Powder Technology 194 (2009) 1832

Contents lists available at ScienceDirect

Powder Technology
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / p ow t e c

Modelling of oversized material ow through a horizontal hydrotransport slurry pipe

to optimize its acoustic detection
K. Albion, L. Briens , C. Briens, F. Berruti
Institute for Chemicals and Fuels from Alternative Resources, Faculty of Engineering, The University of Western Ontario, London, Ontario, Canada N6A 5B9

a r t i c l e

i n f o

Article history:
Received 8 February 2008
Received in revised form 16 January 2009
Accepted 4 March 2009
Available online 19 March 2009
Keywords:
Hydrotransport
Process monitoring
Acoustics
Kurtosis

a b s t r a c t
Hydrotransport of solids through a pipe is a cost and energy efcient method to transport granular solid
materials over long distances. A problem in the hydrotransport of ne particle slurries is the possible
presence in the pipe of undesirable large materials, such as rocks or metal fragments broken off of shovels,
which may enter the slurry pipe through breaks in screens. This large material can damage booster pumps
and downstream equipment resulting in costly repairs and loss of production.
Acoustic sensors along with signal analysis techniques can be used for online detection of oversized material
in a hydrotransport system. Acoustic detection methods are ideal, since they are non-invasive and any probe
located within the pipe would be unlikely to survive the harsh conditions present within the line.
The objective of this study was to model the motion behaviour of large materials, such as rocks, travelling
through a horizontal hydrotransport pipe. This information can then be used to determine optimum
locations for the acoustic sensors to ensure that rocks are detected quickly and effectively.
2009 Elsevier B.V. All rights reserved.

1. Introduction

2. Background information

The oil sands of northern Alberta are a valuable source of bitumen

that can be processed to produce a synthetic crude oil. Many
advancements have been made in the oil sand industry, one of
which has been to reduce the number of draglines and belt conveyors
in the mine, in favour of shovels and hydrotransport pipes [1]. Mined
oil sand is transported from the mine to extraction facilities where
solids and water are separated from the bitumen. The bitumen is then
thermally or catalytically upgraded and processed to produce a light,
sweet crude oil [2] used in the production of fuels, petrochemicals and
lubricants [3].
In the hydrotransport of oil sand, it is not uncommon to have rocks
in the mined ore [2]. Although rocks can be removed from the sand
using screens before the oil sand is slurried, screen wear leads to holes
through which rocks, and metal pieces from shovels, can enter the
pipeline. This oversized material can become lodged in pump
impellers resulting in damage to the pump, downtime and loss of
production.
The objective of this paper is to present a model developed to predict
the motion of oversized material in a horizontal pipe in a hydrotransport
system. The model was veried using signals recorded by acoustic
probes and analysed using advanced methods. The model was used to
determine the optimum placement of the acoustic probes on the
horizontal pipeline for reliable detection of large rocks in the system.

This section provides background information on the use of hydrotransport in the oil sand industry, rock behaviour in open channels, acoustic
emissions and the detection of large material in a hydrotransport system
and an overview of particle trajectory models in transport systems.

Corresponding author. Tel.: +1 519 661 2111x88849; fax: +1 519 661 3498.
E-mail address: lbriens@eng.uwo.ca (L. Briens).
0032-5910/$ see front matter 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.powtec.2009.03.017

2.1. Oil sand mining and extraction

Layers of overburden, mainly clay, silt and gravel, must be removed
from the earth's surface to access the oil sand beneath the earth's
surface [3].
In the current process, trucks from the mine deliver the oil sand to
double roll crushers capable of crushing 6000 tons of oil sand per
hour. Crushed oil sand has a top size range of 50150 mm [4]. The
crushed oil sand is sent to a cyclofeeder, where it is mixed with caustic
soda and hot water and pumped using booster pump houses to the
extraction facility through a 27-inch diameter hydrotransport pipe.
Once in the extraction facility, the oil sand is sent to tumblers where
additional caustic soda, hot water and steam are added. The slurry is
then discharged to vibrating screens to remove solid materials [3].
Hydrotransport is the process of combining oil sand and hot water
to create a slurry that is pumped through the pipeline. Most of the
slurry passes through a screen that vibrates on a 5 to 10 angle which
allows material to move over the screen. Large remaining material is
discarded as waste or recrushed and processed [2]. In hydrotransport,
most large particles are oil sand lumps which ablate during transport.
However oversized material exists which can be oil free rocks or
undigested ore [4].

K. Albion et al. / Powder Technology 194 (2009) 1832

2.2. Rock motion

Chien and Wan [5] state that heterogeneous slurry ow through a
pipe is similar to, and can be compared to, open channel ow. Rock
motion and the distance rocks travel have been examined by many
researchers in open channel ow. Rock motion behaviour is affected
by particle shape, size and by the uid velocity.
Sanders et al. [4] state that in a hydrotransport pipe large, isometric
particles, specically round particles, are as easy to transport as
smaller particles. Shook et al. [6] explains that spherical particles may
roll at very low velocities, and isolated non-spherical particles slide
under the inuence of the drag force. At high slurry velocities particles
often become entrained and collide with the pipe wall.
Osborne [7] examined the effect of particle size on transport. It was
determined that particle transport increases with an increase in
particle size to a critical point when the rate begins to decrease with
an increase in particle size. The decrease of mobility of the largest
particles was attributed to the limited capability of uid forces to
transport the heaviest and largest particles. Ferguson et al. [8]
determined that the velocity of a rock depends on the size of the
particle, with smaller particles travelling faster than larger particles.
Wu and Chou [9] state that particle motion occurs through rolling,
sliding, lifting or bouncing. Properties that inuence particle transport
include turbulent ow uctuations, heterogeneous particle sizes, shapes
and densities, bed geometry, particle exposure and bed roughness.
Drake et al. [10] examined gravel motion using photography. It was
determined that for small particles, entrainment was through rolling
and lift-off. Entrainment of large particles was caused by rolling of large
particles and sliding of angular particles, with some entrainment caused
by particleparticle impacts. Fig. 1 shows entrainment of small rocks,
rolling of round rocks and sliding and entrainment of large angular
rocks. It was determined that, during transport, the average distance
travelled by a particle was equivalent to 15 times the particle diameter.
2.3. Acoustics
Acoustic sensors are inexpensive and provide reliable, online and
non-intrusive monitoring. This technology is easy to implement:
microphones attach to the external surface of the pipe wall and can be
easily relocated.
There are two acoustic monitoring methods:
active acoustics detect the effect of the process on a transmitted
acoustic wave, usually in the ultrasonic range; and
passive acoustics detect the acoustic emissions generated by the process.

19

Generally, passive acoustic methods are much easier to implement and

are preferred when the process acoustic emissions are strong, as in
hydrotransport systems.
Acoustics is dened as the study of sound. Sound is caused by pressure
vibrations that can be detected by the human ear or microphones. In
processes involving the movement of solid particles, acoustic emissions
are caused by particles colliding with each other and vessel walls [11].
The microphones used in this study were prepolarized electret
condenser microphones, which transform sound pressure uctuations
into capacitance variations, which are further converted into an
electrical voltage signal. They produce an oscillating voltage proportional to the original pressure oscillations [12].
2.4. Signal analysis kurtosis
Kurtosis is a measure of the relative peakedness of a distribution.
Kurtosis is calculated by:
M
P

K=

y y4

M 4

where y is the signal value, y is the signal mean, is the signal

standard deviation and M is the number of points in the signal.
Kurtosis is a dimensionless value used to determine the relative
height of a peak, and is classied based on the shape of a single peak:
leptokurtic more peaked, mesokurtic normally peaked, and
platykurtic less peaked [13].
2.5. Acoustic detection of large material in a hydrotransport system
Albion et al. [14] used acoustic probes located along a hydrotransport
system to detect large material in a sandwater slurry. The acoustic signals
were analysed using kurtosis, which clearly identied collisions between
the rock and the pipe wall. The number of collisions between the rock and
the pipe wall was found to be dependent on the rock size and the slurry
velocity, with more collisions occurring for large rocks and at low slurry
velocities. The number of collisions between the rock and the pipe wall led
to the development of correlations, based on the acoustic signal, to
determine if the rock was oversized (rock diameter/pipe diameter N25%)
which would result in immediate damage to the system, or if the rock was
an acceptable size (rock diameter/pipe diameter b25%) and would not
cause immediate damage to the system. It was determined that with an
array of 10 microphones, there was 100% detection of all rocks in the slurry,
which varied in size from 9.119.2 mm.

Fig. 1. Rock motion encountered in sediment transport (Adapted from [10]) of: a) lift-off of a small particle, b) rolling of a large particle, c) sliding and lift-off of a large angular particle.

20

K. Albion et al. / Powder Technology 194 (2009) 1832

2.6. Particle trajectory models

Particle trajectories have been modelled in various systems,
including pneumatic transport, sediment transport and slurry
pipelines. These models have focused largely on incipient motion
and transport of small particles in laminar ow.
Forces that were considered in these particle trajectory models are
illustrated in Fig. 2 and include:
Added mass force: resulting from the relative acceleration or
deceleration of a particle compared to the uid; the added mass is
the mass of the uid that would result in the same inertial effect if it
accelerated with the particle [15].
Basset history force: due to a particle accelerating or decelerating in
a uniform ow, as a function of the history of the relative velocity
during a time interval [15].
Buoyancy force: due to the weight of liquid displaced by the volume
of liquid equivalent to the particle volume [16].
Drag force: results due to the difference between the particle
velocity and the uid velocity for a particle moving with a constant
velocity in a uniform ow [17].
Friction force: opposes relative motion of a particle in the uid and
acts in the opposite direction of the drag force [18].
Lift force: experienced by a particle moving at a constant velocity
due to velocity gradient in high shear regions [17].
Magnus force: a lift force that results from particle rotation and a
velocity gradient in non-uniform ows, which moves the particle
into regions of higher ow velocity, and is negligible for small
particles or slow particle rotation [17].
van der Waals force: attraction between particles, or particles and
the pipe wall, due to interacting dipoles [17].
Cabrejos and Klinzing [18] modelled the incipient motion of
spherical particles in a horizontal pneumatic transport line to
determine the gas velocity required to pick up particles from the
bottom of the pipe. A force balance was performed using the lift,
buoyancy, gravity, adhesion (through the van der Waals force), drag
and friction forces. They determined that several factors inuence the
pick-up velocity of different solid particles, which include: particle
size, density, shape and interaction with other particles. Their model
had good agreement with experimental tests of pick-up velocities of
spherical particles at rest in a pneumatic transport pipe.
Wilson and Sellgren [19] studied the suspension of small particles
in a slurry pipeline. They determined that the shape of the particles
affects the lift force which is responsible for suspending the particle in
the ow. The KuttaZhukovski equation was used to model the lift

force near the pipe wall in the viscous sublayer, which depends on the
shape of the velocity prole. It was determined that for particles
smaller than 0.15 mm, the lift force was not signicant enough to lift
the particle out of the viscous sublayer.
The bouncing motion of spherical particles in stagnant liquids and
air was modelled by Gondret et al. [20]. Experiments were also
conducted into the measurement of the coefcient of restitution in
both uids. Particle trajectories were modelled using gravity, drag,
added mass and the history forces. These forces were applied under
Stokes conditions as the uids were stagnant. They determined that
these four forces were sufcient to accurately model the trajectory of
spherical particles, and the lift force was not required as it was
assumed to be weak near the wall.
Lick et al. [21] modelled the incipient motion of ne and coarse
quartz particles and the effect of cohesivity, which is dependent on the
particle diameter and sphericity. The forces analysed for this model
included gravitational, lift, drag and van der Waals forces. They found
excellent agreement between the model and experimental data,
however an additional binding force was required to predict incipient
motion for the cohesive slurry.
Wang and Shirazi [22] used Computational Fluid Dynamics (CFD)
to model particle trajectories through elbows and bends to determine
erosion rates for slurry pipelines. The velocity prole for the uid was
modelled using the NavierStokes equation. The particle trajectories
were modelled with a force balance containing the drag force, the
buoyancy force due to centrifugal motion, the added mass force, and
the viscous sublayer. The sublayer was used to determine if particles
would rebound, or become trapped in this region. Good agreement
was found between their model and experimental data from literature
for determining the amount of material removed from the elbow at
low particle owrates.
Average ow turbulence for the coarse sediment transport in water
was used to estimate transport velocities, friction coefcients and
particle motion in an unbounded uid [23]. Included in their force
balance was the drag force in the Stokes regime, submerged gravity,
Basset history force, the lift force, and the Magnus force. They found
that their model performed satisfactorily when compared to experimental measurements of particle trajectories, and suggested that lift
forces were important in the modelling of the trajectory.
Willetts [24] studied the behaviour of a sand grain after being
dislodged from the surface of a stream. His model was based on lift
forces due to circulation, buoyant forces and drag forces. The
trajectories were found to be sensitive to the slope of the bed, particle
density and uid velocity. It was found that there was good agreement
between the model and videos of particle trajectories, regarding the

Fig. 2. Forces acting on a transported particle.

K. Albion et al. / Powder Technology 194 (2009) 1832

trajectory shape, and that the trajectory size increased with a decrease
in the particle size.
Salman et al. [25] developed a model to simulate the motion of a
particle in a horizontal pneumatic transport pipe. It was determined
that forces inuencing particle motion are drag in the horizontal
direction, and lift due to the velocity gradient and particle spin. This
model was found to have good agreement with experimental values
regarding the length of time a particle was suspended in the ow,
however, the model was not complete as to the modelling of particle
trajectories and forces.

21

Table 1
Flow properties of the slurries studied.
Slurry concentration
(wt.%)

Slurry density
(kg/m3)

Effective slurry viscosity

(Pa.s)

10
30
50

1064
1228
1450

1.13 10 3
1.58 10 3
2.90 10 3

Ball valves below the storage tank allowed for drainage of the slurry
from the tank and pipeline.

3. The modelled hydrotransport system and experimental methods

3.2. Slurry characteristics

3.1. Hydrotransport system

The slurry was made up of silica sand at various sand concentrations. The Sauter-mean diameter of the silica sand was 180 m and its
density was 2650 kg/m3. The terminal velocity of the sand was
calculated to be 0.022 m/s.
The slurry concentration was controlled by the amount of silica
sand added to the storage vessel. Slurry velocities of 2, 3 and 3.5 m/s,
and slurry concentrations of 10, 30, and 50 wt.%, were studied. The
effective slurry viscosity was calculated with the correlation by Gillies
et al. [26]. The slurry concentrations and their corresponding densities
and viscosities are shown in Table 1.

The hydrotransport system examined in this study consisted of a

0.05 m inside diameter stainless steel pipe, shown in Fig. 3a. Water
and solids were initially added to the storage tank through an opening
at the top of the vessel. Air from a compressor powered the diaphragm
pump used to circulate the slurry through the pipeline. To reduce the
effect of vibration from the pump, sections of pipe at the top of the
inclined line and at the end of the horizontal line were replaced with a
section of 0.05 m diameter reinforced exible hose. A magnetic ux
owmeter was used to set and monitor the slurry velocity, which was
controlled by the owrate of air supplied to the pump. After exiting
the pump, the slurry made a 90 bend and entered the top of the
inclined pipe section. The inclined section had a length of 2.75 m at an
angle of 30 from the horizontal. The slurry then travelled through a
1.7 m vertical section and a 3.6 m horizontal section before returning
to the storage tank. The pipeline dimensions are shown in Fig. 3b. A
0.5 m transparent PVC pipe was located at 1.4 m from the elbow which
allowed for visual observations of the slurry ow and rock motion. A
rock inlet chamber was located at the top of the inclined line. This
chamber consisted of two ball valves, which acted as a water lock and
allowed for the introduction of rocks under running conditions. A rock
cage was located on the end of the pipe, inside the storage tank, to
prevent rocks from re-entering the system and damaging the pump.

3.3. Rock characteristics

Thirteen rocks were used to model motion over a range of particle
sizes, shapes and densities. The volume-equivalent rock diameters,
rock densities, masses, volumes, projected areas, and terminal
velocities in water are shown in Table 2. The rocks are shown in Fig. 4.
3.4. Acoustic sensors
Six prepolarized electret microphones model 130D10, preamplier
model 130P10, manufactured by PCB Piezotronics were used to
simultaneously record acoustic signals at different locations along the
horizontal pipeline to verify the model. Data was acquired using a 12-

Fig. 3. a. Schematic diagram of the hydrotransport system. X indicates the experimental acoustic probe measurement locations. b. Schematic diagram including dimensions of the
hydrotransport system.

22

K. Albion et al. / Powder Technology 194 (2009) 1832

Table 2
Physical properties of rocks studied.
Rock

Volumeequivalent
diameter
(mm)

Density
(kg/m3)

Rock
mass
(g)

Rock
volume
(mL)

Projected
area (mm2)

Terminal
velocity
(in water)
(m/s)

A
B
C
D
E
F
G
H
I
J
K
L
M

19.2
16.8
17.9
15.1
12.0
11.5
10.5
9.1
18.3
17.5
15.4
14.2
9.1

2735
3160
2867
3111
3111
3250
2750
3125
2700
2482
2726
2133
3700

10.1
7.90
8.60
5.60
2.80
1.69
1.65
1.25
8.64
6.95
5.18
3.20
1.48

3.7
2.5
3.0
1.8
0.9
0.8
0.6
0.4
3.2
2.8
1.9
1.5
0.4

411
345
334
211
133
122
111
89
489
345
400
189
122

0.75
0.75
0.63
0.63
0.75
0.63
0.94
0.75
0.46
0.63
0.75
0.75
0.54

bit National Instruments data acquisition card model NI PCI-6071E.

Each measurement was recorded at a frequency of 40 000 Hz. Acoustic
sensors were located directly on the side of the pipe, at each
measurement location. Experimental measurement locations were
evenly spaced along the pipe at 0.03, 0.50, 1.00, 1.50, 2.00 and 2.50 m
along the horizontal section measured from the vertical to horizontal
elbow. The microphone locations are shown in Fig. 3. Acoustic sensors
were secured to the pipeline using a shim shaped to the form of the
sensor, and attached to a thin foam base. This base was held to the pipe
with Velcro wrapped around the pipe circumference.

ball valve was opened and a rock dropped into the chamber, and the
rst valve closed. The second valve was then opened to drop the rock
into the pipe. The rock travelled through the system and was caught in
the rock cage at the end of the pipe in the storage tank. To recover the
rocks, the rock cage was unscrewed off of the pipe, emptied and
replaced. Multiple experiments were conducted for each rock at all
combinations of slurry concentrations and velocities, for a total of
more than 800 experiments.
The terminal velocity of the rocks was determined by using water
and sugar solutions at the viscosities corresponding to the effective
slurry viscosities. Terminal velocities for the rocks in water are
presented in Table 2. Distances were marked on a clear 0.05 m
diameter acrylic pipe and the rocks falling through the pipe were
recorded using a video camera. The video les were recorded at 15
frames per second. The terminal velocity was calculated based on the
number of frames required for the rock to move through a set distance,
after checking that the rock was fully accelerated.
3.6. Signal analysis methods
Raw signals were recorded using National Instruments' LabVIEW
data acquisition software. Kurtosis of the signal was calculated; the
amplitude of peaks in the raw signal correspond to the magnitude of
the peaks of kurtosis. The time interval length to calculate kurtosis
was 0.010 s. A kurtosis threshold limit was set using a limit of 10.
Albion et al. [14] provides the details regarding the selection of the
time interval length and the threshold limit values. The number of
peaks in the signal was calculated based on the number of peaks above
this threshold value.

3.5. Experimental method

4. Model development
For each slurry concentration, 45 L of water were added to the
storage tank. The pump was started and the appropriate amount of
sand was weighed and added to the water in the tank. Pumping the
water prevented the sand from settling out of the slurry as it was
added and kept the slurry mixed. Once the appropriate sand
concentration was reached, the desired slurry velocity was set. After
the ow stabilized, the acoustic measurements were started. One rock
was added to the slurry during each acoustic measurement. The rst

Little research has been conducted into the modelling of large

particle trajectories in slurry transport systems. This model has been
developed to predict the trajectory of rocks in a horizontal hydrotransport pipe at slurry velocities of 2, 3 and 3.5 m/s and at slurry
concentrations of 10, 30 and 50 wt.%. A variety of rocks were used in
the model. The properties of the slurry are given in Table 1. The rocks
are shown in Fig. 4, and the rock properties are given in Table 2. The

Fig. 4. Rocks modelled and used in the hydrotransport system.

K. Albion et al. / Powder Technology 194 (2009) 1832

length of horizontal pipe used in the model was 3 m. The x-direction is

the axial direction along the horizontal pipe, and the y-direction is the
radial direction.
4.1. Detailed model description
4.1.1. Velocity prole
The velocity prole was calculated for turbulent ow of liquid in a
pipe. Eqs. (2) and (3) are valid for Re N 4000. The Reynolds Number of
the ow in the pipe at a slurry velocity of 2 m/s, the lowest slurry
velocity examined, is approximately 94 000. It is assumed that the
radial velocity prole is constant along the length of the pipe, is not
affected by the elbow and the corresponding acceleration region, and
can be used as a rst approximation.
Eq. (2) [27] was used to determine the maximum velocity at the
centreline of the pipe using the average slurry velocity measured by
the magnetic ux owmeter. The magnetic ux owmeter measures
an area-averaged slurry velocity over the cross-section of the pipe
[15], which takes into account the effect of solids on the velocity.
Eq. (3) [27] was used to calculate the slurry velocity at a specic
location in the vertical direction in the pipe, since the velocity prole
changes with the distance from the pipe wall. The vertical position of
the rock was used in Eq. (3) to obtain the slurry velocity in the
horizontal direction at that location.
Umax =

Uslurry;avg
49 = 60

A constant concentration distribution was assumed over the pipe cross

sectional area, as at the velocities examined in this experimental work and
model, solids did not settle on the bottom of the pipe [28]. The viscosity of
the sand slurries was calculated using the Gillies et al. [26] correlation (Eq.
(4)) for sand particles with a narrow size distribution. This correlation was
developed using sand particles of approximately the same size as the sand
in this experimental work, and at similar concentrations to a maximum of
50 vol.%. The density of the slurry was calculated (Eq. (5)) using the weight
fraction of sand in the slurry.
n
o
2
= liquid 1 + 2:5Cv + 10Cv + 0:0019 exp20Cv

1
C
1 Cm
= m +
:
sand
liquid
slurry

4
5

Re =

slurry

CD =



Vrock rock slurry g
2
0:5slurry Arock Ut;rock

FD = 0:5CD j Uslurry Urock j slurry Arock :

7
8

Drag force in x-direction:

FD;x = FD cos :

FD;y = FD sin :

8a

4.1.3. Drag force

In Eq. (7), the drag coefcient is calculated based on the shape of
the rock. The drag coefcient is calculated since the shape of the rock
has a signicant effect on the drag coefcient, and a constant value of
0.43 at high Reynolds Numbers is not applicable. The projected area
was used to take into account the shape of the rock on the drag
coefcient. The projected area of the rock was taken as the projected

8b

4.1.4. Added mass force

The added mass force is a force that is exerted on a moving object
as it accelerates through a uid leading to an acceleration of the uid
surrounding the object [29]. This force has been included in particle
trajectory models [23], however, since the rock is moving slower than
the slurry, this force does not apply in this situation.
4.1.5. Buoyancy force
The buoyancy force is an upward force acting on an object due to
the surrounding uid it is immersed in. The net buoyancy force is
equal to the magnitude of the weight of the displaced uid and was
calculated using Eq. (9).

FB = Vrock slurry g:

4.1.2. Force balance

The Reynolds Number, Eq. (6), was used to determine the drag
force acting on the rock, relative to the slurry velocity. The volumeequivalent diameter of the rock was used since rocks have many axes
that can be measured. The volume-equivalent diameter allows for a
reproducible measured diameter as it is based on the volume of a
sphere having the same volume as the rock.



drock Uslurry Urock slurry

area that is not the largest area or smallest projected area of the rock, as
videos of rocks travelling through the transparent section of pipe have
shown that rocks tumble resulting in a variety of projected areas
normal to the ow.
For spherical particles, a constant drag coefcient value of 0.43 can
be used as the Reynolds Number is in Newton's regime (Rep 970 000
for relative velocity of 1 m/s). However, the calculated drag coefcient
using the terminal velocity and shape of the rock also results in a drag
coefcient found in Newton's regime.
The drag force acts on an object as it moves steadily through a uid
[29]. The drag force in Eq. (8) is necessary since the rock and slurry do
not travel at the same velocity. A drag force acts on the rock in the
vertical and horizontal directions as it travels through the pipe, and
the force must be broken into components as in Eqs. (8a) and (8b)
[30].

Drag force in y-direction:

ry Dpipe 17

:
Uslurry = Umax 1
Dpipe

slurry

23

4.1.6. Lift force

The radial velocity gradient causes an object to lift-off of a surface,
and acts perpendicular to the uid ow. For an object denser than the
uid, the force acts towards the centre of the pipe, away from pipe
walls [15]. The direction of the lift force changes as it crosses the pipe
radius, and is accounted for by the lift coefcient, given as Eq. (10),
with the value of the coefcient greatest in regions with a high
velocity gradient. The shear rate, , is calculated based on the current
position of the rock compared to the previous position of the rock, and
the corresponding slurry velocities at these points, as given in Eq. (11).
The Albertson shape factor [31], given in Eq. (12), has been included in
the calculation of the lift force, as irregular shaped particles
experience a decreased lift force [32]. The lift force was calculated
for the vertical direction using Eq. (13) [25].
Lift coefcient:
CL = 1:16

drock
:
Umax

10

24

K. Albion et al. / Powder Technology 194 (2009) 1832

Shear rate:

Horizontal-direction:

Uslurry;t Uslurry;t 1
:
=
xy;t xy;t 1

11

Albertson shape factor:

s3
= p

s2 s1
5

FL = CL slurry drock 2

12


j U

slurry Urock

j sin

3
2

jrelative j2 :

13

A lift force is also produced due to the Magnus effect. In this

system, particle rotation is slow, and this force has been neglected.
4.1.7. Gravitational force
The gravitational force acts in the downward direction, opposite of
the buoyancy force. Eq. (14) is the gravitational force for a submerged
object.

Fg = Vrock rock g:

14

4.1.8. Force balance

Newton's second law is used for the force balance on the rock.
Eq. (15) is rearranged to solve for the acceleration in each direction,
which is calculated through the use of ordinary differential equations.
The solution to the ordinary differential equation provides the vertical
position, vertical velocity, horizontal position and horizontal velocity
of the rock at any given time.

X
dUrock

:
F = ma = mrock
dt

Fx = FD;x

16

FD;x

ax =
:
mrock

16a

Vertical-direction:
X

Fy = Fg FD;y + FL + FB

ay =

Fg
FD;y
FL
FB

+
+
:
mrock
mrock
mrock
mrock

Fig. 5 illustrates the forces acting on a rock as it travels through the

slurry. The positive force in the horizontal direction is the drag force
and acts in the direction of the ow. Positive forces in the vertical
direction are the lift and buoyancy forces, as they act in the opposite
direction of gravity, and perpendicular to the ow. The force of gravity
and the drag force act in the negative direction. The model was solved
in MATLAB using the 2nd3rd order RungeKutta equation.

17a

An Events function was used in MATLAB to determine when the

rock position was at the top or bottom pipe wall. This function stops
the integration at the wall and changes the direction of the rock,
causing it to rebound.
The model was run under various initial conditions, consisting of
initial rock velocities greater than the rock terminal velocity, and rock
entrance positions spaced throughout the pipe diameter, to determine
the effect of the initial conditions on the number of collisions which
occur along the length of the pipe. Once a collision occurred, a
coefcient of restitution was applied to the new set of initial
conditions. These new initial conditions were the horizontal position
and velocity and the vertical position and velocity at the time of the
collision. However, the vertical velocity has been multiplied by a factor
of 1 to account for the change in direction after the collision. The
coefcient of restitution was calculated using Eq. (18):

Urock;t = e Urock;t 1 :

15

17

18

The coefcient of restitution, e, was experimentally determined to

be 0.7 by dropping rocks onto a steel plate in a vertical column of
water. Videos were taken of these experiments, and the time and
distances were recorded, which allowed for determining the velocities
before and after the rock impacted the steel plate. The coefcient of
restitution was then calculated from the ratio of velocities of the rock
before and after the collision.
A trajectory height threshold was used to determine the number of
collisions. In this case, the threshold value was set to 0.0025 m from

Fig. 5. Force balance on a rock in the slurry.

K. Albion et al. / Powder Technology 194 (2009) 1832

the bottom pipe wall. If the maximum trajectory height was greater
than the height threshold value, a collision is counted. If the maximum
trajectory height between successive collisions is less than the
threshold, a collision is not counted, as it is assumed that this collision
does not make a signicant sound, and the rock is considered to be
rolling along the bottom of the pipe. This was veried by kurtosis
analysis of acoustic measurements of rocks dropped from various

25

distances and hitting a steel plate. It was found that rocks hitting the
plate from a distance of a few millimetres did not cause enough of a
sound for recognition as a collision. Setting this threshold value
between 0.002 m and 0.004 m does not affect the number of collisions
predicted by the model.
As well, a kinetic energy threshold at the time of the collision was
used for the smaller rocks (Rocks E, F, G, H and M) to determine the

Fig. 6. Raw signal of Rock C in the hydrotransport system at a slurry velocity of 3 m/s and concentration of 30 wt.% and different locations along the horizontal line: a) 0.03
m, b) 0.50 m, c) 1.00 m, d) 1.50 m, e) 2.00 m, f) 2.50 m.

26

K. Albion et al. / Powder Technology 194 (2009) 1832

number of collisions. For each rock, the kinetic energy was calculated
using the rock velocity at the time of impact. If the kinetic energy
was greater than 0.0003 J, it was assumed that the rock had enough
energy for a sufcient sound to occur for a collision to be detected
by a microphone. Setting this threshold value between 0.00025 J
and 0.0004 J does not affect the number of collisions predicted by the
model.

5. Results and discussion

5.1. Experimental data the raw acoustic signal
Fig. 6 shows the raw acoustic signals at different locations along the
horizontal hydrotransport pipe. Fig. 6a through f corresponds to measurements recorded at 0.03, 0.50, 1.00, 1.50, 2.00 and 2.50 m respectively along

Fig. 7. Kurtosis of Rock C in the hydrotransport system at a slurry velocity of 3 m/s and concentration of 30 wt.% and different locations along the horizontal line: a) 0.03 m, b) 0.50 m,
c) 1.00 m, d) 1.50 m, e) 2.00 m, f) 2.50 m.

K. Albion et al. / Powder Technology 194 (2009) 1832

27

Fig. 8. Trajectories of Rock C in the horizontal hydrotransport pipe at a slurry concentration of 30 wt.%, initial rock velocity of 1.5 m/s and initial rock height of 0.025 m at slurry
velocities of 2, 3 and 3.5 m/s.

the horizontal line. The rock resulting in this signal was Rock C, which is a
larger sized rounded rock (Table 2, Fig. 4), at a slurry velocity of 3.0 m/s and
a concentration of 30 wt.%.
As shown by the signals in Fig. 6, regular pulses exist in the signal,
which correspond to the diaphragm pump as it pumped the slurry
through the system with a cycle time of approximately 1 s.
A series of additional, much narrower peaks occurred at all microphone locations between approximately 5.5 and 6.5 s. These peaks
indicate the passage of a rock through the pipe. Peaks in the acoustic
signal are a function of the rock size and of the distance between the
impact location and the microphone. On the average, larger rocks and
impacts closer to the location of the microphone were found to result
in larger peaks, whereas smaller rocks and impacts further from the
microphone location caused smaller peaks. However, a small rock
hitting the wall near the microphone caused a larger peak than a large
rock hitting the wall far from the microphone: the peak size decreased
with an increase in distance from the microphone.
The progression of the rock can be seen in the raw signal. In Fig. 6a,
the rock is at 0.03 m at approximately 5.5 s, and by 6.5 s, it has moved

to 2.50 m from the elbow as shown in Fig. 6f. As expected, it was found
that at higher slurry velocities, rocks have a shorter residence time in
the pipe than at lower slurry velocities.

Fig. 9. Effect of initial rock entrance position on the number of collisions at an initial
rock velocity of 1.5 m/s and slurry velocity of 3 m/s.

Fig. 10. Effect of initial rock velocity on the number of collisions at an initial rock
entrance position of 0.025 m from the bottom of the pipe and a slurry velocity of 3 m/s.

5.2. Experimental data kurtosis of the raw acoustic signal

Fig. 7 shows that the kurtosis peaks correspond to the peaks
observed in Fig. 6, between approximately 5.5 and 6.5 s. Again, Fig. 7a
through f corresponds to the acoustic probes at 0.03, 0.50, 1.00, 1.50,
2.00 and 2.50 m in the horizontal line. The kurtosis peaks were
calculated using a time interval of 0.010 s.
Comparing the signals in Fig. 7 with the signals in Fig. 6, the
occurrence of the peaks is the same, and the magnitude of the peaks is
similar: bigger peaks in the raw signal correspond to larger peaks of the
calculated kurtosis. Again, bigger peaks correspond to rock collisions
with the wall closer to the microphone location. There is generally a
large peak with smaller peaks on either side which indicate that the
microphone records signals from the rock as it approached as well as it
travelled away from the microphone location.

28

K. Albion et al. / Powder Technology 194 (2009) 1832

Fig. 11. a. Comparison of the number of collisions counted by kurtosis to the number of collisions predicted by the model for all rocks, at all slurry concentrations, initial rock velocities
and an initial rock position of 0.025 m from the bottom of the pipe at a slurry velocity of 2 m/s. b. Comparison of the number of collisions counted by kurtosis to the number of
collisions predicted by the model for all rocks, at all slurry concentrations, initial rock velocities and an initial rock position of 0.025 m from the bottom of the pipe at a slurry velocity
of 3 m/s. c. Comparison of the number of collisions counted by kurtosis to the number of collisions predicted by the model for all rocks, at all slurry concentrations, initial rock
velocities and an initial rock position of 0.025 m from the bottom of the pipe at a slurry velocity of 3.5 m/s.

While a comparison of Figs. 6 and 7 show that the same peaks

corresponding to the stone impacts were detected, they are much
clearer in Fig. 7 than in Fig. 6. It is to be noted that the pulses in the
signal caused by the pump are not present in the calculated kurtosis
signal. The rock can be detected using kurtosis even if the rock cannot
be distinguished from the pump noise in the raw acoustic signal.
The experimental data used for model verication was the number
of collisions counted by kurtosis from the 6 acoustic probes located

along the horizontal pipe at 0.5 m intervals. These collisions are

counted for slurry velocities of 2, 3 and 3.5 m/s and slurry
concentrations of 10, 30 and 50 wt.%.
5.3. Modelled trajectory of the rock
Fig. 8 shows the motion of Rock C at a slurry concentration of
30 wt.% and at slurry velocities of 2, 3 and 3.5 m/s. Initial conditions

K. Albion et al. / Powder Technology 194 (2009) 1832

29

Fig. 12. a. Comparison of the number of collisions counted by kurtosis to the number of collisions predicted by the model for all oversized rocks, at all slurry concentrations, initial rock
velocities, and an initial rock position of 0.025 m from the bottom of the pipe at a slurry velocity of 2 m/s. b. Comparison of the number of collisions counted by kurtosis to the number
of collisions predicted by the model for all oversized rocks, at all slurry concentrations, initial rock velocities, and an initial rock position of 0.025 m from the bottom of the pipe at a
slurry velocity of 3 m/s. c. Comparison of the number of collisions counted by kurtosis to the number of collisions predicted by the model for all oversized rocks, at all slurry
concentrations, initial rock velocities, and an initial rock position of 0.025 m from the bottom of the pipe at a slurry velocity of 3.5 m/s.

used in this plot were an initial rock velocity of 1.5 m/s and an initial
rock position of 0.025 m from the bottom pipe wall. It is shown in
Fig. 8, that the slurry velocity has a signicant effect on the number of
collisions and the trajectory length of the rock. At a slurry velocity of
2 m/s, many collisions occur along the length of the pipe, however at a
position of 2.5 m from the elbow, the rock is rolling along the bottom
of the pipe, and does not make sufcient sound for collisions to be
detected. The coefcient of restitution also has a signicant effect, as
the maximum trajectory height decreases along the length of the pipe.

At a slurry velocity of 3 m/s, the rock collides along the length of the
pipe, and does not roll towards the end. The number of collisions at
this velocity is less than at 2 m/s, as the trajectories are longer. At a
slurry velocity of 3.5 m/s there are fewer collisions between the rock
and the pipe wall than at the lower slurry velocities, and the rock does
not experience any rolling along the bottom of the pipe, over the
considered length of pipe.
At the higher slurry velocities, the uid is better able to keep the rock
in suspension, which results in fewer collisions with the pipe wall, and

30

K. Albion et al. / Powder Technology 194 (2009) 1832

Fig. 13. Trajectories of oversized rocks in the horizontal hydrotransport pipe at a slurry concentration of 30 wt.%, initial rock velocity of 2.0 m/s and initial rock entrance position of
0.025 m and a slurry velocity of 3 m/s.

the rock travels further between collisions. At the entrance region of the
pipe, the rock is found to impact against the top and bottom pipe walls.
These collisions are recorded by the microphones located on the side of
the pipe. Acoustic signals were recorded simultaneously at the top,
bottom and side of the pipe, and a collision between the rock and the
pipe wall was recorded by all microphones regardless of their position
around the circumference. The trajectory heights shown in Fig. 8 were
qualitatively veried using videos recorded of rocks travelling at each
slurry velocity through the clear PVC pipe section.
5.4. The effect of the initial conditions on the number of collisions
Fig. 9 shows the effect of the initial entrance position of a rock on
the number of collisions predicted by the model for each slurry
concentration, at an initial rock velocity of 1.5 m/s and a slurry
velocity of 3 m/s for Rock C. It can be seen that there was no effect of
the rock entrance position on the number of collisions at slurry
concentrations of 10 and 30 wt.%, however, at a concentration of
50 wt.%, the number of collisions increased by 1 collision when the
entrance location was near the top of the pipe. Since the number of
collisions at this slurry concentration varied by 1 collision, this
additional collision that was recorded for the entrance position near
the top of the pipe can be the result of an extra collision with the top
pipe wall, when it was released close to this boundary. At the slurry
concentration of 50 wt.%, the number of collisions predicted at rock
entrance positions of 0.01 and 0.025 m match the number of collisions
counted by kurtosis, whereas the number of collisions predicted by
the model at the entrance position of 0.04 m is 1 greater than the
number of collisions counted by kurtosis.
Fig. 10 shows the effect of different initial rock velocities on the
number of collisions between the rock and the pipe wall for an initial
rock position, at the pipe centreline, of 0.025 m from the bottom of the
pipe at a slurry velocity of 3 m/s. At slurry concentrations of 10 and
30 wt.%, the initial rock velocity does not affect the number of
collisions predicted by the model. At a slurry concentration of 50 wt.%,
an additional collision is predicted by the model at the highest initial
rock velocity. Again, this can be the result of an additional collision
with the top of the pipe as the rock enters the horizontal pipe. At the
slurry concentration of 50 wt.%, the number of collisions predicted at
initial rock velocities of 1.0 and 1.5 m/s match the number of collisions
counted by kurtosis, whereas the number of collisions predicted by
the model at the rock initial velocity of 2.0 m/s is 1 greater than the
number of collisions counted by kurtosis.

5.5. Comparison of the number of collisions predicted by the model to the

number of collisions determined through kurtosis
For the model to predict rock motion over the 3 m length of pipe,
initial conditions were required. To generate a range of data, each rock
studied used a combination of rock velocities and rock entrance
positions for initial conditions. The initial rock velocities were chosen
to be greater than the terminal velocity of the rock, and were 1, 1.5 and
2 m/s for slurry velocities of 3 and 3.5 m/s, and were 1 and 1.5 m/s for
the slurry velocity of 2 m/s. The initial rock positions chosen were the
centreline of the pipe at 0.025 m, and at 0.01 and 0.04 m, measured
from the bottom of the pipe wall. These heights were chosen to
determine the effect of the pipe walls on the number of collisions, as
one position was close to the bottom pipe wall, and the other close to
the top pipe wall. As was shown in Figs. 9 and 10, the model was not
very sensitive to the initial conditions, as the number of predicted
collisions varied from the number of counted collisions by 1 collision
under some conditions.
Fig. 11ac shows the number of collisions between the rocks and
pipe walls as counted by kurtosis, compared to the number of
collisions predicted by the model based on the rock trajectory. Using
the model, each time the rock was located at the pipe wall, a collision
occurred, unless the maximum trajectory value was less than a set
threshold value (0.0025 m) as then the rock was considered to be
rolling, and would not generate sufcient sound to be recorded by the
microphone and be considered a collision.
Fig. 11a compares the number of rockwall collisions at a slurry
velocity of 2 m/s, Fig. 11b shows the collision comparison at 3 m/s,
whereas Fig. 11c shows the comparison at 3.5 m/s. These gures were
generated using all the rocks, the three slurry concentrations of 10, 30
and 50 wt.% and all the initial condition combinations. The model was
tested at each slurry velocity, as earlier work into the acoustic
detection of oversized material indicated that the slurry velocity, as
well as the rock size, had a signicant effect on the number of
collisions between a rock and the pipe wall [14].
As shown in Fig. 11ac, there is signicant scatter between the actual
and estimated numbers of collisions for the rocks. In particular, there is
poor agreement for the rocks that are considered to be of an acceptable
size (rock diameter/pipe diameter b25%). This poor agreement can be
attributed to the effect of turbulence on the rock, as these rocks are smaller
and lighter and are affected more by turbulent eddies. As well, rocks of an
acceptable rock size may collide with the pipe wall more frequently, or
they may not possess enough energy for a hard collision against the wall,

K. Albion et al. / Powder Technology 194 (2009) 1832

31

Fig. 14. Modelled trajectories of a large rock in a 27-inch horizontal hydrotransport pipe at a slurry concentration of 50 wt.%, slurry velocity of 3 m/s, initial rock velocity of 1.5 m/s and
varying initial rock entrance heights.

and may not generate a loud sound when they collide to be recorded,
unless a collision is close to a microphone.
It can also be seen from Fig. 11ac, that the number of collisions
counted by kurtosis and estimated by the model decreases as the
slurry velocity increases. This is attributed to the ability of the slurry to
carry the rocks. At low slurry velocities, the slurry cannot carry the
rocks large distances, and they collide more often with the pipe wall,
however, as the slurry velocity increases, the slurry can better suspend
the rocks, and they travel further distances between collisions.
Fig. 12ac compares the number of collisions between the rock and
the pipe wall for oversized rocks only (rock diameter/pipe diameter
N25%). The data used in these gures are the same as the earlier
gures: there is a plot for each slurry velocity, and each plot was
generated using 12 sets of initial conditions for each slurry concentration for all the oversized rocks. The agreement for the oversized rocks is
much better than for the acceptable rock sizes. As well, there is much
overlap of the predicted and counted number of collisions.
In Fig. 12ac, only the oversized rocks were considered. In the
hydrotransport of oil sand, there will be many rocks of an acceptable
size present in the slurry that will not cause damage to the equipment.
It is not as important to detect these rocks, as it is to detect oversized
rocks. Oversized rocks can cause immediate damage to the system, and
ideal probe locations must be identied to detect these rocks so that
preventative action can be taken before signicant damage occurs.
5.6. Rock detection at various positions along the horizontal pipe
Fig. 13 shows the trajectory of oversized rocks along the horizontal
hydrotransport pipe at a slurry concentration of 30 wt.% and a slurry
velocity of 3 m/s, as predicted by the model. Examination of the rock
trajectories and collision locations led to the identication of optimum
microphone locations for reliable detection of oversized rocks for all
slurry concentrations and initial conditions at each slurry velocity.
Ideal locations for rock detection occur, when all rocks collide at
approximately the same location in the pipe, and the sound of the
collision can be detected by the acoustic probe. Fig. 13 shows that the
majority of the collisions are concentrated over a short length of pipe
until approximately 1.0 m from the elbow, under these conditions.
Further from the elbow, collisions occur over a longer length of pipe,
and a microphone in this area would result in fewer rocks detected.
The initial conditions effect the location of a rock collision as the
location of the collision can vary a few centimetres.
The distance sound can travel from a rockwall collision was tested
using pipe sections connected by a tee to form a 90 bend. This section
of pipe was lled with water, and microphones were located along the

length of the horizontal section of pipe. Rocks were dropped into the
vertical section of pipe to impact the elbow from different heights,
which varied the velocity of the rock. The distance between the
microphones was increased to determine the distance when the
sound of the collision was no longer detected by the microphone. It
was found that at higher velocities, the sound of the collision was
detected 30 cm from the collision, whereas at slower velocities, the
sound of the collision was detected 40 cm from the collision.
Examining the trajectories of all the oversized rocks at all concentrations and initial condition combinations, the optimum microphone locations for reliable rock detection at each slurry velocity are:
2 m/s:
0.05, 0.50, 1.00 m
3 m/s:
0.05, 0.50, 1.00 m
3.5 m/s: 0.05, 0.50, 1.00, 1.50, 2.00 m.
The optimum microphone locations vary with slurry velocity, due
to the effect of the velocity on the ability to carry the rock. The
difference in the locations of some of the microphones are due to the
decreasing length of the rock trajectories.
These optimum microphone positions correspond well with the
experimentally determined ideal locations. At slurry velocities of 2
and 3 m/s, 3 microphones were required for 100% detection, and 4
microphones were required for a slurry velocity of 3.5 m/s. At 2 and
3 m/s, the experimentally determined ideal locations were found to
be 0.03, 0.5 and 1.0 m in the horizontal line. At a slurry velocity of
3.5 m/s, the ideal locations were experimentally found to be 0.03, 1.0,
1.5 and 2.0 m in the horizontal line [14].
There is perfect agreement between the optimum rock detection
locations predicted by the model and the experimentally determined
locations at all slurry velocities, as the model predicted locations are all
within the rst metre of the pipe after the elbow for slurry velocities of
2 and 3 m/s, and within the rst 2 m at a slurry velocity of 3.5 m/s.
Fig. 13 shows that ideal microphone locations reliable detection of
oversized rocks, regardless of the slurry concentration or the initial
conditions of the rock, are close to the elbow. As compared to Fig. 8,
microphone placement near the elbow is ideal since many rocks
collide in this area as they change direction after coming through the
elbow and accelerate in the ow.
5.7. Application of the model to industrial conditions
The model was applied to a 27-inch (0.69 m) diameter, horizontal
hydrotransport pipe, to simulate industrial conditions. The length of
the pipe used to calculate the trajectories was 10 m. A large rock was

32

K. Albion et al. / Powder Technology 194 (2009) 1832

used in the model with measured properties which included: Vrock =

0.00075 m3, drock = 0.113 m, m = 1.945 kg, rock = 2590 kg/m3,
Arock = 0.0092 m2, Ut,rock = 1.00 m/s and = 0.73. The model was run
for a slurry concentration of 50 wt.% and a slurry velocity of 3 m/s. The
initial conditions were varied, testing the effect of the rock entrance
position (0.173, 0.345, 0.518 m) and the initial velocities (1.0, 1.5,
2.0 m/s). Fig. 14 shows the effect of the initial rock entrance position on
the trajectory of the rock in the large diameter pipe.
Fig. 14 suggests that the optimum location for microphone
placement is at approximately 1.3 m from the elbow. At all
combinations of initial conditions, the rock collided at 1.3 m. After
the rst collision the lengths of the trajectories varied, and the
collision locations became spread over a wider area. Ideal locations
further along the pipe include approximately 3.3 and 5.8 m, as rock
collisions occur in these areas under all initial conditions. Placement of
microphones at these locations may detect a rock collision as they are
within the distance that sound can be recorded by the microphone.
6. Conclusions
A simple model has been developed to predict the number of
collisions between an oversized rock and the pipe wall, as well as to
predict the location of these collisions in a horizontal hydrotransport
pipe. This was accomplished by using a force balance on the rock,
which accounts for the shape of the rock through the drag and lift
coefcients and forces.
The model was veried by acoustic signals recorded at 0.5 m
intervals along the length of the horizontal pipe. Kurtosis was used to
analyse the acoustic signals; a peak exists in the kurtosis signal where
a rock was recorded colliding with the pipe wall by the microphones.
The number of peaks counted in the kurtosis signal is the number of
collisions recorded between the rock and the pipe wall.
There was good agreement for the number of collisions between
the rock and the pipe wall as determined by the model, and the
number of collisions counted by kurtosis. This was examined at each
slurry velocity for the oversized rocks. The agreement between the
predicted and actual number of collisions for acceptable sized rocks
was poor, however, these rocks do not cause damage to equipment
and are not as important to detect as the oversized rocks. The model
was found not to be particularly sensitive to initial conditions, as
changing the initial rock velocity or rock position only varied the
number of collisions by one collision in some cases.
To reliably detect oversized rocks in a hydrotransport system, it is
important to determine the ideal microphone locations. Ideal
microphone locations were determined for each slurry velocity,
under all slurry concentrations and combinations of initial conditions.
For slurry velocities of 2 and 3 m/s, 3 microphones are required for
100% detection of oversized rocks. At a slurry velocity of 3.5 m/s, 4
microphones were required for 100% detection of oversized rocks. It is
important to have the ideal microphone locations to detect oversized
rocks as not detecting an oversized rock will lead to immediate
damage of equipment resulting in downtime and loss of production.
Detection of a large rock allows corrective actions to be taken before
damage occurs.
Acknowledgements
The authors would like to thank the Natural Sciences and
Engineering Research Council of Canada and Syncrude Canada Ltd.
for their nancial support of this research.

References
[1] A. Fair, J. Oxenford, J. Coward, M. Lipsett, Technology visions for mining at Syncrude,
CIM Bull. 92 (1999) 113121.
[2] R.B. Paine, B.M. Wright, Understanding hydrotransport: the key to Syncrude's
success, CIM Bull. 92 (1999) 105108.
[3] M. Valenti, Oil Sand Kicks Into High Gear, Mechanical Engineering Magazine,
1998 http://www.memagazine.org/backissues/december98/features/oilsand/
oilsand.html.
[4] R.S. Sanders, J. Schaan, R. Hughes, C. Shook, Performance of sand slurry pipelines in
the oil sands industry, Can. J. Chem. Eng. 82 (2004) 850857.
[5] N. Chien, Z. Wan, Mechanics of Sediment Transport, ASCE Press, Virginia, 1999.
[6] C.A. Shook, R.G. Gillies, R.S. Sanders, Pipeline Hydrotransport with Applications in
the Oil Sand Industry, Saskatchewan Research Council, Saskatoon, 2002.
[7] P.D. Osborne, Transport of gravel and cobble on a mixed-sediment inner bank shoreline of a large inlet, Grays Harbor, Washington, Mar. Geol. 224 (2005) 145156.
[8] R.I. Ferguson, D.J. Bloomer, T.B. Hoey, A. Werritty, Mobility of river tracer pebbles
over different timescales, Water Resour. Res. 38 (2002) 18.
[9] F.-C. Wu, Y.-J. Chou, Rolling and lifting probabilities for sediment entrainment,
J. Hydraul. Eng. 129 (2003) 110119.
[10] T.G. Drake, R.L. Shreve, W.E. Dietrich, P.J. Whiting, L.B. Leopold, Bedload transport
of ne gravel observed by motion-picture photography, J. Fluid Mech. 192 (1988)
193217.
[11] J.W.R. Boyd, J. Varley, The uses of passive measurement of acoustic emissions from
chemical engineering processes, Chem. Eng. Sci. 56 (2001) 17491767.
[12] M. Valentino, Microphone handbook, PCB Piezotronics Vibration Division, 2005,
http://www.pcb.com/Linked_Documents/Vibration/Microphone_Handbook.
pdf.
[13] L.W. Flott, Quality control: nonnormal frequency distributions, Met. Finish. 93
(1995) 5255.
[14] K. Albion, L. Briens, C. Briens, F. Berruti, S. McDougall, Detection of oversized
material in a hydrotransport slurry pipe using a non-invasive acoustic method.
Powder Technol. 190 (2009) 361371.
[15] C.A. Shook, M.C. Roco, Slurry Flow: Principles and Practice, ButterworthHeinemann, Toronto, 1991.
[16] B. Abulnaga, Slurry Systems Handbook, The McGraw-Hill Companies, Inc., New
York, 2002.
[17] L.-S. Fan, C. Zhu, Pneumatic Conveying of Solids in Principles of GasSolids Flows,
Cambridge University Press, United Kingdom, 1998.
[18] F.J. Cabrejos, G.E. Klinzing, Incipient motion of solid particles in horizontal
pneumatic conveying, Powder Technol. 72 (1992) 5161.
[19] K.C. Wilson, A. Sellgren, Interactions of particles and near-wall lift in slurry
pipelines, J. Hydraul. Eng. 129 (2003) 7376.
[20] P. Gondret, M. Lance, L. Petit, Bouncing motion of spherical particles in uids, Phys.
Fluid 14 (2002) 643652.
[21] W. Lick, L. Jin, J. Gailani, Initiation of the movement of quartz particles, J. Hydraul.
Eng. 130 (2004) 755761.
[22] J. Wang, S. Shirazi, A CFD based correlation for erosion factor for long-radius elbow
and bends, Trans. ASME 125 (2003) 2634.
[23] Y. Nino, M. Garcia, Gravel saltation: 2 modeling, Water Resour. Res. 30 (1994)
19151924.
[24] B.B. Willetts, The inuence of lift due to circulation on saltation and suspension of
solid particles in streams, Hydrotransport 1: First International Conference on the
Hydraulic Transport of Solids in Pipes, Coventry, England, vol. H2, 1970, pp. 1736.
[25] A.D. Salman, D.A. Gorham, M. Szabo, M.J. Hounslow, Spherical particle movement
in dilute pneumatic conveying, Powder Technol. 153 (2005) 4350.
[26] R.G. Gillies, M. McKibben, K.B. Hill, C.A. Shook, Sand transport by Newtonian uids
in laminar pipe ow, Powder Technol. 104 (1999) 269277.
[27] C.J. Geankoplis, Transport Processes and Unit Operations, 3rd EditionPrentice-Hall
Inc., Toronto, 1993.
[28] K. Albion, L. Briens, C. Briens, F. Berruti, Flow regime determination in horizontal
hydrotransport using non-intrusive acoustic probes. Can. J. Chem. Eng. 86 (2008)
9891000.
[29] M.C. Roco, Particulate Two-Phase Flow, Butterworth-Heinemann, Toronto, 1993.
[30] C.E. Lapple, C.B. Shepherd, Calculation of particle trajectories, Ind. Eng. Chem. 32
(1940) 605617.
[31] M.L. Albertson, Effects of shape on the fall velocity of gravel particles, 5th Iowa
Hydraulic Conference, Iowa University, Iowa, 1953.
[32] E. Gavze, M. Shapiro, Particles in shear ow near a solid wall: effect of nonsphericity on forces and velocities, Int. J. Multiph. Flow 23 (1997) 155182.