b
a
Surface Engineering and Tribology Group, School of Engineering Sciences, University of Southampton, Higheld, Southampton SO17 1BJ, UK
b
School of Chemical, Environmental and Mining Engineering, The Nottingham Mining and Minerals Centre,
University of Nottingham, University Park, Nottingham NG7 2RD, UK
Received 22 April 2003; received in revised form 24 July 2003; accepted 9 September 2003
Abstract
Computational models for the impact velocity and impact angle in a bend have been successfully applied to the ow eld and validated
by electrical resistance tomography (ERT) to conrm the position of particle burdens. Particle impact parameters from this work have been
used as inputs to erosion models to predict wall wastage rates in a R
c
/D = 1.2 bend and a upstream straight pipe section. The location of
damage and the levels of wall wastage agree well with those obtained from erosion trials conducted on a full scale loop rig with AISI 304
stainless steel pipework. Well-distributed but asymmetric particulate ows of sand in water have been studied at average ow velocities of
3 ms
1
. Predicted erosion damage levels and locations are compared to non-destructive ultrasonic and gravimetric measurements as well
as wall thickness measurements made with a micrometer after cutting the pipe and bends sections.
2003 Elsevier B.V. All rights reserved.
Keywords: Erosion; Pipes; Bends; Sand slurry; Microcutting; Wall wastage; CFD; Models
1. Introduction
There is little disagreement that erosion damage in con-
ventional pipelines is costly and inconvenient. In the UK
alone, the Department of Trade and Industry in 2000 put ero-
sion costs at approximately 20 million per annum and Jost
[1] estimated wear-related gures as high as 1.5% of GNP.
The smooth interior prole of cylindrical ducts encour-
ages settling of particles from low-speed slurries. Designers
tend to specify a ow velocity which will entrain settling
particles into the ow but this is usually higher than needed
satisfactorily to transport the slurry. Pumping power require-
ments are high as is the propensity for components to wear.
Bends act like concave mirrors and reect the ow (with any
entrained particles) leading to localised wear. Fig. 1 illus-
trates the geometry of straight and curved ducts and the po-
sition of asymmetric particle burdens within the bore. Please
note the denitions of the pipe invert, intrados and extrados
as applied to the geometry of ducts, appear in the Glossary
and in Fig. 1. It is tempting to view the lower part of the
duct (pipe invert) as an obvious location for accelerated wear
in straight pipes and the outer extreme (extrados) for accel-

Corresponding author. Tel.: +44-1703594881; fax: +44-1703593230.


E-mail address: rjw3@soton.ac.uk (R.J.K. Wood).
erated wear in bends, but this is not always the case. The
discussion which follows sets the scene for a wider consid-
eration of particle positions and wear.
1.1. Particle positions and wear
Most published results refer to maximum wear at the bot-
tom of a pipe. Little information is available about the dis-
tribution or circumferential location of erosion modes. A
simple model for erosive wastage at a particular orientation,
, in the duct includes primarily a factor for the energy dis-
sipation of impacting particles plus friction and directional
impingement at that orientation [2]. There must also be a
multiplicative factor for the mean supply of impacting par-
ticles at that orientation (

particles min
1
). Factors such
as impact angle and velocity-based corrections for plastic
shear deformations are left for later discussion.
Asymmetric slurry ow can be modelled as two layers
following Wilson [3,4]:
a contact layer where moving particles are in continuous,
indirect or sporadic contact with the duct,
a suspended layer.
The principle of the supply of energetic particles, elabo-
rated above, leads us to consider three possible modal posi-
tions for wear:
0043-1648/$ see front matter 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.wear.2003.09.002
938 R.J.K. Wood et al. / Wear 256 (2004) 937947
Nomenclature
C
k
cutting characteristic velocity (ms
1
)
C
v
volume fraction
D internal diameter of bend (m)
D
k
modied deformation characteristic
velocity (ms
1
)
E
f
deformation erosion factor (J m
3
)
E
p
Youngs modulus for the particle (Nm
2
)
E
t
Youngs modulus for the target (Nm
2
)
Er erosion rate (m
3
min
1
)
L pipe length (m)
L
i
input leg of bend (m)
L
o
output leg of bend (m)
M total mass of particles
M
p
particle mass (kg)
M
L
target mass loss (kg)
n velocity ratio exponent

particle supply at a given position on


the interior surface of a duct, particles (s
1
)
q
p
Poissons ratio for particle
q
t
Poissons ratio for target
Q
v
volumetric ow rate of solids (m
3
s
1
)
r
p
particle radius (m)
R pipe radius (m)
R
c
centreline radius of bend (m)
R
f
roundness factor for particle (value 01)
U
p
particle velocity (ms
1
)
W
c
cutting wear rate (m
3
per impact)
W
d
deformation wear rate (m
3
per impact)
W
t
total wear rate (m
3
per impact)
Y yield stress of target (Nm
2
)
Greek letters
wall penetration (mm)

mean
mean penetration depth (mm)
difference in wall thickness (mm)
orientation of a point on the interior surface
of the bore of a duct (

i
peripheral orientation of interface between
contact and suspended layers (

p
density of particle (kg m
3
)

t
density of target (kg m
3
)
plastic ow stress for target (Nm
2
)
1.
i
: The boundary region between the contact and
suspended layers where particle velocities can be high
and the particle supply can also be high.
2. 0

: The base of the contact layer where particle


supply is relatively high but particle velocities, and
therefore erosion efciencies, are relatively low.
3. 180

: The top of the suspended layer where par-


ticle velocities are relatively high, but particle supply
can be extremely low.
Fig. 1. Asymmetric suspensions in straight sections and bends, showing
key circumferential positions.
Roco and Cader [2] show wear modes at the boundary
region when velocities are near or below the deposition ve-
locity, and at the base of the contact layer when velocities
equal or exceed the deposition limit. At higher slurry veloc-
ities, wear is distributed around the periphery of the bore, in
concert with the asymmetric distribution of particle supply.
In a loop operating just above the deposition velocity, the
local particle velocity and particle supply will vary about
the mean values. A mixture of wear modes is therefore to
be expected.
1.2. Slurry ow patterns
The pattern of slurry ow is important in determining
the kinetic energy and supply of particles at a point on the
periphery of a pipe. Slurry ow patterns are commonly
categorised into several identiable states of suspension or
partial settling. At one extreme is the stationary bed and
at the other is symmetrically suspended ow. Govier and
Aziz [5] subdivided the transition into ve main categories
while Chien [6] classied sand/water mixtures into an even
greater number of categories.
Williams and Beck [7] show that electrical resistance to-
mography (ERT) can be used to determine these patterns ex-
perimentally. Evenly spaced electrodes around the pipe and
in contact with the slurry are used to construct a grid of con-
ductivity measurements which can be used to plot contours
of equal slurry concentration. It is important to note that the
technique does not measure the density of each particle but
assigns a concentration to regions of the cross-section. For
studying the patterns of ow this presentation is ideal.
Flow pattern determinations were performed at the Uni-
versity of Nottingham using a straight duct and white beads
of median size 1.75 mm (0.257 at 1 s) and median relative
density of 1.45 (0.005). Measurements were made using a
commercial ITS-P2000 ERT system. Acurrent injection pro-
tocol based on excitation of adjacent electrodes using 10 kHz
alternating current was used. The gains of pre-ampliers for
all measurement were acquired by automatic calibrations.
Images were reconstructed using a sensitivity-coefcient
back-projection algorithm after Wang et al. [8]. The tomo-
graphic concentration curves were obtained from a ring of
16 electrodes which were evenly spaced around the periph-
ery of the pipe at the measurement plane. Concentration
fractions were obtained fromconductivity measurements be-
tween electrodes. Fig. 2 shows the concentration contours
obtained at three distinct ow patterns: a stationary bed in
R.J.K. Wood et al. / Wear 256 (2004) 937947 939
(a) 0.5 m/s (b) 1 m/s (c) 2 m/s
Fig. 2. ERT data for the ow of beads in a straight duct (2 mm beads, relative density 1.45, nominal in situ concentration 5% (v/v)): (a) stationary bed
pattern; (b) moving bed pattern; (c) asymmetric suspension.
which there was no discernable movement of packed beads
at the pipe invert, a moving bed in which settled beads were
progressing along the pipe and lastly an asymmetric suspen-
sion. The diagrams in Fig. 2 demonstrate the possible wear
modes mentioned above. The use of 16 electrodes caused the
grid to be fairly coarse. The slight misorientation of these
diagrams was found to be caused by digitisation effects.
1.3. Wear from ow of slurry
Slurry erosion is complex and good models have yet to be
developed. This is certainly true for erosion of pipelines and
is compounded by the fact that most published literature on
the wear of pipework relates to pneumatic conveying sys-
tems. Meng and Ludema [9] quote 33 independent parame-
ters in a recent review of 22 erosion models and predictive
equations found in the literature. Thus, the complexity of
slurry erosion is due to the number of important variables
present, some of these are listed in Table 1.
Table 1
Some of the main variables which inuence erosion
Slurry variables Component variables
Liquid: viscosity, density, surface
activity, lubricity, corrosivity,
temperature
Bulk properties: ductility or
brittleness, hardness and
toughness, melting point,
microstructure, shape and
roughness
Particles: brittleness, size, density,
relative velocity, shape, relative
hardness, concentration,
particle/particle interactions
Surface properties: work
hardening, corrosion layers,
surface treatments, coating type,
coating bond, microstructure
Flow eld: angle of impingement,
particle impact efciency,
boundary layer, particle rebound,
degradation, particle drop-out,
turbulence intensity
Service variables: contacting
materials, pressure, velocity,
temperature, surface nish,
lubrication, corrosion, hydraulic
design, intermittent slurry ows
For pipe bend erosion the problem is to develop a model
and in parallel develop a better understanding of the ow
eld and particle trajectories to generate accurate inputs into
the model, and then compare the model predictions with
experimental results.
Discussion (recently in an excellent review by Kaushal
and Tomita [10]) has focussed on the distribution of erod-
ing particles in a cross-section through a straight duct, but
understanding the particle trajectories through bends is of
great importance if wear rates are to be predicted at specic
locations. For bends in a pneumatic conveying line, Mills
and Mason [11] found signicantly greater wear depth for
70 m sand compared to 230 m sand. Blanchard et al. [12]
attempted a two-dimensional model of particle trajectories
in liquid ows within bends but with limited success due to
the inability of the model to predict secondary ows. At the
University of Southampton, Forder [13] has studied solid
particle trajectories in liquid ows in pipe bend geometry
before modelling more complex ows in choke valves us-
ing a commercial computational uid dynamics (CFD) code.
Forder [13] extended this commercial CFD programme to
include a predictive erosion element based on a combined
deformation and cutting erosion model and this will be dis-
cussed in detail in the following section. The model was
tested by comparing the predicted wear rates and wear lo-
cations in pipe bends with the experimental air/sand erosion
results of Bourgoyne [14], with excellent correlations. Fur-
ther experimental wear results for pipe bends could be used
to test this approach in the future such as reported by King
et al. [15] and Wiedenroth [16] for a limited range of bend
geometries, materials and liquidsolid ows. The redistri-
bution of solids inside horizontal bends for multisized par-
ticulate slurries has been experimentally studied by Ahmed
et al. [17]. Conclusions from this work point to the redistri-
bution of large particles outwards to be the likely cause of
rapid bend erosion.
940 R.J.K. Wood et al. / Wear 256 (2004) 937947
2. Component erosion models
The prediction of erosion rates and locations is difcult
for equipment handling particle-laden process streams. The
evolution of erosion models and commonality between them
has been reviewed by Meng and Ludema [9], Wood et al.
[18] and Hutchings [19].
Recent erosion models recognise the two erosion mech-
anisms of cutting and deformation erosion, with discrete
models representing each, e.g. Forder [13]. A cutting ero-
sion model was rst proposed by Finnie [20] and later
modied by Hashish [21]. For low impact angles (typi-
cally 040

) of relatively hard particles on ductile targets,


cutting is likely when the shear stresses induced by the
impact exceed the shear strength of the target. A defor-
mation model, Bitter [22], is thought applicable at higher
impact angles (3090

). A stress eld is generated within


the contact inducing plastically deformed sub-layers where
the stress (typically sub-surface) exceeds the yield strength
of the target material. This leads to erosion by delamina-
tion, micro-cracking and coalescence of voids to produce
surface fragmentation. These models operate in parallel and
the summation of the two presents the overall volumetric
erosion per impact and can be given by:
W
t
=
_
100
2

29
r
3
p
_
U
p
C
k
_
n
sin 2

sin
_
+
_
M
p
(U
p
sin D
k
)
2
2E
f
_
(1)
where
C
k
=
_
3R
0.6
f

p
(2)
and
D
k
=

2
2

10
(1.59Y)
2.5
_
R
f

t
_
0.5
_
1 q
2
p
E
p
+
1 q
2
t
E
t
_
2
(3)
and n = 2.54 for carbon steel, Hashish [21], E
f
= 1.910
10
for AISI 4130 steel.
This model, as used by Forder [13], includes important
variables such as particle shape (roundness R
f
) and material
properties for both particle and target that are typically not
considered by earlier simple models.
This approach is similar to that of a recent paper by Li and
Shen [23] which describes a computer simulation of pipe
bend erosion in pneumatic systems conveying low concen-
trations of granular material. Forder [13] and Li and Shen
[23] use computer modelling to determine the velocity eld
of the particles for given boundary geometries, particle size
and density, particle loading, uid type and driving pressure
gradient. This information is placed into an erosion model
based on Eq. (1). The empirical constants in these mod-
els were set by [13,23] to agree with experimental work by
Table 2
Values of parameters used in Eqs. (1)(3)
Parameters Value
Volume ow rate, Q
v
(m
3
s
1
) 0.01
Pipe bore (m) 0.078
Particle density,
p
(kg m
3
) 2670
Roundness factor, R
f
0.5
Plastic ow stress, (Pa) 1.00E+09
Yield stress of target, Y (Pa) 3.20E+08
Target density,
t
(kg m
3
) 7850
Target Poissons ratio, q
t
0.3
Particle Poissons ratio, q
p
0.23
Particle Youngs modulus, E
p
(Pa) 5.90E+10
Target Youngs modulus, E
t
(Pa) 2.07E+11
Deformation erosion factor, E
f
(J m
3
) 1.9E+10
Particle radius, r
p
(m) 500
Velocity ratio exponent, n 2.54
Particle velocity on entry, U
p
(ms
1
) 3.0
Particle mass, M
p
(kg) 1.4E06
Shimizu et al. [24] using copper and glass or established us-
ing a slurry jet impingement rig at various jet angles up to a
jet velocity of 35 ms
1
on stainless steels and sintered tung-
sten carbides, Forder [13]. Wallace et al. [25] used a similar
approach coupling a CFD package with an alternative ero-
sion model, with fewer variables, based on that of Neilson
and Gilchrist [26]. These models still need to be veried
by experiment and in the case of the Neilson and Gilchrist
model at least four sets of experimental data are required
obtained at different impingement angles and velocities to
evaluate the empirical constants.
2.1. Computational simulations
Computational simulations have been made of a straight
pipe and bend that evaluate the impact angle and velocity
distribution over the internal surfaces. These values are in-
corporated into the BitterHashish component erosion mod-
els based on Eq. (1) to predict erosion patterns within these
components. FLUENT v5.4 CFD code
1
with an embed-
ded multi-phase (algebraic slip) model was used. A bend
of 0.078 m bore diameter and a R
c
/D ratio of 1.2 with a
0.078 m bore diameter straight pipe section 0.56 m in length
located directly upstream of the bend was modelled. Both
components were modelled in a horizontal plane. The liq-
uid phase was set to typical values for water (viscosity =
0.0009 Pa s, density = 1000 kg m
3
), the solids had a den-
sity of 2670 kg m
3
and a size of 1 mm. Both liquid and solid
entry velocities were set to 3 ms
1
with a uniform solid dis-
tribution. A standard k turbulence model was used with
standard wall functions and zero roughness. Impact angles
() were dened as those from the tangent to the bend wall.
Table 2 summarises the parameters used.
1
Fluent Inc., Centerra Research Park, 10 Cavendish Court, Lebanon,
NH 03766, USA.
R.J.K. Wood et al. / Wear 256 (2004) 937947 941
180
90
0
270
-16
-14
-12
-10
-8
-6
-4
-2
0
2
4
6
0 90 180 270 360
in plane angle ( )
I
m
p
a
c
t

a
n
g
l
e

(

)
0 plane
45 plane
pipe cr oss-section
0
2
4
6
8
10
12
0 90 180 270 360
in plane angle ()
S
a
n
d

v
o
l
u
m
e

(
%
)
0 plane
45 plane
0
0. 5
1
1. 5
2
2. 5
3
0 90 180 270 360
in plane angle ()
I
m
p
a
c
t

v
e
l
o
c
i
t
y

(
m
/
s
)
0 plane
45 plane
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 90 180 270 360
in plane angl e ()
E
r
o
s
i
o
n

r
a
t
e

(
m
m

3
/
m
i
n
u
t
e
)
0 plane
45 plane
Fig. 3. Computational results.
The wear rates can be calculated by using the modied
HashishBitter model, Eq. (1). The variation of the overall
wear rate W
t
with impingement angle is plotted elsewhere,
Wood et al. [18], and predicts maximum erosion to occur at
slightly different angles to that of Neilson and Gilchrist.
The results from the coupled CFD and erosion model are
illustrated in Fig. 3. Particle impact angles and velocities at
a plane midway along the straight section and perpendicu-
lar to its axis (labelled 0

plane within the yz plane) were


evaluated as well as for a plane at 45

or midway around
the bend. The impact angles, sand volume, impact velocity
and erosion rates are plotted for both the 0

plane within the


straight section and the 45

plane within the bend. Although


Alternative
Route
Magnetic
Flow Meter
Centrifugal
Slurry Pump
Straight 43
Bend 65
Stirred
Tank ~3m
3
2.64m
2.48m
4.68m
Fig. 4. Details of the experimental loop.
secondary recirculating ows (negative impingement angles)
are evident between the in-plane angles of 90

and 270

the
particle loading and impact velocities between these angles
are such that little wear is predicted with most damage oc-
curring 270

and 90

.
3. Experimental loop
The experimental loop comprised horizontal straights and
bends, Fig. 4. The rig could be broken down into sections
to allow inspection after each test. For each test, sandwater
mixtures at 10 vol.% were pumped through the rig at a ow
942 R.J.K. Wood et al. / Wear 256 (2004) 937947
C
A
G
E
0
0.1
0.2
0.3
0.4
0.5
0 90 180 270
all straight pipe sections
0 90 180 270
D C B
A
G
E
0
0.1
0.2
0.3
0.4
0 90 180 225 270 315
Circumferential position of wear mode:
Circumferential position of wear mode:
all bends
P
r
o
b
a
b
i
l
i
t
y
P
r
o
b
a
b
i
l
i
t
y
0 90 180 225 270 315
Flow into Page
A
B
C
D
E
F
G
H
0
90
180
270
(c) (d)
Wear mode (0.76 m/h)
(a) (b)
Fig. 5. Wear modes and their distribution over straight and curved ducts: (a) measurement points (not all were used on every plane); (b) measurements in
one plane (ACEG) on a straight duct showing wear mode at position C; (c) distribution over all straight pipe sections (13 planes in total); (d) distribution
over all curved ducts (18 planes in total).
velocity of 3 ms
1
over 30 min periods, with diversion to
an alternate loop of the same geometry for 30 min. The tests
were repeated to a total of 210 h with mean 1000 m di-
ameter sub-rounded to sub-angular silica sand (D14/25 size
range 5001400 m and hardness 1100 Hv) from Hepworth
Minerals and Chemicals Ltd., Leighton Buzzard, UK. The
two test components, straight 43 and bend 65, are shown
in Fig. 4 because wear patterns in these components will
be described below. These patterns will be compared to the
predicted wear rates from the CFD and erosion model based
on these component geometries.
3.1. Wastage measurements
Following each test, pipe wall losses were measured by
a combination of techniques. The rig was dismantled so
that each section could be weighed, measured by ultrasonic
thickness probe and visually inspected by endoscope. Before
this could be done, the sections were cleaned with warm
5% (v/v) citric acid to remove calcite deposits on internal
surfaces. This was necessary to allow the change in mass of
the pipe sections to be attributed solely to wear. The pipes
were then allowed to dry in air prior to weighing.
Gravimetric losses in each pipe were obtained by sub-
tracting its weight before and after each test. Each pipe
section was placed on an aluminium frame which dis-
tributed its weight between four electronic balances. Each
balance was nominally capable of weighing up to 8.1 kg
to the nearest 0.01 g. The outputs of the balances were
controlled by means of an RS232 interface to a personal
computer.
Ultrasonic thickness gauges (Krautkramer type DM4 DL)
were used to measure the thickness at a selection of eight
circumferential points labelled ABCDEFGH in a clock-
wise direction. Point A was at the top of the pipe (or
180

in-plane angle), point C was at the 3 oclock posi-


tion and so on. The planes at which the surveys were car-
ried out were dened so that results could be compared
from test to test. Straight sections were tested at the car-
dinal points ACEG. Bends were tested at ABCDE and
G to capture the outside of the bends, the top and bot-
tom point and one point (G) at the inside of the bends.
Fig. 5(a) shows the measurement points and Fig. 5(b)
shows a typical plane through a pipe section showing
the mode or maximum of the wear distribution at that
plane. Fig. 5(c) and (d) show the distribution of the wear
modes (maximal wear points) in straight sections and
bends.
Following the series of tests, some components were de-
structively tested and measured by ball ended micrometer
at the locations at which ultrasonic measurements had been
taken. The results of these tests on the bend are shown in
Fig. 6. Ultrasonic readings showed a small but signicant
bias. This was explained as follows. The contact area, or
zone of the ultrasonic beam was different from the area of
contact of the ball-ended micrometer. Both instruments are
most accurate when estimating the thickness of a rectilin-
ear piece. When the increased contact zone of the ultrasonic
beam is superimposed onto the curved surface of the pipe,
a slight over-estimate of the thickness of the pipe occurs.
The smaller contact area of the micrometer caused a smaller
over-estimate. When the ultrasonic readings were compared
with the micrometer readings the former had a positive bias,
i.e. the ultrasonic estimates were higher than the micrometer
estimates. The calibration of the ultrasonic instrument was
done on a square at test piece of the same steel as that used
for the pipes and did not take pipe curvature into account,
so the full regression line in Fig. 6 incorporates a small
intercept. The ultrasonic transducer was therefore expected
to read slightly high for lower wastage in straight pipe 43
and slightly low for high wastage in bend 65. The modal
bias in Fig. 6(c) differed slightly from the regression value.
R.J.K. Wood et al. / Wear 256 (2004) 937947 943
4.0
4.5
5.0
5.5
6.0
4.0 4.5 5.0 5.5 6.0
Micrometer (mm)
U
l
t
r
a
s
o
n
i
c
(
m
m
)
0.0
1.0
2.0
3.0
4.0
5.0
6.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Micrometer (mm)
U
l
t
r
a
s
o
n
i
c

(
m
m
)

Data
y = 0.9731x
y = 1.0064x - 0.177
0
1
2
3
4
5
6
7
0
0
.
0
4
0
.
0
8
0
.
1
2
0
.
1
6
0
.
2
0
.
2
4
0
.
2
8
0
.
3
2
0
.
3
6
0
.
4
Micrometer-Ultrasonic Reading
F
r
e
q
u
e
n
c
y
(a) (b)
(c)
Fig. 6. Comparison between ultrasonic gauge and micrometer at the same measurement points over every measurement plane on four pipe components: (a)
calibration lines with and without intercept showing crossover; (b) good correspondence at small scale; (c) distribution of deviations between micrometer
and ultrasonic readings.
The standard error was 0.04 mm, giving an error range of
0.090.17 mm (using a modal bias of 0.13 mm). Typical er-
rors from micrometer readings, obtained by repeating the
measurements, were found to be in the order of 0.03 mm.
The uncertainty from gravimetric readings, also obtained
by repeating measurements, was estimated at 0.09 g at
1 S.E.
Destructive testing of the components also allowed a de-
tailed analysis of the erosion patterns using a scanning elec-
tron microscope. Fig. 7 presents micrographs of the erosion
patterns from the test loop at the top (point A or 180

) and
side (point D or 315

) at the 45

plane of the bend and bot-


tom (point E or 360

) of the straight pipe at the 0

the plane.
From Fig. 5(d) it is clear from the ultrasonic measurements
that point E (bottom) suffers the most damage while point D
(side) suffers approximately half as much and point A (top)
the least damage. This agrees well with erosion damage in-
formation from the micrographs of each point as explained
below. Fig. 7(a) shows point A has a complex surface to-
pography matching that of the as-received internal pipe sur-
face with evidence of some erosion damage to protruding
areas. Fig. 7(b) shows point D that is within the boundary
region between the suspended and contact layers and shows
a low density of microcutting scars (mainly aligned with
the ow direction). This is consistent with the particle posi-
tion model described in Section 1.1. Relatively large num-
bers of particles of relatively high kinetic energy impinge on
the pipe surface. In the present study, these particles have
eroded away most of the original surface features, although
some original features are seen in the centre of the micro-
graph. Fig. 7(c) shows typical severe damage patterns found
at point E with multiple microcutting impact sites (mainly
aligned with the ow direction) and complete removal of all
as-received surface features. Further details can be found in
Wood and Jones [27].
The comparison of gravimetric data for bend 65 and
straight pipe 43 should be done on the basis of the area
of the pipe interior which is in contact with the particle
burden. The rst step is to calculate the entire area of the
pipe interior or wetted area. For the purpose of discus-
sion we assume that the area in contact with the particle
burden is 25% of the wetted area. It is not absolutely nec-
essary to make such an assumption for comparison of two
gravimetric estimates, but does yield more realistic data for
944 R.J.K. Wood et al. / Wear 256 (2004) 937947
Fig. 7. SEM micrograph showing wear pattern damage of the bend after 210 h at: (a) point A or 180

; (b) point D or 315

; (c) the straight pipe at point


E or 0

.
comparison with other methods. For both sections,

mean
=
M
L

t
0.25 wetted area
=
4M
L

t
wetted area
(4)
The wetted areas can be obtained from Eqs. (5) and (6):
wetted area = 2RL (for straight 43) (5)
where R is the pipe radius and L the pipe length
wetted area =

2
4
_
R
2
c

_
D
2
_
2
_
+2R(L
i
+L
o
)
(for bend 65) (6)
where R
c
is the bend radius, and L
i
, L
o
the input and output
legs.
Table 3 shows gravimetric data from the two pipe sections
converted to a mean penetration as described above.
The micrometer results for both test geometries are pre-
sented in Fig. 8. A total of 57 measurements were taken for
each plane (0

for the straight and 45

for the bend). To al-


low for variations in wall thickness induced in component
manufacture values obtained at planes of no or little erosion
(just downstream of inlet weld for both components) have
been subtracted from the in-plane values to generate a differ-
ence in wall thickness (). Not all the wall variations in the
manufacture of the bend can be removed by this method as
seen in Fig. 7(a) with positive differences in wall thickness.
Fig. 8(a) indicates that, for the bend, an asymmetric
erosion damage pattern has occurred between the in-plane
angles of 225

and 90

(points BCDEFG). Although, as


expected, it is noted that there is more (approximately 1.5
times) erosion of the outer bend radius between in-plane
angles 225

and 360

(points BCDE) compared to the in-


ner radius erosion that occurs between 0

and 90

(points
EFG). The CFD code predicts a fairly symmetrical erosion
pattern, see Fig. 3.
Table 3
Gravimetric measurements, M
L
, after 210 h
Straight 43 Bend 65
Gravimetric (g) 4.75 9.45
Wetted area (m
2
) 0.14 0.0613
Mean penetration over 25%
wetted area,
mean
(mm) from
Eq. (6)
0.017 0.077
R.J.K. Wood et al. / Wear 256 (2004) 937947 945
-1.50
-1.00
-0.50
0.00
0.50
1.00
0 90 180 270 360
In plane angle ()


w
a
l
l

t
h
i
c
k
n
e
s
s

(
m
m
)
-0.25
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20
0.25
0 90 180 270 360
In plane angle ()


w
a
l
l

t
h
i
c
k
n
e
s
s

(
m
m
)
(a)
(b)
Fig. 8. Difference in micrometer derived wall thickness between uneroded
and eroded areas: (a) wall thickness measurements for the bend section
at plane 45

; (b) for the straight pipe section at plane 0

.
Aless clear trend in measured penetration depth is seen for
the straight pipe, Fig. 8(b). Although, erosion has occurred
between 290360

and 3080

that is similar to the erosion


locations predicted in Fig. 3. The maximum and mean values
of these results are compared with the other measurements
and CFD predictions in Table 4.
A histogram of the results is given in Fig. 9. The ratios
of the mean values of bend to straight erosion are listed
in Table 5. It should be noted that whilst comparisons be-
tween the ultrasonic, micrometer and CFD predictions are
Table 4
Summary of measured, calculated and predicted erosion penetration depths
measurements Bend 65 Straight 43
Ultrasonic thickness (mm)
Maximum 0.41 0.40
Mean 0.11 0.09
Micrometer thickness (mm)
Maximum 0.70
a
0.19
Mean 0.21 0.02
CFD (mm)
Maximum 0.72 0.113
Mean 0.24 0.035
Gravimetric (mm)
Mean 0.077 0.017
a
Corrected for manufacturing processes that thin the bottom and
thicken the top walls.
Fig. 9. Comparison of wastage estimates (using results from Fig. 3 and the
wetted areas from Eqs. (5) and (6)) with micrometer and ultrasonic mea-
surements from pipe loop experiments at 3.0 ms
1
, 1 mm sand, 10 vol.%
sand concentration after 210 h.
Table 5
Ratios of bend to straight mean erosion
Mean bend/mean straight
erosion ratio
Ultrasonic thickness 1.2
Micrometer thickness 10.5
CFD 6.9
Gravimetric (all planes) 4.5
in-plane at one section, the gravimetric results reect an
average loss of all planes and are unable to detect ero-
sion maxima. Fig. 10 plots the measured (ultrasonic and
micrometer) results against the CFD predicted results and
shows good agreement between the micrometer and CFD
data while the ultrasonic measurements show some scatter.
The ultrasonic probe appears to yield higher values of wall
thinning on the straight compared to those predicted while
the reverse is true for the bend. Thus, caution should be
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 0.2 0.4 0.6 0.8
CFD predicted erosion penetration (mm)
m
e
a
s
u
r
e
d

e
r
o
s
i
o
n

p
e
n
e
t
r
a
t
i
o
n

(
m
m
)
Ultrasonic
Micrometer
Fig. 10. Comparison between CFD predicted results (using results from
Fig. 3 and the wetted areas from Eqs. (5) and (6)) with micrometer and
ultrasonic measurements from pipe loop experiments at 3.0 ms
1
, 1 mm
sand, 10 vol.% sand concentration after 210 h.
946 R.J.K. Wood et al. / Wear 256 (2004) 937947
used when interpreting ultrasonic thickness data from pipe
systems.
4. Conclusions
Sandwater experiments using a 78 mm bore pipe loop
system to investigate damage to a straight pipe and bend
found that wear maxima were distributed to either side
of the pipe bore for the straight pipe, in contradiction of
the common assumption of predominant wear at the base
of a pipe. In the bend there was an expected increase in
wear at the outsides of the bends compared to the in-
sides, but also signicant wear occurred at the base of the
bend.
The predominant erosion mechanism of the straight pipe
and bend surfaces was microcutting from low angle solid
particle impingements. Lower densities of cutting scars
were seen at points within the boundary region between
the suspended and contact layers of sand while higher
densities were evident at points within the base of the
contact layer.
Good agreement was found between the in-plane wall
penetration depths obtained using a ball ended micrometer
and the coupled CFD and HashishBitter erosion model
for both straight pipe and bend components.
Comparison between the coupled CFD/erosion models
and micrometer/ultrasonic measurements of the straight
pipe and bend show reasonable agreement between pre-
dicted and experimental erosion patterns.
Interpretation of ultrasonic thickness data from pipe sys-
tems was hampered by higher scatter in the data compared
to the micrometer data.
Gravimetric measurements, corrected for wetted area,
show the bend to erode at 4.5 times that of the upstream
straight pipe but gravimetric measurements of this kind
are not meaningful in the context of life prediction as they
refer to a smeared-out estimate of erosive wear. The
maximum erosion rates and wear locations are important
for pipe lifetime and failure prediction.
Acknowledgements
The authors would like to thank Mr. Tony Gospel of the
University of Nottingham and Dr. Julian Wharton of the
University of Southampton for their help throughout this
project.
References
[1] H.P. Jost, Lubrication (Tribology): Education and Research, UK
Department of Education and Science, 1966.
[2] M.C. Roco, T. Cader, Numerical method to predict wear distribution
in slurry pipelines, in: G. Jones, J. Thorn (Eds.), Advances in Pipeline
Protection, BHRA, Craneld, UK, 1988.
[3] K.C. Wilson, Slip point of beds in solidliquid pipeline ow, Proc.
ASCE J. Hyd. Div. 96 (1970) 112.
[4] K.C. Wilson, A unied physically-based analysis of solidliquid
pipeline ow, in: Proceedings of the Hydrotransport 4, BHRA,
Craneld, UK, 1976.
[5] G.W. Govier, K. Aziz, The Flow of Complex Mixtures in Pipes, Van
Nostrand, New York, 1972.
[6] S.-F. Chien, ASME FED, LiquidSolid Flows 189 (1994) 231
246.
[7] R.A. Williams, M.S. Beck (Eds.), Process Tomography: Principles,
Techniques and applications, Butterworths, Heinemann, 1995.
[8] M. Wang, T.F. Jones, W. Yin, J. Ganeshalingham, R.A. Williams,
N.J. Miles, D. Li, Y. Lai, Y. Wu, Measurement of swirling ow in a
hydraulic conveying loop using electrical resistance tomography, in:
Proceedings of the World Congress on Particle Technology, vol. 4,
July 2125, 2002, Sydney, Australia.
[9] H.C. Meng, K.C. Ludema, Wear models and predictive
equationstheir form and content, Wear 181183 (1995) 443
457.
[10] D.R. Kaushal, Y. Tomita, Solids concentration proles and pressure
drop in pipeline ow of multisized particulate slurries, Int. J.
Multiphase Flow 28 (2002) 16971717.
[11] D. Mills, J.S. Mason, Particle size effects in bend erosion, Wear 44
(1977) 311328.
[12] D.J. Blanchard, P. Grifth, E. Rabinowicz, Erosion of a pipe bend
by solid particles entrained in water, J. Eng. Ind. 106 (1984) 213
217.
[13] A. Forder, A computational uid dynamics investigation into the
particulate erosion of oileld control valves, Ph.D. Thesis, School
of Engineering Sciences, University of Southampton, 2000.
[14] A.T. Bourgoyne, Experimental study of erosion in diverter systems
due to sand production, in: Proceedings of the SPE/IADC Drilling
Conference, SPE/IADC 18716, Society of Petroleum Engineers,
1989, p. 807.
[15] R. King, B. Jacobs, G. Jones, Factors affecting the design of slurry
transport systems for minimum wear, in: Proceedings of the Pipe
Protection Conference, Cannes, France, Elsevier, Organised and
Sponsored by BHR Group, 1991, p. 67.
[16] W. Wiedenroth, An experimental study of wear of centrifugal pumps
and pipeline components, J. Pipelines 4 (1984) 223228.
[17] M. Ahmed, S.N. Singh, V. Seshadri, Distribution of solid particles
in multisized particulate slurry ow through a 90

pipe bend in
horizontal plane, Bulk Solids Handling 13 (2) (1993) 379385.
[18] R.J.K. Wood, T.F. Jones, N. Miles, J. Ganeshalingam, Upstream
swirl-induction for reduction of erosion damage from slurries in
pipeline bends, Wear 250 (112) (2001) 771779.
[19] I.M. Hutchings, TribologyFriction and Wear of Engineering
Materials, Arnold, London, 1992.
[20] I. Finnie, Some observations on the erosion of ductile metals, Wear
19 (1972) 8190.
[21] M. Hashish, An improved model of erosion by solid particles, in:
Proceedings of the Seventh International Conference on Erosion
by Liquid and Solid Impact, Paper 66, Cavendish Laboratory,
1988.
[22] J.G.A. Bitter, A study of erosion phenomena, Part I, Wear 6 (1963)
521.
[23] X. Li, H.H. Shen, A computer simulation of pipe bend erosion in
a dilute pneumatic transport of granular materials, Particulate Sci.
Technol. 14 (1996) 5973.
[24] A. Shimizu, Y. Yagi, H. Yoshida, T. Yokomine, Erosion of gaseous
suspension ow duct due to particle collision. 1. Experimental-
determination of erosion rate by individual collision, J. Nucl. Sci.
Technol. 30 (9) (1993) 881889.
R.J.K. Wood et al. / Wear 256 (2004) 937947 947
[25] M.S. Wallace, J.S. Peters, T.J. Scanlon, W.M. Dempster, S.
McCulloch, J.B. Ogilvie, in: Proceedings of the ASME 2000 Fluids
Engineering Summer Meeting, Paper FEDSM2000-1124, Boston,
MA, June 1115, 2000.
[26] J.H. Neilson, A. Gilchrist, Erosion by stream of solid particles, Wear
11 (1968) 111122.
[27] R.J.K. Wood, T.F. Jones, Investigations of sandwater induced
erosive wear of AISI 304L stainless steel pipes by
pilot-scale and laboratory-scale testing, Wear 255 (2003) 206
218.
Glossary
Wear distribution: The pattern of wear on the circumference of a duct
cross-section, viewed as a distribution.
Wear mode: The position on the circumference of pipe cross-sections
where the incidence of wear damage is greatest.
Pipe invert: The pipe invert is the region at the bottom of the bore where
settling particles tend to gather.
Extrados: The extrados is the outermost extremity of a curved duct.
Intrados: The intrados is the innermost extremity.

You might also like