Parametric Modeling of Wear in a Slurry Pipeline with Bed Flow
S. El Sayed, M.G. Lipsett*
Department of Mechanical Engineering
University of Alberta, Edmonton
Alberta, Canada T6G 2G8
ABSTRACT
Slurry transport is a key process in a number of industries. In transporting mined oilsand,
slurry pipelining promotes conditioning to release and aerate bitumen prior to separation
from water and solids. Reliability of slurry transport pipelines is a major ongoing
problem for operating oilsands companies due to unexpected piping failures. To date, no
accurate model has been developed to predict wear rates in slurry transport pipelines,
although previous studies have shown that some process variables are important such as
flow rate, slurry density, and particle size distribution.
This work investigates erosion wear mechanisms causing inner pipe wall wear due to
sand slurry flow in a horizontal section of pipe under steady-state conditions. A lumpedparameter erosion wear model is presented based on a simplification of the physics of
oilsands slurry flow. An apparatus was developed and tested to measure the forces acting
on the pipe inner wall to monitor forces related to erosion in a laboratory-scale sand
slurry loop. Preliminary results are presented with some recommendations for future
work that would be required to validate the model.
Keywords: Slurry wear, pipelines, modeling, prognosis
Introduction
Alberta oilsands deposits comprise one of the world’s largest oil reserves. Significant
investments have been made in oilsands mining and processing operations, which have
resulted in oilsands-derived synthetic crude oil supplying 60% of Canada’s needs.
Oilsands contain mixture of bitumen, sharp-edged silica sand with average diameter of
approximately 120 um, clay fines of less than 44 um diameter, and water. The bitumen
content ranges between 1% and 20% depending on the quality of the ore body.
Shallow oilsands deposits are mined using front shovels and transported using large offroad haul trucks to a collection point. At this location, the oilsands are crushed and
screened to remove oversized rocks and any foreign material that may harm equipment
downstream. Warm treated water is added to the mixture, and the resulting oilsands
slurry is transported by pipeline to a separation plant. During slurry transport, turbulent
mixing causes entrained air bubbles to collide with bitumen droplets that have been
liberated from the oilsand. In favourable process conditions, the bubbles attach to the
bitumen, which promotes efficient separation of bitumen from water and solids. Sand and
fine tailings are subsequently transported from the plant to settling basins. An oilsands
*
Corresponding author. Email: mlipsett@ualberta.ca
mining operation typically employs approximately 90 km of slurry piping, with up
to10,000 tonnes per hour of oilsands being transported in a single pipeline.
Wear is the dominant damage mechanism for steel slurry pipelines transporting oilsands.
Any failure to a pipeline component results in downtime and lost production.
Conservative maintenance strategies are used to reduce risk of failure, because mean time
between failures is not predictable due to variable operating conditions and ore types. As
well, there is a lack of understanding of the wear mechanisms affecting oilsands slurry
pipelines. The current industry practice is to predict wear in a hydrotransport pipeline
based on a linear relationship relating pipeline metal loss to the number of operating
hours, based solely on prior operating experience but not the variability of process
conditions. Consequently, operating costs are substantially increased, with much piping
removed from service with some remaining usable wall thickness, and yet sudden
pipeline failures still occur on occasion.
For these reasons, a predictive model for hydrotransport pipeline wear rate would help
maintainers to schedule maintenance activities more cost effectively and reduce the
occurrence of sudden failures.
1.1 Oilsand Slurry Properties
A slurry of oilsand in water tends to be heterogeneous, with a significant fraction of its
particles settling to the bottom of a container due to their own weight. The particle size
distribution of oilsand varies depending on the mined oilsand location, with sand particles
generally larger in deeper deposits. Transported oilsands slurries have an average
concentrated of approximately 35% by volume, and the particle size diameter varies
between 0.18 to 0.3mm [1]. The flow regime of heterogeneous slurries varies with the
slurry average flow velocity. In order to prevent any accumulation of solid particles at the
bottom of the pipeline, the mixture average velocity must exceed the slurry deposition
velocity. Oilsands slurries are transported above deposition velocity at velocities between
3 m/s and 5.5 m/s with the minimum deposition velocity being around 2 m/s [1]. As the
velocity of the mixture is increased, more particles are suspended by fluid turbulence, and
as a result the velocity and concentration profile across the pipe diameter changes as well.
Slurry pipeline wear is complex, due mainly to three wear processes: erosion, corrosion,
and combined erosion-corrosion. Slurry pipeline wear is difficult to predict because of
the challenges to isolate and quantify the contribution of each of the wear processes and
their synergistic effects. Erosion is usually the dominant wear process, with corrosion
playing a minor role, especially in fresh water slurry applications. On the other hand, the
combined erosion-corrosion wear process can be the dominant wear process in some
slurry applications. The combined erosion-corrosion wear process becomes dominant
when there is continuous removal of pipe wall material due to erosion and as a result
exposing new material to corrode [2].
The wear profile around the circumference of slurry piping may not be uniform,
especially for heterogeneous slurry flow, likely due to the changes on concentration and
flow conditions in different regions of the cross-section. There may also possibly be
changes in chemistry near the wall, particularly in parts of a dense bed where there is not
a lot of mixing.
Oilsands hydrotransport pipeline wear measurements are taken regularly as part of
equipment monitoring programs. In one case, the maximum wear rate reported in a
straight section of 76 cm diameter low-carbon steel pipe occurred at the bottom of the
pipe, with average velocity of 4.5m/s, bitumen content of between 11% and 12%, and
fines content between 21% and 26%. The material loss rate was mainly due to bottom
bed friction where the slurry bed concentration was believed to be a main contributor [1].
Similarly, Roco and Addie reported the maximum wear rate to be at the bottom of the
pipe in a sand slurry test loop, with pipe diameter dpipe = 200mm and concentration of
10% by volume [3]. Gupta, Singh, and Seshadri reported on the uneven wear rate in a
brass-pipe / sand-slurry test loop, with average velocity between 1.95 to 2.75 m/s and
concentration between 17.23 to 34.5% by weight and a wide particle size distribution
range; again, the maximum wear rate was found to be at the bottom of the pipe [4].
There have been several attempts to model slurry pipeline wear, using either empirical
models or energy approaches.
An empirical model based on a large number of tests was presented by Salama to predict
the erosion wear rate in multiphase gas/liquid well production pipelines [5]. Similarly,
Gupta, Singh, and Seshadri developed an empirical model to predict the erosion wear rate
in a slurry brass test loop around the pipe circumference [4]. The empirical correlation
produced was a function of the average slurry velocity, average particle size, and average
slurry in-situ concentration. The empirical correlation values developed for wear rates
appear to apply only to narrow ranges of operating conditions and pipe materials. No
insight was provided to the wear mechanisms affecting slurry pipelines; and, as a result,
these equations cannot be used with confidence to predict wear in slurry pipelines
operating at different conditions.
Other researchers have used computer simulations to predict uneven wear rates in slurry
pipelines. For example, Roco and Addie divided erosion wear mechanisms in a
heterogeneous slurry pipeline into directional impact, random collision, and Coulombic
friction. Directional impact was found to be dominant in elbows and bends, while
Coulombic friction was dominant in flows with a dense bed. A small-scale test loop was
constructed to verify results based on simulated near wall particle velocities and
concentration profile, for particle diameter dparticle = 0.5mm and concentration of 10% by
volume [3]. Similarly, Wood, Jones, and Miles used computer simulations to predict
wear in a 78mm straight stainless steel pipe with slurry flow at 10% slurry concentration
[6].
The key contribution in the work of Roco et al. is the mechanistic approach utilized,
although the contribution of corrosion and erosion/corrosion was not isolated from the
erosion wear mechanisms. Consequently, one can not call the model presented a pure
erosion model. There is room for improving erosion models of slurry pipelines. The work
of Wood et al. was limited by the computer power available and the number of particles
included in the simulation [6]. In case of high slurry concentrations, more powerful
simulations are needed. Numerical simulations of dense slurries remain challenging to
generate and validate especially with broad ranges of particle size distributions and in
larger pipe diameters.
Since wear in oilsands hydrotransport pipelines is a complex process that is difficult to
observe, an alternative approach is necessary to reduce the complexity of the wear model
by tackling each of its aspects separately. The main objective of this work is to introduce
a simple parametric model for erosion wear based on the physics of sand slurry flow in a
straight section of pipe at steady-state conditions. While there are good models of
heterogeneous slurry flow, there is uncertainty about the forces associated with wear in a
slurry pipeline. An apparatus was designed and tested to measure the overall forces on a
section of pipe wall, due to fluid-wall and particle-wall interactions.
Sand Slurry Erosion Wear Model Development
In the improved two layer model presented by Gillies, Shook, and Wilson, heterogeneous
slurry flow at moderate flow velocities was divided into two layers [7].
In the top layer, particles are fully suspended by mixing turbulence where there is only
kinematic friction. In the bottom layer, a fraction of the particles is fully suspended by
turbulence. The remaining fraction of particles is supported through contact by pipe wall.
As a result, both kinematic friction and Coulombic friction exist in the bottom layer. The
idealized concentration and velocity profile in the two-layer model of heterogeneous
slurry flow is shown in Figure 1.
Figure 1. Two layer model concentration and velocity profiles [9]
A simple lumped-parameter flow model was developed for a straight section of an
oilsands hydrotransport pipeline, based on the improved two layer model for high
concentration slurries at moderate flow velocities [8]. A graphical representation of the
model presented is shown in Figure 2.
Figure 2. Physical system model of oilsands slurry pipeline section [8]
This model uses the convention of dividing the flow in a straight section of a dense slurry
pipeline into two layers: a top layer and a bottom layer. In both layers it was assumed that
the flow processes occur independently and in sequence [8]. It was also assumed that the
flow conditions are steady-state. Three processes occur in the top layer. There is top layer
friction represented by resistance component R4, by top layer mixing represented by R5,
and bitumen conditioning in the top layer is represented by resistance component R6. The
flow is each process is assumed to be the same, and resistance is with respect to the
average flow in the layer rather than any near-wall effect [8].
Three flow processes occur in the bottom layer. There is bottom layer friction represented
graphically by resistance component R1 followed by bottom layer digestion represented
by R2. Mixing in the bottom layer is represented by R3.
There may be some transfer of material between the bottom and top layers. During
transportation of oilsands in hydrotransport pipelines, larger particles (also known as
lumps) lose cohesion and ablate or even break apart; smaller particles are released into
the flow. Some smaller particles are transferred to the top layer due to carrier fluid
turbulence. In the case of oilsands, bitumen in the bottom layer gets released during
transportation and it is also carried by carrier fluid turbulence to the top layer. If aeration
occurs, the bitumen has a net migration from the bed to the less dense top layer. This
transitional flow process between the bottom layer and the top layer is represented by
flow component R7 [8].
The pressure difference between the top and bottom layers at the upstream end of the
horizontal pipe section is represented by Ph, and downstream by Pl [8]. PA, PB, and PD
represent the average pressure in the top layer of the upstream, intermediate and
downstream locations of the pipe section [8]. PC represents the average pressure in the
bottom layer at an intermediate state where bitumen and digested particles flow to the top
layer [8].
The oilsands hydrotransport pipeline lumped parameter flow model can be simplified
further to represent sand slurry flow in a straight section of pipe. A graphical
representation of the simplified sand slurry flow model can be seen in Figure 3.
Resistance components related to bitumen conditioning can be safely neglected in sand
slurry flow since very little bitumen is present in the sand slurry [10], and so resistance
component R6 is removed. In addition, a sand slurry does not contain as many large
lumps as an oilsands slurry. As a result, resistance component R2 representing solid
particle digestion and bitumen conditioning in the bottom layer can also be safely
neglected in a sand slurry. (During experiments to determine the coefficient, sand
particles shape and size would be monitored over the course of each experiment to ensure
that this assumption is valid.) Resistance component R7 can also be removed, because no
particle digestion and no bitumen conditioning occurs in the bottom layer [10].
Consequently, it can be assumed that no transfer of solid particles or bitumen occurs
between the bottom layer and the top layer.
Figure 3. Simplified physical model of sand slurry pipeline section [10]
As seen from Figure 3, the sand slurry flow model is similar to the oilsands
hydrotransport flow model, in that the flow is divided into two layers: a top and a bottom
layer. Sand particles are fully suspended in the top layer due to carrier fluid turbulence.
Some particles are fully suspended in the bottom layer due to carrier fluid turbulence,
with the remaining particles directly supported by the pipe wall. The latter fraction of
particles slides against the bottom of the pipe wall forming a sliding bed. Pressure drop in
the top layer is mainly due to top layer kinematic stress and top layer mixing flow
processes. These flow processes are graphically represented in figure 2 by resistance
components R4 and R5. Similarly, pressure drop in the bottom layer is attributed to bottom
layer kinematic and Coulombic friction represented by R1 and bottom layer mixing
represented by R3. To isolate the contributions of kinematic friction from Coulombic
friction, resistance component R1 can be split into two components R11 and R12
representing kinematic and Coulombic friction respectively [10].
There are two flow processes in each of the layers in sand slurry flow; and energy is
dissipated in each of the resistance components. For example, kinetic energy is dissipated
by the kinematic friction component in the top flow layer. Kinetic energy is dissipated
due to the mixing flow process. A fraction of the energy dissipated by friction is lost in
the form of heat, which heats up the slurry flow and raises its temperature without
causing any damage to the pipe inner wall.
There may be another fraction of the friction energy that causes damage to the pipe if the
energy transfer exceeds a threshold for the pipe material. Part of the energy lost in
suspending the slurry particles in the top layer or mixing energy causes some particles to
exit the flow streamlines and impact randomly on the straight pipe inner wall. A fraction
of the energy transmitted to the pipe inner wall by these random impact causes damage to
the pipe [10]. These random impacts hit the pipe wall at low angles [6]. A similar process
occurs in the bottom layer of flow; however, in the bottom layer, the energy lost due to
friction can be divided into kinematic and Coulombic friction terms.
The wear contribution s′ of each of the wear components was described by Roco and
Addie [3] as:
s '   * [V * (   o )] ,
(1)
where wall shear stress and wear coefficient measurements are necessary to quantify the
contribution of each of the wear mechanisms on the inner pipe wall for both top and
bottom layers. In a slurry, there may be a difference between the bulk velocity of the fluid
in the region and the effective velocity near the wall that results in damage accumulation.
As a result, wear coefficients can be introduced to an energy model with a coefficient 
proportional to the energy Ed , which is the energy transferred to the pipe above some
threshold, and , which attenuates the damage due to a difference between process
velocity and local velocity of impacts. For the different processes in the pipe section there
will be a set of damage terms, as illustrated in Figure 4. The overall erosion wear rates in
a straight section of sand slurry pipeline can be estimated if the contribution of each of
the erosion wear mechanisms can be quantified.
Figure 4. Simplified wear damage model of sand slurry pipeline section
The total energy dissipated in the system can be expressed as:
2
2
Pt1345
P1345
Pd21345


2Rt1345 2 R1345 2Rd1345
(2)
2
Pt1345
U total 
2Rt1345
U threshold
U damage 
(3)
2
P1345
2R1345
(4)
Pd21345
2 Rd1345
(5)
The energy contributing to damage can therefore be expressed as:
U d1345 
2
Pt1345
P 2
 1345
2 Rt1345 2 R1345
(6)
The damage mechanism is related to damage of the pipe wall by a wear coefficient; and
the damage rate can be expressed as:
2
2
Pt1345
P1345
s  1345Ed1345  1345(U d1345 / area)  1345[(

) / area]
2Rt1345 2R1345
(7)
'
s  1345 * [V * ( 1345   o1345 )]
'
(8)
Model Validation Issues
In order to validate the proposed model it is necessary to evaluate the parameters R, α, β,
and E d for each of the erosion wear mechanisms. It is also required to isolate each of the
wear mechanisms so as to calculate the corresponding parameters of interest. Isolation of
the kinematic friction and random impact of particles from the Coulombic friction can be
achieved in a fully suspended heterogeneous slurry flow by increasing the flow velocity.
In a fully suspended flow condition, a symmetric concentration and velocity profile exists
in the pipe, and Coulombic friction can be ignored. The resulting flow no longer consists
of two layers, but rather fully suspended slurry flow within the entire pipe. In that case,
the kinematic friction and random impact of particles can be lumped together into a
single equation: q0 = R1345P1345, as illustrated in Figure 5. This relationship can then be
exploited later when validating the flow model for dense bed cases, when the velocity of
the homogeneous slurry case matches that of the top layer for two-layer conditions.
ΔP1345
q0
R1345
Figure 5. Simplified physical model of sand slurry pipeline section for fully suspended
flow
The resistance coefficient R1345 corresponding to the kinematic friction and random
impact of particles due to slurry mixing in the fully suspended slurry flow can be
determined for a variety of flow conditions by measuring the pressure drop across a
laboratory test section in a slurry loop. By controlling the slurry flow rate, the in situ
particle concentration, density of particles, carrier fluid viscosity, and particle size
distribution, a variety of near wall particle velocities and concentrations around the pipe
circumference can be achieved. A set of R1345 values can be determined for a range of
flow conditions. At this stage, it is important to prevent damage accumulation from
occurring in the pipe by covering the pipe with an abrasion-resistant overlay. Care must
be taken to choose an overlay material that has the same roughness as the pipe itself. This
will ensure that no energy is lost due to pipe wall erosion wear damage process, while
still maintaining the same head loss due to friction.
Erosion damage in a fully suspended heterogeneous slurry flow condition causes
additional pressure drop in the piping system above a threshold point which can be
represented as shown in 6 [12].
ΔP1345
q0
R1345
ΔPd1345
Rd1345
Figure 6. Parametric representation of sand slurry flow and erosion damage for
fully suspended flow
The damage caused is a function of the pressure drop and the flow rate, which is also a
function of velocity and concentration [12]. The threshold energy for incipient pipe wear
can be estimated for a variety of flow conditions by coating the pipe wall with a thin
surface stain. Flow can be increased until the stain paint layer begins to be ablated to
determine the threshold point for incipient wear. Another potential method to determine
the threshold energy for incipient wear is by recording acoustic emissions in the vicinity
of the pipe [13]. The amplitude of some features in the acoustic signature for damage
should correlate well with the threshold stress for incipient wear; but the acoustic
emission method needs to be tested in a slurry application to identify the unique
signatures at which damage occurs.
By measuring the total pressure drop in the pipe section, velocity, and concentration at
the same flow conditions used to find R1345 values, Rt1345 values can be determined. The
total energy in the system and the threshold energy for incipient wear can then be
computed using Equation 3.
Also, by measuring the pipe erosion damage rate using an ultrasonic thickness
measurement device, and the stress experienced by the pipe wall, and the threshold stress
for incipient wear, the erosion wear coefficient α1345 can be determined using Equation 8.
Once the threshold for incipient wear, α1345 values, R1345 values, and Rt1345 values for a
wide range of flow conditions are determined, the contribution of kinematic friction to
erosion pipe wear in the bottom layer of heterogeneous sand slurry flow can be
determined. Similarly, erosion damage in heterogeneous slurry flow due to Coulombic
friction causes additional pressure drop in the piping system which can be represented as
shown in Figure 7. Since kinematic friction, Coulombic friction, and random impact of
particle due to bottom slurry mixing do not occur sequentially, isolation of Coulombic
friction contribution to pipe erosion damage is difficult.
ΔP12
ql
ΔPd12
R12
Rd12
Figure 7. Parametric representation of sand slurry flow and erosion damage for sliding
bed
In order to determine α12, an ASTM G65 rubber wheel test can be used, or another
similar wear test that can emulate process conditions, such as the oscillating table test [3]
and the wet sand rubber wheel test [14]. The ASTM G65 test [14] is a standard wear test
in the oilsands industry because it yields fairly repeatable relative wear results, provided
that the solid particle distribution used in the test resembles that in pipe flow conditions,
and the normal force of the particle and the tangential speed of the wheel are close to how
a particle slides along the actual pipe. The apparatus is illustrated in Figure 8.
Once-through flow
of abrasive solids
Figure 8. A simplified schematic illustration of the ASTM G65 test.
By knowing the suspended arm weight, the sliding friction shear stress can be calculated
by multiplying the normal force by a dynamic friction factor (usually assumed to be 0.5
for sand particles). The velocity of the flowing sand particles can be assumed to be equal
to the velocity at the tip of the rubber wheel. Therefore, the wear coefficient for sliding
wear in the bottom layer can be calculated for the slurry/material combination, since the
wear rate experienced by the specimen can be measured directly. Once the threshold
conditions for incipient pipe wear are determined for a range of wheel velocities and
normal forces imposed by the weight, a curve of the wear coefficients calculated at
different particle velocities can be constructed and used to estimate wear in the bottom
layer of the pipe due to particle sliding abrasion.
To predict wear due to sliding bed abrasion, wall stress measurements need to be
performed around the circumference of pipe. The contribution of kinematic friction to
damage around the circumference of the pipe by knowing α1345 values, R1345 values, and
Rt1345 values obtained in the fully suspended flow tests. The velocity and concentration
profile across the pipe wall should be also developed in order to isolate the contribution
of kinematic friction to pipe wall damage. To understand local wall effects, additional
flow monitoring and visualization techniques are required to measure the near wall
velocity and concentrations of suspended particles and sliding bed. Advanced flow
visualization techniques may include a high-speed camera estimating change in particle
speed near the wall at a viewing section (which will admittedly not likely yield realistic
results for the actual pipe material) or a boroscope introduced into the flow (which will
affect the flow and will require calibration); but a correlation based on the estimate of
bulk velocity in the layer may be sufficient for
The threshold stress for incipient wear due to Coulombic friction wear mechanism can be
determined by measuring the stress required to start wear in the laboratory-scale tests.
Once the stress, threshold stress for incipient wear, particle sliding velocity, and wear
coefficient are determined, a model for erosion due to Coulombic forces can be validated
and used for estimating erosion wear rates around the circumference of the pipe during
heterogeneous slurry flow.
Experimental Setup and Testing Methodology
Experiments are needed to find the constitutive relationships for each of the elements of
the wear model. A slurry loop of 2-inch inner diameter has been constructed and
commissioned at the University of Alberta. The process flow diagram of the test loop is
shown in Figure 9. The purpose of the test loop is to be able to produce various slurry
pipeflow conditions in a controlled environment. The construction material of the pipe
loop was chosen as SA-106 grade B seamless carbon steel pipe, which is a common
piping construction material. The loop also has a removable section to test the wear
resistance of other materials such as polymers and composites. The slurry loop consists of
a 7.5 kW pump equipped with a variable frequency drive, a Coriolis flow meter, a slurry
tank, and a straight run test section, with a total loop length of approximately 12 m. The
loop is also equipped with a clear sight-glass section for flow visualization and particle
tracking using particle image velocimetry to estimate the momentum loss during impacts
against the pipe wall.
Figure 9. Preliminary process diagram of the slurry loop at the University of Alberta
Since wall wear rate is a function of wall shear stress exerted by the slurry on the pipe
inner wall, a floating element sensor assembly was designed and constructed to measure
shear stress on a small area of pipe wall. The floating element sensor assembly is shown
in Figures 10 and 11.
Wear Sample
Gap
Strain Gage location on Cantilever
Supports
Figure 10. Cross section view of the floating element sensor assembly
In [11], Kiewicki, Saric, Marusic, and Eaton describe the floating element sensor as the
simplest sensor used in measuring wall stress due to its simple working principle. The
sensor is equipped with a floating wear sample mounted on elastic supports. The floating
wear sample can translate slightly due to friction force because of the small gap present
between the wear sample and the pipe wall; the displacement is proportional to the force
on the floating element.
The wear sample was machined out of the same pipe material using Electrical Discharge
Machining Technology in order to eliminate alignment concerns. The material of the
wear sample can be varied in order to measure the difference in overall force
measurement for different materials. The difference between readings may be used to
give an indication of the magnitude of force contribution due to random collisions of
particles and the sliding bed. Two flush water connections were added to the sides of the
floating element assembly to flush out trapped sand particles between the wear sample
and the pipe. This can be performed by maintaining a small positive differential pressure
between the flush water and the slurry flow at small flow rates. Alignment of the wear
sample was performed prior to commissioning of the floating element assembly.
Figure 11. Photo of the floating element assembly
The cantilever supports were equipped with two full Wheatstone bridges. The shear force
bridge provides strain measurements due to the slurry friction and eliminates the
contribution of the axial forces due to pipe internal pressure according to standard
formula relating simple beam loading to strain gauge output:
G.F . 4 Fy (d )( y )
(
)
.
(9)
Ei
4
IE
The axial force bridge measures the strain experienced by the cantilever support due to
axial force from high angle impacts and slurry turbulence while eliminating the strain
contribution of the friction and random impact force, as
EO
EO
Ei


G.F . 2.6 Fx
(
)
4
AE
.
(10)
A strain gauge conditioner fabricated at the University of Alberta was used to balance the
axial and shear strain gage bridges installed on the floating element cantilever supports.
The strain gauge conditioner contains two channels with separate circuits and an output
voltage variable amplifier with a gain range between 0 and 2100 for each channel. The
output voltage of each of the strain gauge bridges was acquired using a computer data
acquisition system.
In order to keep the gap from plugging with sand particles, flush water was used. By
maintaining a small positive differential pressure between the flush water and the slurry
flow, sand particles are prevented from entering the gap. The effect of flush water flow
across the gap can be minimized provided that the pressure difference is kept small. If
necessary, the force-strain relationship can be corrected empirically for circumstances
when the flush rate through the gap is high.
The floating element bridge outputs were calibrated using known forces applied in both
shear and axial loading as shown in Figure 12. After bridge balancing, the friction and
axial bridge outputs were recorded for each axial and shear force applied and the
corresponding [2x2] calibration matrix was calculated in Matlab. After calibration was
completed, the element was ready for laboratory testing in a slurry flow loop.
Figure 12. Calibration of Floating Wear Sample Using Known Weights
Preliminary testing of the floating element assembly was conducted on a test loop at
Syncrude, constructed using commercially available SA-106 Gr. B seamless carbon steel
pipe with nominal diameter of 76 mm and a test section of 50 mm diameter. The loop is
driven by a 3/2 AH Warman gland sealed belt driven pump with 40 HP Hyundai heavy
duty AC motor equipped with variable frequency drive. A Coriolis Krohne flow meter
and venturi flow meter measure flow rate. There is a hopper with a 50 mm outlet valve
for loading particulate solids into the loop.
Downstream Clear
Section
Floating Element
Upstream Clear
Assembly
Section
2” x 3” Transition Pipe Spool
Figure 13. Photo of the floating element assembly installed on Syncrude slurry loop
Test #1 was conducted with water flow without flush water supply. The water bulk flow
rate was increased from 0 USGPM to 250 USGPM and then decreased back down to 0
USGPM. A repeatability test (Test #2) to assess the variation in measured data by the
axial and friction bridges was also conducted. During test #2, the pump rpm was varied
from 0 to 706 rpm and then back to 0 rpm again. This process was repeated one more
time while continually recording the shear and axial full Wheatstone bridges, volume
flow rate, density, temperature, pump rpm, and pump discharge pressure. During the
second round of preliminary testing (test #3), heterogeneous sand slurry was pumped
through the test loop at average flow rates between 110 and 220 USGPM. Before the start
of test #3, the floating element assembly was removed from the piping system and
cleaned to ensure no sand was present in the chamber of the assembly.
Table 1. Test Summary
Test #
Description
Flow rate
Density
d50
Velocity
USGPM
Kg/m3
mm
m/s
1
Water Flow
0 to 250
1014
N/A
0 to 7.3
2
Repeatability
0 to 136
1014
N/A
0 to 4.0
3
Slurry Flow
110 to 220
1350
0.27
3.2 to 6.4
Results from test #1 and test #3 are represented in Figures 14, 15, 16, and 17.
Figure 14. Plot of measured water friction force versus bulk velocity
Figure 15. Plot of measured slurry friction force versus bulk flow rate
Similar to test #1, the friction force was measured for the duration of test #3, and
measured friction force increased with increasing flow rate as expected. The difference
between the starting zero point and the final zero point is due to a small drift in both of
the axial and friction bridge sensor output.
Figure 16. Plot of measured water axial force versus bulk flow rate
Figure 17. Plot of measured slurry axial force versus bulk flow rate
The increase in the axial force measured was more substantial compared to the water
flow. The axial final zero point was also plotted on the same plot.
The theoretical kinematic stress due to water flow in the pipeline is given by
 kw  0.5w f fwVw2
.
(11)
Knowing the water density, velocity and the corresponding fanning friction factor, the
theoretical kinematic stress at the pipe wall was calculated (using water viscosity value
 f = 10-3 Pa.sec). The corresponding theoretical friction force was computed by
multiplying the theoretical stress calculated by the area of the floating element wear
sample in contact with the flow. The comparison graph shown in Figure 18 highlights the
difference between the theoretical friction force and the measured friction force.
Figure 18. Theoretical and measured water friction force versus flow bulk velocity
The small difference between the theoretical and measured water friction force can be
attributed to several factors. First of all, the viscosity of the water in the theoretical
calculations was approximated to be 0.001 Pa.sec for water at 20C. The actual viscosity
of the tested water must be measured in the lab in order to verify this assumption. There
may be slight error in the volume flow rate and density measurements made by the
Coriolis flow meter.
Similarly, the theoretical kinematic stress due to slurry flow in the pipeline was
calculated according to the two-layer model assuming fully stratified flow conditions:
 kw  0.5w f fwV 2  0.5s f fsV 2
(12)
According to the improved two-layer model [9], the kinematic stress experienced by the
flow is increased due to the presence of solid particles in the mixture, which can be
computed by calculating a modified friction factor f fs . The mean diameter of sand
particle size distribution used was 0.27 mm. A plot showing the slurry flow theoretical
and measured friction force is shown in Figure 19.
Figure19. Theoretical and measured slurry friction force versus flow bulk velocity
The measured slurry friction force is high for the measured flow conditions compared to
the friction force calculated from the kinematic stress based on the two-layer model. The
presence of the gap between the floating element wear sample and the pipe has
contributed to unaccounted discrepancy in the measurement. Fouling of the assembly and
the flow of flush water through the gap contributed to error in the measurements, but this
error was not quantified. These predicted sources of error are considered to be the main
sources of error in the measurements made.
Due to Coriolis flow meter malfunctioning at higher slurry densities, flow rate was
measured using a backup venturi flow meter. The flow rate measurement of the Venturi
meter was thought to be slightly out of calibration for correcting the venturi flow rate
estimate for slurry, because the slurry flow rate was measured using a venturi flow meter
which is calibrated to measure water flows, and the meter had previously given slightly
erroneous measurements compared to the Coriolis output for the same water flow rate.
The measured slurry friction force was 10 times higher than the theoretical friction force.
Since the area of the element is constant, the measured kinematic stress acting on the
floating element is 10 times higher than the theoretical kinematic stress. The required
flow rate would have to be 2.2 times the measured flow rate in the pipe loop to produce
the same stress level experienced by the floating element sensor. The error in the flow
measurement could not be that high, as the pump motor could not produce enough power
to drive such a high flow rate in the loop. The decrease in water viscosity due to test
temperature increase during testing is small and as a result its effect on the kinematic
stress value is also considered negligible. It was assumed that the errors in the f fs and
f fw are negligible relative
to the error in the flow rate measurement; and it was assumed
that the error in the solids density and carrier fluid density is negligible. (Additional
velocity and concentration profile measurements must be made during future testing in
order to verify the floating element friction measurements.)
There may be errors in relating strain measurements to forces on the element surface. The
floating element sensor measures both axial and shear forces at the same time. As a
result, cross channel interference occurs between the axial and shear bridges. Errors in
the axial bridge measurement get transferred due to cross-channel interference to the
shear bridge output. Consequently, it would improve the design of the floating element to
conduct the axial and shear force measurements separately.
Cross-channel interference error can be attributed to the pressure drop across the floating
element sensor, which applies an axial load. The axial load applied causes errors in the
axial and friction force measurement because the element was calibrated using a
combination of axial and shear loads. The pressure drop across the floating element wear
sample during slurry flow is estimated to be 600 Pa using the two layer model equation at
220 USGPM flow rate. This pressure drop causes an additional axial force equal to 1.8 N
knowing that the surface area of the floating wear sample is 0.003 m2. By inspection of
calibration Table 1, the voltage difference due to the pressure drop is 0.019 V in the axial
bridge and 0.027 V in the friction bridge which causes a maximum relative error in the
bridge output equal to 3% in the axial bridge output and 15% in friction bridge output.
Post test calibration of the floating element sensor was conducted to make certain that
the element was not damaged during testing. Initial visual inspection of the assembly
indicated there was no visible damage done on the element since all the components were
in good condition. The post calibration shear force results agreed with the pretest shear
calibration results. No axial force post test calibration was performed since the sensor
was in good condition.
During test #2, the friction force measured was repeatable with a standard deviation in
the measured data equal to 4%, as shown in Figure 20. The axial force output was less
repeatable compared to the friction force with a calculated standard deviation equal to
5%, as shown in Figure 21.
Figure 20. Friction force versus time during repeatability test #2
Figure 21. Axial force versus time during repeatability test #2
The dynamic response of the sensor was also investigated. The first natural frequency of
the moving components of the floating element assembly was found by mounting the
wear sample assembly on a shaker table, shown in Figure 22, consisting of an articulating
drum that vibrates a shaker table placed on a lubricated surface. The frequency of
vibrations was varied until the first natural frequency of the element was found to be 38
Hz.
Wear Sample Holder
Assembly
Vibrating Drum
Shaker Table
Figure 22. A photo of the frequency test setup
Both axial and shear bridge output signals were acquired at 500 samples per second data
acquisition rate. A power spectral density (PSD) analysis of the friction bridge signal was
developed in Matlab using Welch averaged spectral estimation method with several
windowing functions such as Hanning, Blackman and Chebychev. The input signal was
divided into segment sizes of 64 samples for all three window functions with default
value of 50% overlap between segments. The PSD plot using Chebychev window
showed the power spectrum without exhibiting apparent spectral leakage, and as a result
it was chosen as the window of choice. The peak of the power spectral density plot
occurs at 5 Hz for the friction signal as shown in Figure 23. The power spectral density
plot drops sharply after 10 Hz and reaches a minimum at 30 Hz as the plot flattens out.
Another small peak occurs at around 39 Hz due to the natural frequency of the floating
element sensor. As a result, the amplitude of these fluctuations is amplified due to
resonance resulting in another peak in the power spectral density. Fortunately, the
magnitude of the latter peak due to resonance is very small compared to the dominant
peak of fluctuation in the signal occurring at 5 Hz.
Figure 23. Sample friction bridge power spectral density plot for both slurry and water flows
As shown in Figure 24, the peak of the power spectral density plot occurs at 5 Hz for the
axial signal during water and slurry testing at 207 and 190 USGPM respectively. The
power spectral density plot drops sharply after 10 Hz and reaches a minimum at 30 Hz as
the plot flattens out. As a result, the signal dominant measurement fluctuations occur at
around 5 Hz, which is well below the first natural frequency of the floating element in the
axial directions as expected. Small amplitude fluctuations in the signal occurring above
30 Hz are likely due to mechanical force fluctuations from turbulence or low angle
impact of particles, or may be an artifact of the low-pass filtering.
Figure 24. Sample axial bridge power spectral density plot for both slurry and water flow
From these tests, it was concluded that the signal measurements were not distorted by
excitation of a natural frequency of the element. Small amplitude fluctuations in the
signal occurring above 30 Hz were likely due to mechanical force fluctuations from
turbulence or low angle impact of particles.
Conclusions and Future Work
A model has been formulated for estimating erosion wear in a horizontal section of pipe
under steady state conditions for sand slurry flow. The model is based on the physics of
sand slurry flow represented by the two-layer model. The presented erosion wear model
can be improved to embody erosion wear in oilsands straight horizontal section of pipe
under steady state conditions by including the contribution of bitumen and particle
digestion to the total erosion wear.
A floating element assembly has been designed and tested to measure axial and shear
forces inside a straight pipe section. Both axial and shear forces acting on the floating
element were measured during water and sand slurry experiments in addition to other
flow variables such as flow rate, density, temperature, pump rpm, and pump discharge
pressure. The results of the measured forces did not fully agree with theoretical results
due to possible discrepancies in the flow rate measurement and a possible concentration
and possibility of a concentration profile across the pipe cross section during slurry
testing.
Future work will improve the design of the floating element sensor assembly so that the
floating element sensor can be used in future slurry wear tests. The steady-state, lumped
parameter model will be validated. The wear model will be enhanced to add additional
elements to capture erosion wear in the transition zone between top layer and bottom
layer, and to include the contribution of corrosion and combined erosion-corrosion to
wear rate. Conditions under which particles contact the surface will be investigated, and a
practical approach will be developed and field-tested to measure stresses experienced by
industrial slurry transport pipelines.
Acknowledgments
Funding support is gratefully acknowledged from Syncrude Canada Ltd. and The Natural
Sciences and Engineering Research Council of Canada (NSERC). Victor Jaimes
contributed to the slurry loop design. Derek Loewen contributed to the spectral analysis
of the sensor signals. Dan Wolfe coordinated the slurry loop experiments at Syncrude.
Nomenclature
PA is the averaged pressure of the top layer at the start of pipe
PD is the averaged pressure of the top layer at end of pipe
q0 is the average flow rate of the slurry in pipeline
ql is the average flow rate of slurry in the lower layer of pipe
qu is the average flow rate of slurry in the upper layer of pipe
R1 is the resistance coefficient due to bottom layer friction (Columbic + kinematic
friction)
R3 is the resistance coefficient due to bottom layer mixing (Energy contributing to
particle suspension and also to particle impact with the pipe inner wall in bottom layer)
R4 is the resistance coefficient due to top layer friction (Kinematic friction)
R5 is the resistance coefficient due to top layer mixing (Energy contributing to particle
suspension and also to particle impact with the pipe inner wall in upper layer)
s ' is the wear rate by the corresponding wear mechanism in units of (thickness/time)
E d is the rate of energy transfer to the pipe wall contributing to damage in units of
(energy per unit area/time)
Et is the total rate of energy transfer to the pipe wall in units of (energy per unit
area/time)
E o is the threshold energy rate for incipient wear above which damage starts to
accumulate (energy per unit area/time)
 is the stress caused by the wear component on the pipe wall in units (force/area)
 o is the threshold stress for incipient wear caused by the wear component on the pipe
wall in units of (force/area)
V is the local tangential velocity component for each wear mechanism in units of
(distance/time)
α is the wear coefficient of each wear component in units of ((thickness/(energy per unit
area))
α11 is the wear coefficient due to bottom layer kinematic friction in the lower layer
α12 is the wear coefficient due to bottom layer Columbic friction in the lower layer
α3 is the wear coefficient due to particle impact with pipe inner wall in the lower layer
α4 is the wear coefficient due to top layer kinematic friction in the top layer
α5 is the wear coefficient due particle impact with pipe inner wall in the upper layer
β3 is the energy fraction causing particles to exit flow streamlines and impact pipe inner
wall in the lower layer
β5 is the energy fraction causing particles to exit flow streamlines and impact pipe inner
wall in the upper layer
Fy is the shear force in the y direction
Fx is the axial force in the x direction
I is the moment of inertia
A is the cross sectional area of the cantilever element
d1 is the distance between the shear force and strain gage 1
y is equal to half of the cantilever element thickness
Eo is the bridge output voltage
Ei is the bridge excitation voltage
G.F . stands for the gage factor
 w is the density of water = 1014 kg/m3
f fw is the water fanning friction factor
 s is the density of sand particles = 2650 kg/m3
f fs is the friction factor computed using the two layer model
V is the slurry flow velocity in the bottom layer of the flow
U total is the total energy dissipated in the system,
U damage is the energy fraction contributing to damage in the system,
Pt1345
is the total pressure drop in the system in Pa,
P1345 is the pressure drop in the system due to slurry flow in Pa,
Pd1345 is the additional pressure drop in the system due to erosion damage in Pa,
Rt1345 is the resistance coefficient representing total pressure drop in the system,
R1345 is the resistance coefficient representing pressure drop in the system due to slurry
flow, and
Rd1345 is the resistance coefficient representing erosion damage
References
1. Schaan, J., Cook, N., & Sanders, R. S. (2007). On-line wear measurements for
commercial-scale, coarse-particle slurry pipelines. The 17th International
Conference on the Hydraulic Transport of Solids, The South African Institute of
Mining and Metalurgy and the BHR Group.
2. Wilson, K. C., Addie, G. R., Sellgren, A., & Clift, R. (2006). Slurry transport
using centrifugal pumps (3rd ed.). New York: Springer Science.
3. Roco, M. C., & Addie, G. R. (1987). Erosion wear in slurry pumps and pipes.
Powder Technology, 50(1), 35-46.
4. Gupta, R., Singh, S. N., & Seshadri, V. (1995). Study on the uneven wear rate in a
slurry pipeline. Bulk Solids Handling, 15(4), 603-607.
5. Salama, M. M. (2000). An alternative to API 14E erosional velocity limits for
sand-laden fluids. Journal of Energy Resources Technology, Transactions of the
ASME, 122(2), 71-77.
6. Wood, R. J. K., Jones, T. F., Ganeshalingam, J., & Miles, N. J. (2004).
Comparison of predicted and experimental erosion estimates in slurry ducts.
Wear, 256(9-10), 937-947
7. Gillies, R. G., Shook, C. A., & Wilson, K. C. (1991). Improved two layer model
for horizontal slurry pipeline flow. Canadian Journal of Chemical Engineering,
69(1), 173-178.
8. M. Lipsett, “Modeling wear damage accumulation in slurry pipeline systems,”
Fifth International Conference on Condition Monitoring 2008, Edinburgh, July 17
2008
9. Gillies, R. G., & Shook, C. A. (2000). Modeling high concentration settling slurry
flows. The Canadian Journal of Chemical Engineering, Volume 78, 709-716.
10. Lipsett, M., & El-Sayed, S. (2009). A simple predictive model for slurry wear in a
pipeline with bed flow. Global Petroleum Conference, Calgary.
11. Kiewicki, J. C., Saric, W. S., Marusic, I., & Eaton, J. K. (2007). Wall bounded
flows. In C. Tropea, A. Yarin & J. F. Foss (Eds.), Springer handbook of
experimental fluid mechanics (pp. 871-902)
12. M. Lipsett, “Incorporating Damage Models in Deterministic System Models to
Model Reliability in Time-Varying Systems,” Sixth International Conference on
Condition Monitoring 2009, Dublin, June 23 2009.
13. Jamaludin, N., & Mba, D. (2001). Monitoring low-speed rolling element bearings
using acoustic emissions. Proceeding of the 14th International Conference on
Condition Monitoring and Diagnostic Engineering Management, 37-48.
14. Budinski, K. G. (2007). Guide to friction, wear and erosion testing. West
Conshohocken, PA: ASTM International.