energies

Review
Effects of Fluid Viscosity and Two-Phase Flow on
Performance of ESP
Peng Liu 1 , Yumo Wang 1, *, Feng Yan 1 , Chaofei Nie 1 , Xin Ouyang 1 , Jiashuang Xu 2
and Jing Gong 1
1 CNPC Key Laboratory of Oil & Gas Storage and Transportation, China University of Petroleum,
Beijing 102249, China; [email protected] (P.L.); [email protected] (F.Y.);
[email protected] (C.N.); [email protected] (X.O.); [email protected] (J.G.)
2 China Petroleum Engineering & Construction Corporation Southwest Company, Chengdu 610094, China;
[email protected]
* Correspondence: [email protected]




Received: 22 September 2020; Accepted: 12 October 2020; Published: 20 October 2020


Abstract: Electric submersible pumps (ESPs) are widely used in the oil and gas industry for crude-oil
lifting, especially in subsea oil fields or underground storage caverns. The failure of ESPs causes
a large economic cost mainly attributed to a break in production continuity, as the ESP cannot be
easily replaced. Therefore, the assurance of safe and efficient operation of ESPs has attracted high
attention in recent years, although the problem still remains challenging given the complexity of
carrying fluid and the mechanical structure of the ESP. In this article, we systematically review both
the high-impact, classic contributions and the most up-to-date, current opinions in experimental
and numerical advances of viscous effects and two-phase flow in ESPs. We specifically focus on the
applications in the oil and gas industry and point out a few current challenges in the operation of
ESPs. We aim to guide the audience which is new to the area of ESPs to the correct articles related to
their interests, including classic work and recent advances.

Keywords: electric submersible pumps (ESPs); two-phase flow; computational fluid dynamics (CFD);
flow pattern; viscosity effect

1. Introduction
With rapid development of the world economy and continuous exploitation of land resources,
much attention has been paid to offshore oil fields. Oil in gusher wells can be extracted with natural
power. For commonly existing low-pressure wells, however, artificial lift methods are usually needed,
such as gas lifts [1], progressive cavity pumps [2], and electric submersible pumps (ESPs) [3]. Unlike rod
pumping, ESPs can be employed without gas wells nearby and conveniently coupled with electrical
motors, which, together with their small footprint and high energy efficiency [4], make them widely
used in subsea oil field development.
The electric submersible pump oil production system is composed of aboveground and
underground parts, as shown in Figure 1a. Usually, aboveground parts of ESP consist of a switchboard,
junction box, wellhead, etc., while underground parts contain a motor, protector, separator, and pump.
The motor is the prime motion-generator of the system, which is generally located at the bottom.
The protector lays in the middle to isolate well fluid from motor oil. The main role of the separator is to
reduce the gas volume fraction inside the pump. There are tens or even hundreds of stages in a pump
and each stage consists of an impeller (Figure 1b) stacked under a diffuser (Figure 1c). When working,
blades in impellers rotate at high speed, contributing to an increase in fluid pressure and velocity.
The kinetic part of fluid energy is further converted to fluid pressure through the diffuser before the

Energies 2020, 13, 5486; doi:10.3390/en13205486 www.mdpi.com/journal/energies

Energies 2020, 13, 5486 2 of 20
Energies 2020, 13, x FOR PEER REVIEW 2 of 21

pressure through
fluid is directed to the diffuser
the next before
stage. the fluid
Eventually, theisfluid
directed to the
pressure can next stage.elevated
be greatly Eventually, the fluid
through fluid
pressure can be greatly elevated
flow past multiple stages in series.through fluid flow past multiple stages in series.

Figure 1. (a) Conventional electric submersible pump (ESP) installation (Carvajaldiaz 2012 [5]); (b) an
Figure
example1.of
(a)impeller
Conventional electric
front view submersible
(Takacs 2009 [4]);pump
(c) an (ESP) installation
example (Carvajaldiaz
of diffuser 2012 [5]);
front view (Takacs 2009(b)[4]).
an
example of impeller front view (Takacs 2009 [4]); (c) an example of diffuser front view (Takacs 2009 [4]).
For different flow directions, centrifugal pumps are classified as axial, radial, and mixed flow
For different radial
pumps. Typically, flow directions,
and mixedcentrifugal
flow pumpspumps are classified
are selected for ESPsas foraxial,
their radial, and mixed
larger head. flow
For lighter
pumps. Typically,
application when the radial
flow and
ratemixed
is lessflow
thanpumps are selected
3000 barrels per dayfor(bpd),
ESPs for theirflow
radial larger
pumpshead.areForsuitable.
lighter
application
On the otherwhenhand,the flowflow
mixed ratepumps
is less than 3000 barrels
are typically usedper
for day (bpd),
a large radialofflow
capacity up topumps
40,000are suitable.
bpd. For all
On
ESPs,thestable
other operation
hand, mixed andflow pumps
lifespan are typically
assurance used for
are vitally a large capacity
important of upthe
in reducing to 40,000 bpd.cost
economic For
all ESPs,production,
during stable operation
as theand lifespan
operating assurance are
environment vitally
of ESP canimportant in reducing
be inaccessible the economic
and harsh. cost
For instance,
during production,
the maintenance costasofthe operating
a pump under environment
deep sea is ofupESP can
to 40 be inaccessible
times that of setting andupharsh.
a new For instance,
pump [5].
the maintenance
Factors thatcost of a pump
influence pump under deep sea
operation and is performance
up to 40 timesinclude
that of setting
flow rate,up arotational
new pump [5].
speed,
pumpFactors
geometrythat[6],
influence pump operation
and oil viscosity [7–11]. Theandflowperformance includespeed
rate and rotational flow arerate,subject
rotational speed,
to operation
pump geometry
conditions [6], and easy
and relatively oil viscosity
to control. [7–11]. The flow
The pump rate and
geometry, whichrotational
also hasspeed
a largeare subject on
influence to
operation
pump conditionshowever,
performance, and relatively
needs toeasy to control.
be taken The pump
into account geometry,
in advance. which also
The structure has acan
of ESPs large
be
influence
increasinglyon pump performance,
sophisticated due to however, needs to
the interactions be taken
between into account
stages, especially in advance.
when a larger The structure
number
of ESPs
stagescan
arebe increasingly
introduced sophisticated
to gain due to the
a larger pressure interactions
head. For flowbetween stages, especially
rate, rotational speed, andwhen pump a
larger number
geometry, of stages
the affinity lawsarecan
introduced
be applied to gain
undera larger
various pressure head.
conditions inFor flow rate,
general. However,rotational speed,
the viscous
and pump
effect, whichgeometry, the affinity
alters with the change laws in canspecific
be applied
speedunder
and various conditions
rotational speed, is in difficult
general. toHowever,
predict
the viscous effect,
analytically with awhich
generalalters
form.with the change in specific speed and rotational speed, is difficult to
predict
Theanalytically with a general
possible generation form. flow in ESPs from the appearance of bubbles and particles
of two-phase
The possible generation of two-phase
can cause damage to critical parts of the pump, flow especially
in ESPs from underthe certain
appearanceworkingof bubbles
conditionsand like
particles
high
can causedrop
pressure damage andto critical
heavy parts
fluid of the pump,
medium. Whenespecially
the pressure under certain
drops below working conditions
the bubble like high
point pressure,
pressure
cavitationdrop and heavy
appears fluid medium.
and damages When critical
the pump’s the pressure
parts.drops below thesome
Furthermore, bubble point
fine pressure,
sands in the
cavitation appears and damages the pump’s critical parts. Furthermore,
mining process can flow through the pump channels with the fluid, despite the presence of filtering some fine sands in the
mining
equipment.process
Thecan flow through
problems the pump
that solid channels
particles give risewith
to the
are fluid, despite
not only the presence
pressure loss, butofalso filtering
wear
equipment. The problems that solid particles give rise to are not only pressure loss, but also wear
Energies 2020, 13, 5486 3 of 20

damage, which could lead to vibration and even the failure of facilities. When bubbles and particles
both appear in ESPs, the flow pattern is difficult to predict accurately [12,13].
Therefore, the effects of fluid viscosity and two-phase patterns have attracted much attention
in recent years, and there were numerous studies on these problems, using three approaches:
(1) experimental, phenomenological, or empirical curves/coefficients that are more practical in
reality and engineering; (2) analytical or semi-empirical approaches; (3) computational fluid
dynamic simulations.
To understand the effects of fluid viscosity and two-phase flow on ESP performance, various
experimental tests on different operation conditions were conducted to achieve empirical coefficients,
which were combined with analytical equations and semi-empirical formulas to explain pump
performance. Meanwhile, with the development of computational fluid dynamics (CFD) and relevant
commercial software, the method of numerical simulation, which has advantages of convenience
and accuracy, is widely becoming a routine technique after verification from experimental results,
although the high cost of calculation resources and time can sometimes become a problem depending on
the complexity of geometry [6]. Moreover, the simulation results of a pumping head are usually higher
than that from experimental data, owing to neglecting the leakage of clearance between impellers
and diffusers.
Given the complexity of these approaches, we present this review to guide our audience to
the proper references according to their specific needs without being overwhelmed by the pool of
literature. This review covers two topics: viscous effects and two-phase flow in ESPs; under each
category, the references are categorized as experimental or numerical analysis. The criteria used for
selecting references was as follows: we first included the key classic references on the basis of our
knowledge and experience, and we searched for articles most closely related using the keywords “ESP”
and “performance curve” before refining our search results using the keywords “viscosity effects”,
“particles”, “bubbles”, and “flow mechanism”. We then selected the higher-impact, classic contributions
and the most up-to-date, novel results for the current framework of the review article to systematically
evaluate the experimental and numerical advances in the study of viscous effects and two-phase flow
in ESPs. Note that the effects of corrosive wear in the ESP are not covered in this review, as we focused
on the applications in the oil and gas industry, specifically, when crude oil is the carrying fluid. In this
case, corrosive damage is usually not as important as erosion/abrasion.

2. Viscous Effects
A pump’s performance is evaluated by its head, power, and efficiency at a certain flow rate. Usually,
suppliers provide the performance curves of pumps with water handled. In working conditions,
however, the pump is submerged under crude oil or another heavy fluid, which have much larger
viscosity. It was shown that high fluid viscosity clearly results in a degradation in pump performance
(Figure 2) by reducing head and efficiency and by increasing power consumption. In particular,
Stepanoff [14] found that the centrifugal pump performance at the best efficiency point (BEP) is
degraded at constant specific speed. Nevertheless, the flow rate and viscosity of fluid in ESPs can
change substantially in a short time under operational conditions. Solano [15] then confirmed that
this approach is also applicable under off-design conditions when proper dimensionless numbers
are chosen.
Energies 2020,13,
Energies2020, 13,5486
x FOR PEER REVIEW 44 of
of 20
21

Figure 2.2.Curves
Figure Curvesofof
normalized
normalizedpressure headhead
pressure versus flow rate
versus flowatrate
a certain rotational
at a certain speed forspeed
rotational different
for
fluid viscosity (Ofuchi 2017 [16]).
different fluid viscosity (Ofuchi 2017 [16]).

To
To understand
understand the the effects
effectsofofviscosity
viscosityonon pump
pump performance,
performance, several
several ESP ESP
tests tests
[7,8,11][7,8,11]
were
were conducted. According to the experimental data of fluid with different viscosity,
conducted. According to the experimental data of fluid with different viscosity, empirical coefficients empirical
coefficients [14,17]toare
[14,17] are given given
guide pumpto guide pump For
operation. operation.
comparingFor comparing
various pumps,various pumps, dimensionless
dimensionless analysis is
analysis is also an effective method. Fitting and correcting cannot
also an effective method. Fitting and correcting cannot explain the flow phenomenon explain the flow phenomenon
in ESPs from in
ESPs from the mechanism.
the mechanism. Therefore,Therefore,
conservation conservation
equations equations of mass, momentum,
of mass, momentum, and energy and energy
need to beneed to
solved
be solved to understand the flow pattern. Considering the flow rate and geometry
to understand the flow pattern. Considering the flow rate and geometry parameters in ESPs, various parameters in ESPs,
various
turbulent turbulent flow equations
flow equations are employed.
are employed. After choosing
After choosing a propera proper
modelmodel in Computational
in Computational Fluid
Fluid
Dynamics (CFD) commercial software, the distribution of related physical parameters such as
Dynamics (CFD) commercial software, the distribution of related physical parameters such as
velocity
velocity andand pressure
pressure can
can be
be obtained
obtained through
through iterative
iterative computations,
computations, whichwhich cancan then
then bebe utilized
utilized toto
illustrate
illustrate the
the flow
flow feature.
feature.
In
In this section, researches
this section, researcheson onpump
pumpperformance
performancecurvescurves when
when pumping
pumping high-viscosity
high-viscosity fluidfluid
are
are introduced. In order to understand the viscous effect, many experiments
introduced. In order to understand the viscous effect, many experiments and numeric computations and numeric
computations
[7–11] were carried[7–11]out.
were carried
With out.toWith
respect flowrespect
patternto flow
and patternloss,
pressure andthe
pressure
mechanismloss, the mechanism
of the viscosity
of
effect behind performance is explained to some extent. In addition, the effects of fluid viscosity of
the viscosity effect behind performance is explained to some extent. In addition, the effects in
fluid viscosity
multistage ESPsinare
multistage ESPs are briefly
briefly mentioned, althoughmentioned, although
their thorough their thorough
understanding remainsunderstanding
challenging.
remains challenging.
2.1. Performance Curves
2.1. Performance Curves
The best indicator of pump behavior is achieved by evaluating the shift in the performance
The best indicator of pump behavior is achieved by evaluating the shift in the performance curve.
curve. To acquire a typical performance curve, four important factors (flow rate, head, power, and
To acquire a typical performance curve, four important factors (flow rate, head, power, and efficiency)
efficiency) are usually rendered dimensionless as a flow coefficient, head coefficient, power
are usually rendered dimensionless as a flow coefficient, head coefficient, power coefficient,
coefficient, and efficiency coefficient.
and efficiency coefficient.
Q gH ωΓ φψ
φ= Q 3 ,ψ = 2gH 2 , Π = ωΓ ,η =φψ (1)
φ =ω D2 3 , ψ = ω 2 D22 , Π = ρω3 D52 , η = Π
3 5
(1)
ωD 2 ω D 2 ρω D Π2

φ is the
where φ is the
where flow coefficient,
flow ψ isψtheishead
coefficient, the coefficient,
head coefficient, Πshaft
Π is the power
is the shaftcoefficient,
power coefficient, η is
η is the pump
Q flow rate, ω is the angular speed, D2 is the impeller outlet
efficiency, Q is the volumetric
the pump efficiency, ω is the angular speed, D 2 isdiameter, g is
gravitational acceleration,isHthe volumetric
is the flow
head, Γ is the rate,
torque, and ρ is the fluid density. Three thecurves
impeller
(ψ
outlet diameter, g is gravitational acceleration, H is the head, Γ is the torque, and ρ is the fluid
vs. φ, Π vs. φ, and η vs. φ) were drawn through several experiments to evaluate performance under
different
density. operation
Three curves ( ψ vs.
conditions for φ
each Π vs.
, type
φ , and
of pump [18].η vs. φ ) were drawn through several
experiments to evaluate performance under different operation conditions for each type of pump [18].
 Energies 2020, 13, x FOR PEER REVIEW 5 of 21
Energies 2020, 13, 5486 5 of 20

When the viscosity of pumping fluid is much higher than that of water, performance curves
deviate
Whenfrom the reality.
the viscosity The Hydraulic
of pumping fluid isInstitute Standard
much higher than[17]
that provides
of water, correction
performance factors for
curves
different
deviate fluid
from the viscosities
reality. Theon the basisInstitute
Hydraulic of empirical parameters
Standard from experiments.
[17] provides These
correction factors forresults are
different
sometimes
fluid questioned
viscosities because
on the basis the pumps’
of empirical specified
parameters speed
from in their experiments
experiments. is limited
These results to only a
are sometimes
small range
questioned [19]. The
because influence
the pumps’ of specific
specified speed
speed on head,
in their power,isand
experiments efficiency
limited to onlyisa plotted in Figure
small range [19].
3, asinfluence
The the trendsofofspecific
these parameters with respect
speed on head, power,to capacity
and (flowisrate)
efficiency can change
plotted in Figurewith specific
3, as speed.
the trends
ofFor
thesea larger specificwith
parameters speed, the power
respect no longer
to capacity (flowincreases
rate) canwith capacity.
change with specific speed. For a larger
specific speed, the power no longer increases with capacity.

Figure 3. Classification of centrifugal pumps (Zhu 2019 [20]).

Figure 3. Classification of centrifugal pumps (Zhu 2019 [20]).
When pumping fluid with higher viscosity, Reynolds numbers for ESPs decrease to the hydraulic
transition zone. Thus, dimensionless numbers according to both specific speed and Reynolds number
When pumping fluid with higher viscosity, Reynolds numbers for ESPs decrease to the
can be used. With constant Reynolds numbers, dimensionless head or efficiency curves overlap,
hydraulic transition zone. Thus, dimensionless numbers according to both specific speed and
in spite of different viscosities and rotational speeds. Morrison et al. [21] added a rotational Reynolds
Reynolds number can be used. With constant Reynolds numbers, dimensionless head or efficiency
ρωD2 2
number (Re =
curves overlap, inµ spite
w ) to of
understand pump performance
different viscosities and rotational(µ is dynamic viscosity),
speeds. Morrison et They proposed
al. [21] added a
a new variable φ · Rew −Mo to establishρωD2 the relationship with ψ. Here, exponent Mo is known as the
2
Rew =
Morrison
rotationalnumber
Reynolds which
numberis related
( to μpump design.
) to understand pump performance ( μ is dynamic viscosity),
The performance curve for aφspecific ⋅ Rew−Mopump can be drawn using CFD or experimental techniques.
ψ . Here, exponent
They proposed a new variable to establish the relationship with Mo
According to simulation results, the best efficiency point shifted to lower flow rates with increased
is knownInasother
viscosity. the Morrison
words, a pumpnumber withwhich
higher is related
flow rateto was
pump design.
needed. The study combined the viscosity
effects at different specific speeds and provided a simple tool to predictCFD
The performance curve for a specific pump can be drawn using pumpor experimental
performance.techniques.
However,
According
the research to
was simulation results, and
based on results, the best
thereefficiency
was littlepoint shiftedof
explanation toalower flow rates with increased
mechanism.
viscosity. In refined
Patil [22] other words,
Morrison’sa pumpstudywith and higher
gave anflow rate was
analytical formneeded. The study
of the Morrison combined the
number.
viscosity effects at different specific speeds and provided a simple tool to predict pump performance.
However, the research  was based + on results,
) ∗ Reand( Nthere was
s −0.11 little explanation of a mechanism.

Ns )
(− 20000 0.2698 w 50000 Mo ≤ Momax
Mo =Morrison’s , number. (2)

Patil [22] refined 
 study and gave an analytical
−5 form
(Ns ∗ 10−6 − 0.0008) ∗ Rew (−8∗10 ∗Ns +0.5474) Mo > Momax of the Morrison

 N (
Ns
−0.11)
 (−As shown + 0.2698)*Re Mo ≤ Mo
where Ns is the specific speed. s here, Morrison’s 50000
number is function of both specific speed
w max
Mo =number.
and rotational Reynolds  20000
When the specific speed is a constant, Morrison’s number increases (2)
 −6 ( −8*10 −5
* N + 0.5474)
( N s *10 − 0.0008)*Rew Mo > Momax
with increasing rotational Reynolds numbers until the peak Mo s max , which depends upon the specific

speed. For larger Rew , the Morrison number attains the peak and then gradually reduces , to a value of
around 0.052. The sharp change in Morrison number reveals a transition of flow regime, which can be
Energies 2020, 13, 5486 6 of 20

used to explain the decrease in pump head for larger-viscosity fluid to some extent. However, it is not
adequate to explain the effects of viscosity on the power and efficiency, as the relationships among Π,
η, and φ · Rew −Mo are not yet clear.
Another method for predicting the viscous effect for different pump geometries is normalization.
With dimensionless numbers, normalized groups are utilized according to design operation parameters.
Ofuchi et al. [16] compared curves between constant normalized Reynolds numbers and constant
normalized specific speeds. The effects of viscosity on the head curve degradation can be seen for
both comparisons, but the degradation trends for different pumps match only for the latter. A general
nφ nψ 1.5
correlation ( φn,ω = [ ψn,ω ] ) between head and flow coefficient was proposed, which can be directly
used to predict performance degradation without major geometry parameters. Here, φn and ψn are
the normalized specific flow rate and head at a given rotational speed and fluid viscosity, whereas φn,ω
and ψn,ω are the normalized specific flow rate and head at the design rotational speed with water.
Dimensionless analysis is a simple and effective method to consider the viscous effect and predict
pump performance. However, it cannot explain why degradation happens. Thus, research on the flow
pattern and pressure loss was conducted to understand the mechanism.

2.2. Flow Mechanism

Under operating conditions, the flow rate and viscosity of fluid may vary greatly in a short time.
Pump performance in reality is likely to be degraded compared to that at BEP. In order to investigate
the viscous effect in practical application, Amaral et al. [8] carried out experiments with glycerin,
the viscosity of which varies within the viscosity range of crude oil. Simultaneous changes of fluid
viscosity and pump speed were revealed, and related experimental results showed large deviations
from data at BEP that HI-USA-charts provided. By assessing the head gain and comparing the pressure
difference, the interaction between impellers and diffusers was proposed as the reason for the viscous
effect and was not negligible in the theoretical calculation. The experimental data for various operation
conditions were precious for following research, and the conclusion provided a direction for specific
CFD simulations.
The flow field pattern in mixed flow pumps (Figure 3) is complex to understand even for
single-phase operation. Thus, CFD simulations are commonly carried out. Specific steps include
geometry creation, mesh generation, model selection, condition setting, and iterative computation.
For the flow pattern in ESPs, the selection of turbulence model plays a key role. To acquire solutions
with adequate accuracy within an acceptable period of time, the Reynolds average Navier-Stokes
(RANS) equations were widely employed in CFD solvers in recent studies [9,10,23]. Due to the inclusion
of the Reynolds stress term, it is necessary to add complementary equations. Several turbulence
models, including standard k-ε, RNG (renormalization group) k-ε, standard k-ω, BSL (baseline) k-ω,
and SST (shear stress transport) k-ω, are usually used. Most times, there is no large difference between
numerical results using any of the above models for pump head. However, for part-load operation,
the SST k-ω model is usually preferable as it can recover the appearance of separation zones while
other models cannot. On the basis of this model, Stel [9] and Ofuchi [10] conducted much research on
the viscous effect.
However, the sudden rising head effect was observed [24], which means that the pump head
increases a little with higher fluid viscosity. Li [23] found this phenomenon to disappear with the
SST k-ω model in CFD simulations. When the standard k-ε is employed instead, the effect reappears.
The wet surface roughness was found to be related, and skin friction factors were defined as key
parameters. In the hydraulic transition zone, the skin friction factors decrease with lower Reynolds
numbers, resulting the sudden rising head. Providing the occurring condition of rough wet walls,
the SST k-ω model may be improper. Thus, turbulence models need to be compared with experimental
results under special circumstances.
Further researches were conducted for pressure loss due to viscosity. Various losses (Figure 4) in
ESP stages were defined [25]. Pump head is influenced by hydraulic loss, disc friction, leakage, and so
 numbers, resulting the sudden rising head. Providing the occurring condition of rough wet walls, the
SST k-ω model may be improper. Thus, turbulence models need to be compared with experimental
results under special circumstances.
Further researches were conducted for pressure loss due to viscosity. Various losses (Figure 4)
Energies
in ESP2020, 13, 5486
stages were defined [25]. Pump head is influenced by hydraulic loss, disc friction, leakage, 7 of 20

and so on. The hydraulic loss involves friction losses in the flow paths in the impeller and diffuser.
TheThe
on. dischydraulic
friction represents lossfriction
loss involves effects between
losses in impellers and diffusers.
the flow paths In addition,
in the impeller the influence
and diffuser. of
The disc
leakage is inevitable and should also be noted. The model is claimed to be applicable for various
friction represents loss effects between impellers and diffusers. In addition, the influence of leakage is
fluid
properties, rotational speeds, and pump types. Sun [26] established a one-dimensional analytical
inevitable and should also be noted. The model is claimed to be applicable for various fluid properties,
model for an ESP channel to achieve boosting head. In this work, the hydraulic friction and shock
rotational speeds, and pump types. Sun [26] established a one-dimensional analytical model for an
loss were taken into account. Vieira [27] combined several definitions of pressure loss via theoretical
ESP channel to achieve boosting head. In this work, the hydraulic friction and shock loss were taken
calculation to study the effects of viscosity in a single stage of an ESP. The dominant factors
into account. Vieira [27] combined several definitions of pressure loss via theoretical calculation to
considered were the impeller friction and disc friction losses. However, different combinations of
study the effects of viscosity in a single stage of an ESP. The dominant factors considered were the
parametrizations were needed for each flow case, due to the complexity of viscous effects. There was
impeller friction and disc friction losses. However, different combinations of parametrizations were
no general combination form that could explain all ESPs. With a two-dimensional laser Doppler
needed for each flow case, due to the complexity of viscous effects. There was no general combination
velocimeter (LDV), Li [7] also concluded that increased hydraulic loss and the disc loss were
form that could explain all ESPs. With a two-dimensional laser Doppler velocimeter (LDV), Li [7]
dominant factors. Furthermore, pressure changes between impellers and diffusers for various
also concluded that increased hydraulic loss and the disc loss were dominant factors. Furthermore,
viscosities were observed [11] to be quite close. In consideration of the strengthened recirculation due
pressure changes between impellers and diffusers for various viscosities were observed [11] to be quite
to flow regime transition, Zhu et al. [20] proposed a mechanistic model to predict pump performance.
close. In consideration of the strengthened recirculation due to flow regime transition, Zhu et al. [20]
On the basis of conventional Euler equations and known losses, the recirculation loss was introduced
proposed a mechanistic model to predict pump performance. On the basis of conventional Euler
to the model. With the Tulsa University Artificial Lift Projects (TUALP) database, it was validated to
equations and known losses, the recirculation loss was introduced to the model. With the Tulsa
be applicable for several types of pumps. Knowing related coefficients, the boosting pressure is not
University Artificial Lift Projects (TUALP) database, it was validated to be applicable for several types
hard to get. Thus, more accurate coefficients and better closure relationships could contribute to the
of pumps. Knowing related coefficients, the boosting pressure is not hard to get. Thus, more accurate
improvement.
coefficients and better closure relationships could contribute to the improvement.

Figure 4. (a) Head losses versus flow rate in an ESP (Divine 1993 [25]). The difference between an
Figure 4. (a) Head losses versus flow rate in an ESP (Divine 1993 [25]). The difference between an
ideal Euler formula and actual H-Q curve is shown, including vane losses, hydraulic losses, shock
ideal Euler formula and actual H-Q curve is shown, including vane losses, hydraulic losses, shock
losses, and leakage losses. (b) Power losses versus flow rate in an ESP (Divine 1993 [25]). The gap
losses, and leakage losses. (b) Power losses versus flow rate in an ESP (Divine 1993 [25]). The gap
between brake horsepower and fluid horsepower is divided into turbulence, friction, leakage, disc
between brake horsepower and fluid horsepower is divided into turbulence, friction, leakage, disc
friction, and bearing loss.
friction, and bearing loss.
There exist some special phenomena in multistage ESPs. Stel et al. [9] used CFD simulations to
There
explore exist and
average sometransient
special phenomena
flow featuresininmultistage ESPs.
a multistage Stel
ESP. et al. [9] used
Numerical CFD
results for simulations to
a single-stage
explore average and transient flow features in a multistage ESP. Numerical results
ESP are considered not convincible because the flow is heavily influenced by boundary conditions. for a single-stage
ESP are considered
Basically, nothas
the first stage convincible because therelative
its own uniqueness flow isto
heavily influenced
the following by boundary
stages. Considering conditions.
that the
Basically, the first stage has its own uniqueness relative to the following stages.
flow medium is water, pump performance with a high-viscosity fluid is even more complicated. Considering that the
flowOfuchi
medium is water, pump performance with a high-viscosity fluid is even more complicated.
et al. [10] also analyzed the interaction between stages and the effects of a previous stage on
the upstream flow, with a numerical study on three stages of a mixed-type ESP. The pressure head curve
of the first stage was saddle-shaped, which was greatly different from the following stages. As viscosity
increased, this effect was weakened. In general, the first stage provides a higher head compared to
other stages. In addition, separation zones appear near the second and third impeller trailing edges
when the rotational Reynolds number is lower. As the fluid is viscous enough, the separation zones
are yet smoothed out. The first stage is likely to vary in performance and flow pattern on different
operating conditions, as Stel concluded. The main reason is that the turbulence generated inside the
impeller spreads to the diffuser and the backflow of the diffuser affects the following stages. The ESP
geometry used in the numerical model was the same as that used by Amaral. Thus, the computational
Energies 2020, 13, 5486 8 of 20

results agreed with related experimental data. However, the effects of pump geometry were not taken
into account, which is also an important factor.
In terms of ESP performance degradation caused by high viscosity, classic and up-to-date
researches are summarized in Table 1. Key findings, drawbacks, and assumptions are listed in the
overview column.

Table 1. Summary of studies on ESP performance with high-viscosity fluid. SST, shear stress transport;
CFD, computational fluid dynamics.

Approach Article Overview

Hydraulic Institute Standards Correction coefficients were proposed within a
(1948) [17] narrow range of specific speeds.
Empirical correlation
A new correlation at constant specific speed
Stepanoff (1957) [14]
was proposed.
The change in flow pattern was observed through
Li (2000) [7]
experimental tests.
Experimental tests
The interaction between impellers and diffusers
Amaral et al. (2009) [8]
was proposed to be vital.
Skin friction factors were supposed to result in
Li (2014) [23] the sudden rising head effect with the standard
k-ε turbulence model in Fluent.
Performance of the first stage was found to be
different from that of the following stages with
CFD simulations Stel et al. (2015) [9]
the SST turbulence model in ANSYS CFX 14.5
(ANSYS Inc., Pittsburgh, PA, USA.).
Turbulence spread and backflow were used to
Ofuchi et al. (2017) [10] explain difference between stages with the SST
turbulence model in ANSYS CFX 14.5.
The flow regime transition was thought to
Zhu et al. (2016) [11] strengthen the recirculation with the SST
Experimental tests and turbulence model in ANSYS CFX 15.
CFD simulations A mechanistic model was proposed with
Banjar (2018) [28] oil–water emulsion effects studied using the SST
turbulence model in ANSYS CFX.
Degradation was confirmed at constant specific
Solano (2009) [15]
speed for off-design conditions.
Normalized specific speeds were proposed to
Ofuchi et al. (2017) [16]
predict performance without geometry.
Dimensionless analysis
Morrison’s number was defined to get the
Morrison et al. (2018) [21]
relationship between parameters.
The specific formula to get Morrison’s number
Patil and Morrison (2019) [22]
was given.
A new analytical model related to Reynolds
Ippen (1946) [24]
number was given.
A new analytical procedure was given with all
Gülich (1999a,1999b) [19,29]
the losses considered.
Analytical model A one-dimensional analytical model was given
Sun and Prado (2003) [26]
with hydraulic friction and shock loss included.
Combinations of pressure losses were proposed
Vieira (2015) [27]
to agree with experimental data.
A new mechanistic model was said to be
Zhu et al. (2019) [20]
applicable for all pumps.
Energies 2020, 13, 5486 9 of 20

As stated above, pump performance degradation would appear when fluid viscosity is high.
The initial response is to give empirical coefficients for correction. This includes not only experimental
fitting but also theoretical dimensionless analysis. Under operating conditions, pump performance
is much harder to predict. In order to find major sources, various definitions of pressure loss and
analytical models were then proposed. Increased hydraulic loss and disc loss between impellers and
diffusers were finally thought to be dominant factors. However, the flow pattern in ESP channels is not
yet achieved and the mechanism is not completely clear. With CFD simulations and experimental tests
combined, the problem can be better solved. In addition, phenomena in multistage ESPs are much
more complicated and difficult to explain. At present, improvement of the affinity laws is made on the
basis of large amounts of CFD simulation results. It decreases the time cost to use simulations rather
than experiments. Significantly, CFD models often oversimplify the geometry by using assumptions
related to factors such as leakage. There is a long way to go before the simulation results are completely
consistent with reality. The single-phase flow model is the basis of multiphase flow calculation, which is
also necessary because of the existence of bubbles and particles in fluid.

3. Two-Phase Flow in ESPs

Influenced by a bad operating environment, fluid in ESPs usually contains other components.
If insoluble impurities exist or sand particles are mixed during production, solid-liquid two-phase flow
would appear in the pump. If pumped liquid contains gas, gas-liquid two-phase flow arises. When both
are included, gas-solid-liquid three-phase flow comes into being, which is the difficulty in current
research. Common ESPs are applied with high flow rate and low gas content in subsea conditions [3].
Compared with radial-type pumps, mixed-type pumps are proven to have better performance with
inlet gas. When the gas void fraction (GVF) is large, energy loss between phases and the gas pocket can
make the pump head decrease. Studying bubble behavior with visualization equipment, experiments
were carried out to understand pump performance [30]. Lots of mechanistic models were put forward
to predict performance degradation. Nevertheless, some factors like pump geometry and slippage
between phases were overlooked [31]. Lack of versatility is also an important reason for its difficulty to
use. In view of the above problems, experimentally validated CFD simulations with bubble processing
models are commonly conducted to understand the flow feature.
Analogous to bubbles, much attention was paid to performance degradation due to existing
solid particles in ESPs [32,33]. Considering material and mechanistic differences between bubbles and
particles, adopted methods are different. Gas-liquid flow is studied using the Eulerian-Eulerian model,
which is also applicable to pressure drop calculation for solid-liquid flow because of the advantage of
turbulence computation. In order to achieve particle trajectories and erosion rates, the Lagrangian
particle tracking model was employed. However, the method cannot be applied for a high particle
volume fraction due to the assumption that solid particles do not occupy space. Differently, Marsis [34]
proposed that the discrete phase model (DPM) seems to be more sensitive to near-wall gradients and
the Eulerian model may be more accurate in predicting the erosion rate.
The existence of solid particles not only reduces pump operation efficiency, but also causes some
damage to pump elements. Wear in ESPs is composed of abrasive wear [35] and erosive wear [36].
According to the number of objects in contact, abrasive wear is divided into two-body wear and
three-body wear. For abrasive grains between two surfaces, positions in which abrasion usually occurs
are bearings, bushings, and clearances. Erosion usually occurs at the shrouds of impellers and diffusers,
impeller blades, and diffuser vanes. The direct cause is the impact of solid particles. In accordance
with impact angles, erosive wear includes shear wear and impingement wear, and the mechanism is
different. When the angle is small, shear wear takes place, and impact in and near the normal direction
leads to impingement wear. This article focuses on erosion, and abrasion is not included. It can be
seen from the formation mechanism that particle velocity is a dominant factor for erosion in ESPs.
Furthermore, target material properties are also important according to experience. It is known that the
erosion resistance of ductile material is higher than that of brittle material. Considering related factors
Energies 2020, 13, 5486 10 of 20

that influence the erosion rate, various models were proposed, including empirical equations [37–43]
and mechanistic models [44–49].

3.1. Erosion Models

In order to predict the erosion rate, empirical equations were firstly proposed on the basis of
direct impingement tests. The direct fitting method is used to consider the effects of particle velocity,
impingement angle, and material hardness. The general form for predicting the erosion rate is
given below.
ER = AV n f (θ) (3)

where ER is the erosion rate, V is the particle impact velocity, and θ is the impact angle. A and n are
constants related to material properties, and f(θ) takes different forms according to material properties.
For instance, for a brittle target material, the maximum value of f(θ) appears at 90◦ . With a ductile
material, however, the maximum erosion rate occurs at impact angles of 20–40◦ . Different forms of
functions mean different erosion mechanisms.
In the work of Ahlert [37], a particle shape coefficient Ks was further added. The value of Ks is 1
for angular, 0.53 for semi-rounded, and 0.2 for rounded particles. In addition, f is given as shown below.

2
aθ + bθ
 θ≤α
f (θ) = 

(4)
x cos2 θ sin θ + y sin2 θ + z

θ>α

The model provided two groups of constants for carbon steel and aluminum. For other materials,
the coefficients need to be adapted.
Taking carbon steel as a brittle material, Haugen [38] adopted the form of a polynomial function
for f. Unlike with direct fitting, Oka [39] used E90 as a reference erosion rate. The equations are
given below.
ER = 10−9 × ρt E90 f (θ)
Vp K2 dp K3
E90 = K(aH )bK1 ( V∗ ) ( d∗ ) (5)
f (θ) = (sin θ)n1 [1 + H (1 − sin θ)]n2
where ρt is the target material density, H is the target material hardness, Vp is the particle velocity, dp is
the particle diameter, V* and d* are the reference velocity and diameter, a and b are constants related to
material properties, and other coefficients are known constants. The model adopted the method of
comparison with the benchmark erosion rate and introduced target material properties in contrast to
former equations. Furthermore, a new form of f was proposed.
Zhang [40] replaced Ahlert’s impact angle function with a polynomial function and added
hardness directly to the erosion equation. According to Oka’s impact angle function, Mansouri [41]
drew hardness into the equation in the same way. The DET NORSKE VERITAS (DNV) model [42]
extended Haugen et al.’s model for both brittle and ductile materials. The impact angle functions
of different mechanistic models [33] were as shown in Figure 5. As the effects of material properties
are gaining more attention lately, the method of direct fitting is a simple but not precise approach.
Existing coefficients are not universal, and constants in the above models need to be determined again
by experiments in special circumstances.
Gülich [43] gave another empirical prediction of erosion in pumps. The method divided the whole
pump into several regions for consideration and set up coefficients for the related region. The value
was determined by geometry and flow parameters. Material properties were also taken into account
in the form of material factors. The entire procedure produced a simple tool to predict the erosion
rate. For the roughness of influencing factors, however, the accuracy of the prediction method is not
guaranteed. In addition, empirical equations can only be used for simple prediction and they cannot
explain the mechanism.
drew hardness into the equation in the same way. The DET NORSKE VERITAS (DNV) model [42]
extended Haugen et al.’s model for both brittle and ductile materials. The impact angle functions of
different mechanistic models [33] were as shown in Figure 5. As the effects of material properties are
gaining more attention lately, the method of direct fitting is a simple but not precise approach.
Existing
Energies coefficients
2020, 13, 5486 are not universal, and constants in the above models need to be determined11again
of 20
by experiments in special circumstances.

Figure 5. Impact angle functions of six erosion models (Zhu 2018

2018 [50]).
[50]).

Due to [43]
Gülich possible
gaverepeated impact, rebounding
another empirical prediction of
of particles
erosion in(Figure
pumps. 6) The
ought to be divided
method considered, the
and
whole pump into several regions for consideration and set up coefficients for the related region.used
empirical equations were proposed. The polynomial function as below is a commonly The
method,
value was although different
determined coefficients
by geometry were
and proposed
flow [51,52],
parameters. where eproperties
Material is the restitution coefficient,
were also taken into eN
and e
account are the normal and tangential restitution coefficient, θ is the impact angle,
T in the form of material factors. The entire procedure produced a simple tool to and A , A ,
1 predict
2 A , A
3 the4,
A 5 , A6 can
erosion beFor
rate. found in Table 2. of influencing factors, however, the accuracy of the prediction method
the roughness
is not guaranteed. In addition, empirical equations 2
can only
3
be used
4
for simple
5
prediction and they
Energies 2020,
cannot explain the x FOR e =A
13, mechanism.
PEER 1 + A2 · θ + A3 · θ + A4 · θ + A5 · θ + A6 · θ .
REVIEW 12 of 21

Due to possible repeated impact, rebounding of particles (Figure 6) ought to be considered, and
empirical equations were proposed. The polynomial function as below is a commonly used method,
although different coefficients were proposed [51,52], where e is the restitution coefficient, e N
and eT are the normal and tangential restitution coefficient, θ is the impact angle, and A1 , A 2 ,
A3 , A4 , A5 , A6 can be found in Table 2.

e = A1 + A2 ⋅ θ + A3 ⋅ θ 2 + A4 ⋅ θ 3 + A5 ⋅ θ 4 + A6 ⋅ θ 5 .

Table 2. Coefficients in the empirical model.

Figure 6. Particle reflection (θ1 and θ2 are particle velocity before and after impact) (Zhu 2019) [32].
A1 A2 A3 A4 A5
A6
Figure 6. Particle reflection ( θ1 and2 θ are particle velocity before and after impact) (Zhu 2019) [32].
Table 2. Coefficients
eN 0.993 −1.76 in1.56
the empirical
−0.49 model.
- -
Tabakoff
In order to explain theeTerosion
0.988
A1 −1.66
phenomenon,
A2 2.11 A3 −0.67
research on the
A4 -
impact ofA5 -
a single particle
A6 and
mechanistic erosion modelseNare 0.988
necessary. Finnie [44,45] firstly
−0.78 0.19 −0.024 0.0027 proposed a single-particle
- erosion
model for Forder
ductile
eN Basic assumptions
material.
0.993 −1.76
during model−0.21
−0.49were that
1.56application - impact particles
- are
Tabakoff e T 1 −0.78 0.84 0.028 −0.022
e T 0.988 −1.66 2.11 −0.67 - -
much harder than the surface and that the surface deforms plastically. The physical condition was that
eN surface
a particle impacts the target when the−0.78
0.988 impingement −0.024
0.19angle is 0.0027
θ. The erosion - below.
rate is given
Forder
eT 1 −0.78 0.84 −0.21 0.028 −0.022
cρt 2
ER = V f (θ ) (6)
pK ,
In order to explain the erosion phenomenon, research on the impact of a single particle and
whereerosion
mechanistic c is a nonideal
modelsparticle
are coefficient
necessary.taken as 1/2,
Finnie ρt is thefirstly
[44,45] target surface
proposedmaterial density, p is theerosion
a single-particle
surface
model for stress,
ductile and K isBasic
material. the ratio of vertical toduring
assumptions horizontal
modelforce. Similar to Ahlert,
application f wasimpact
were that dividedparticles
into are
two parts according to the angle.

 6 K
sin 2θ − K sin θ tan θ ≤
2

6
f (θ ) =  (7)
 K cos 2 θ K
tan θ >
 6 6 .
Finnie took K = 2, giving a transition angle of 18.43°. Additionally, the angle causing maximum
erosion rate was 16.85° when K = 2. Although the proposal of Finnie’s model is of historical interest,
Energies 2020, 13, 5486 12 of 20

much harder than the surface and that the surface deforms plastically. The physical condition was that
a particle impacts the target surface when the impingement angle is θ. The erosion rate is given below.
cρt 2
ER = V f (θ), (6)
pK

where c is a nonideal particle coefficient taken as 1/2, ρt is the target surface material density, p is the
surface stress, and K is the ratio of vertical to horizontal force. Similar to Ahlert, f was divided into
two parts according to the angle.

sin 2θ −
 6
K sin2 θ tan θ ≤ K
6
f (θ) = 

. (7)
 K cos2 θ

tan θ > K
6 6

Finnie took K = 2, giving a transition angle of 18.43◦ . Additionally, the angle causing maximum
erosion rate was 16.85◦ when K = 2. Although the proposal of Finnie’s model is of historical interest,
the prediction results above 45◦ are unacceptable and hard to improve.
Extending from Finnie’s model, Bitter [46,47] combined a ductile and brittle erosion model by
bringing in the concept of a threshold velocity. Erosion occurs only if the impact velocity is larger
than the threshold, when the target material is brittle. The suggested threshold velocity is related to
the Young’s modulus of elasticity and the Poisson’s ratios of the particle and the surface. The wear
curve for brittle material is incremental and reaches a maximum at 90◦ , and that for ductile material is
similar to Finnie’s (Figure 7). The same weakness appears whereby there is no erosion at zero impact
angle, which does not correspond with reality. Afterward, Neilson and Gilchrist [48] simplified Bitter’s
model to a certain extent with an adopted empirical constant. Ding [49] took repeated impact into
account with the kinetic theory of granular flow. Despite some progress, the mechanism of erosion is
still not2020,
Energies fully13,explained.
x FOR PEER REVIEW 13 of 21

Figure 7.
Figure Wear trend
7. Wear trend versus
versus impact
impact angle
angle for
for brittle
brittle and
and ductile
ductile materials
materials (Lyczkowski
(Lyczkowski 2002)
2002) [53].
[53].

The erosion rate is important to calculate local penetration rates of grid cells in CFD simulations.
The erosion rate is important to calculate local penetration rates of grid cells in CFD simulations.
For larger deviations of mechanistic models, empirical equations are more widely employed in
For larger deviations of mechanistic models, empirical equations are more widely employed in
commercial software. However, mechanistic models are necessary to understand the phenomenon.
commercial software. However, mechanistic models are necessary to understand the phenomenon.
With the help of CFD simulations, key parameters can be determined and new mechanistic models
With the help of CFD simulations, key parameters can be determined and new mechanistic models
are expected.
are expected.

3.2. Erosion in ESPs

It is known from erosion models that parameters that influence the erosion rate can be split into
two groups: primary parameters and secondary parameters. Primary parameters include particle size
and velocity, while secondary parameters involve material properties such as Young’s modulus of
Energies 2020, 13, 5486 13 of 20

3.2. Erosion in ESPs

It is known from erosion models that parameters that influence the erosion rate can be split into two
groups: primary parameters and secondary parameters. Primary parameters include particle size and
velocity, while secondary parameters involve material properties such as Young’s modulus of elasticity
and Poisson’s ratio. In ESPs, many particles impact pump elements. Thus, more macro parameters
such as solid concentration need to be considered. Furthermore, an ESP’s lifespan is influenced by
operating conditions such as flow rate, pressure, and load. A large number of experimental tests and
CFD simulations were conducted to understand the erosion phenomenon.
Levy and Chik [54] investigated the effects of particle hardness and shape on the erosion.
The erosion rate increased with higher hardness until 700 kg/mm2 (Vickers hardness number) when
particles were strong enough to not crack. Two shapes were compared, and the damage caused
by angular particles was four times that of spherical particles. However, the object material was
1020 carbon steel and the critical hardness may change for other materials. Although not universal,
a research method was provided where the influence of some parameters could be eliminated for
special materials.
Khalid and Sapuan [55] designed a wear equipment and conducted impeller rotating experiments
in slurry. By selecting specific positions of the wear surface for observation, the conclusion was drawn
that the rim of the impeller encounters more wear due to centrifugal force. In order to measure the
degree of wear, several parameters were compared such as weight loss, diameter loss, thickness loss,
and height loss of the impeller. The average height loss of the blade was finally determined to represent
the wear performance of the impeller, which could be used to predict the pump’s lifespan. The research
was mainly related to the erosion results and the mechanism was not given. Moreover, attention was
paid to the impeller and other parts of pumps were not mentioned.
Batalović [56] divided factors affecting erosion into three groups, including characteristics of solid
particles, pump material, and slurry flow. Theoretical analysis was combined with a statistical process
of experimental data in order to build a new model. The small deviation was attributed to difficulties
in precisely describing slurry flow. Although the erosion mechanism was not explained, the model
could be used to find the optimal materials and predict the working life as a simple tool.
Morrison [36] generated a mixed flow ESP under 117 h of erosion. Both overall head and efficiency
were observed to have declined. The primary cause was believed to be the significant clearance change
in the seals, which led to increased leakage. When calculating by theory, therefore, leakage needs to be
taken into account at some point.
Compared to costly and time-consuming experiments, the method of simulating erosion with the
CFD technique is widely used. In CFD simulations, the total removed material mass can be given by
calculating the cumulative results of particles with the discrete phase model (DPM). Due to two phases
(solid and fluid) existing in a pump simultaneously, interactions between them need to be taken into
account. Universally, the flow feature is important, and fluid governing equations need to be solved
firstly. Then, particle motion equations considering fluid force can be solved for particle trajectories.
During the process, the erosion rate is calculated. The iteration calculation of a time step ends here and
the flow field at the next moment is achieved when the next time step starts. In this manner, the erosion
in ESPs can be eventually predicted.
After determining coefficients using erosion experiments, Zhong [57] confirmed Bitter’s model.
A numerical prediction of wear in the casing was then conducted to study the effects of particle diameter
and flow rate on the erosion rate. With the increase in particle diameter and flow rate, the wear in the
pump casing increased. Combining experimental data and numerical calculation, the accuracy of the
prediction method was demonstrated.
Using Gülich’s empirical model along with CFD simulations, Kruger et al. [58] conducted some
research on the erosion in a radial flow centrifugal pump. Erosion processes were spilt into shock-like
and friction-like processes.
 diameter and flow rate on the erosion rate. With the increase in particle diameter and flow rate, the
wear in the pump casing increased. Combining experimental data and numerical calculation, the
accuracy of the prediction method was demonstrated.
Using Gülich’s empirical model along with CFD simulations, Kruger et al. [58] conducted some
research on the erosion in a radial flow centrifugal pump. Erosion processes were spilt into shock-
Energies 2020, 13, 5486 14 of 20
like and friction-like processes.
Noon and Kim [59] conducted three-dimensional numerical analysis for slurry through
centrifugal
Noon andpumps.Kim [59]Finnie’s model
conducted was adopted, and
three-dimensional a wear analysis
numerical map for for
theslurry
pumpthrough
casing was achieved.
centrifugal
The tongue
pumps. Finnie’s(near
modelθ was
= 35°) and belly
adopted, andarea (near
a wear mapθ =for300°) were damaged
the pump casing wasmost seriously.
achieved. Effects of
The tongue
temperature were also mentioned due to corrosion-erosion. Unlike Khalid and Sapuan
(near θ = 35 ) and belly area (near θ = 300 ) were damaged most seriously. Effects of temperature
◦ ◦ [55], emphasis
was put on the volute casing (Figure 8).
were also mentioned due to corrosion-erosion. Unlike Khalid and Sapuan [55], emphasis was put on
the volute casing (Figure 8).

Figure 8. (a) Centrifugal pump casing model (Noon 2016); (b) erosion rate density (Noon 2016) [59].
Figure 8. (a) Centrifugal pump casing model (Noon 2016); (b) erosion rate density (Noon 2016) [59].
Using Finnie’s model, Xiao et al. [60] also conducted systematic numerical research on the
evolution Using
of theFinnie’s
erosionmodel,
patternXiao et al. [60] also
in a centrifugal pump.conducted
With the systematic
blade surfacenumerical
damaged, research on the
the erosion
rateevolution of the erosion
in the impeller pattern
was reduced into
due a centrifugal
the change pump. With parameters.
in geometry the blade surface damaged,
The direct reason theiserosion
that
rate in the impeller was reduced due to the change in geometry parameters. The direct
local particle velocities decreased because of the strengthened recirculation and increased clearances. reason is that
local particle velocities decreased because of the strengthened recirculation and increased
The results showed that the flow characteristics and particle trajectories in pumps would be influenced clearances.
by The results material
the eroded showed in that the which
turn, flow characteristics
provides someand particle
advice for thetrajectories
design andin maintenance
pumps would of be
influenced
centrifugal pumps. by the eroded material in turn, which provides some advice for the design and
maintenance
Consideringofthecentrifugal pumps. Pirouzpanah [61] took balance holes as an additional mass flow
effects of leakage,
Consideringathe
inlet. Furthermore, new effects of leakage,
empirical model Pirouzpanah [61] tookkinetic
related to turbulence balance holes solid
energy, as anconcentration,
additional mass
flow inlet. Furthermore, a new empirical model related to turbulence
and particle velocity was proposed to predict the erosion rate. Turbulence kinetic energy was kinetic energy,
believedsolid
to be the dominant factor. Although it was a simple equation obtained by fitting, the prediction of the
other pump was verified.
Zhu et al. [32,33] conducted sand erosion experiments in an ESP and acquired paint-removal
photos. The most seriously damaged positions were the vane edge area in impellers and the diffuser
throat area. Numerical simulations with different turbulence models and erosion models in ANSYS
were also carried out. The SST k-ε model was considered best for erosion simulation. Due to differences
in average impingement velocity and angle, other turbulence models may produce large deviations.
The results were compared with the experimental data and great differences appeared in the erosion
rate magnitude for various erosion models. Oka et al.’s erosion model showed the least difference
with the actual situation for the impeller, while Haugen et al.’s model agreed best for the diffuser.
It can be seen from this that a single model may be inappropriate for erosion throughout the ESP.
Two mixed-type pumps and one radial-type pump were also compared, and less erosion was observed
with mixed-type pumps. Considering the complexity of the flow field, this conclusion deserves further
validity for different pump types.
Research on erosion in ESPs was conducted using erosion models, experimental tests, and CFD
simulations, as outlined in Table 3. Key findings, drawbacks, and assumptions are listed in the
overview column.
Energies 2020, 13, 5486 15 of 20

Table 3. Summary of studies on erosion in ESPs.

Approach Article Overview

Ahlert (1994) [37] The impact angle function was split into two parts.
Haugen (1995) [38] Carbon steel was taken as a brittle material.
Oka (2005) [39] The erosion rate was calculated with a reference situation.
Zhang (2007) [40] Ahlert’s model was extended with a polynomial function.
Empirical model
An ESP was divided into several parts according to
Gülich (2008) [43]
geometry and flow parameters.
Mansouri (2014) [41] A trigonometric impact angle function was introduced.
Haugen et al.’s model was extended for both brittle and
DNV (2015) [42]
ductile materials.
A single-particle erosion model for ductile material was
Finnie (1960,1972) [44,45]
proposed.
A threshold velocity was defined to combine ductile and
Bitter (1963) [46,47]
Mechanistic model brittle erosion model.
Neilson (1968) [48] Bitter’s model was simplified with an empirical constant.
Repeated impact of particles was included with the kinetic
Ding (1992) [49]
theory of the granular flow.
Particle hardness could be neglected when it was high
Levy (1983) [54]
enough.
Average height loss of the blade was determined to
Khalid (2007) [55]
Experimental tests represent the wear content.
A new statistical model was introduced to predict pump
Batalović (2010) [56]
working life.
Leakage was believed to cause the decrease in overall head
Morrison (2015) [36]
and efficiency.
Bitter’s model was confirmed, and prediction of pump
Zhong (1996) [57]
casing wear was conducted with programming calculation.
Erosion processes were split into shock-like and
Kruger (2010) [58]
friction-like processes with the Eulerian-Eulerian model.
The tongue and belly area were found to be damaged most
Noon (2016) [59]
seriously with Finnie’s model in ANSYS CFX.
CFD simulations The erosion rate was believed to be reduced during the
Xiao (2019) [60] evolution of pump wear with Finnie’s model in ANSYS
CFX 17.
Turbulence kinetic energy was considered the key
Pirouzpanah (2019) [61] parameter to predict the erosion rate with the
Eulerian–granular model in Fluent.
Various turbulence and erosion models were compared
Zhu (2019) [32,33] with experimental data of erosion with the
Eulerian-Lagrangian model in ANSYS Fluent 17.2.

Although erosion in ESPs is frequently observed, there exist difficulties in the quantitative
description of erosion degree in ESPs. A commonly used measure for erosive degree is the target
material removed by a unit mass particle. With the complexity of geometry, the wear of different
parts of an ESP is different. The vane edge area in impellers and the diffuser throat area are believed
to be damaged most seriously. The erosion rate was connected with particle velocity, flow rate,
and target material parameters by fitting. Therefore, empirical models were proposed as a simple
prediction method. For deeper understanding of this problem and better calculation accuracy, however,
Energies 2020, 13, 5486 16 of 20

the numerical model should go down to a level of single-particle impact. On the basis of these
efforts, continuous improvements were proposed to improve the prediction’s accuracy. In CFD
simulations, the Lagrangian particle tracking model was employed, as well as the Eulerian–granular
model. Combined with erosion models, numerical calculation is commonly used to predict the erosion
rate. Subject to various material properties and complex geometry, the numerical results of erosion in
ESPs struggle to match the experimental data. Even if realized, CFD simulations are hard to apply
universally. Further research needs to be conducted to completely understand the mechanism.

4. Challenges

4.1. Effects of Viscosity

The complex pump geometry for ESPs plays a large role in the viscous effect. Thus, special designs
need to be carried out in order to improve pump performance. Passage width and outlet angle of
the impeller [62,63] are believed to be key parameters considering the effects of viscosity. The best
ranges are 27.5–32.5◦ and 17–21 mm respectively. Before manufacture and production, dimensionless
analysis and CFD simulations are necessary to predict pump head and efficiency under operating
conditions. Although the viscous effect is relatively well understood at present, the specific speed
of pumps suitable for theory has a certain range. If the specific speed of the adopted pump is too
large or too small, the theoretical results are no longer credible. When the viscosity of pumping fluid
changes with shear rate or when a multistage ESP is going to be used, the current theory cannot yet
make accurate predictions.

4.2. Effects of Particles

When pump failure occurs, economic loss is attributed to the suspension of production, but not the
expenses for repairing pumps [5]. Thus, enough attention ought to be paid to the design procedure to
reduce possible wear, as well as emergency treatment after equipment failure. Materials more resistant
to wear should be used in key parts of ESPs to extend their working life [54]. Another way is to control
operating conditions such as the flow rate and rotational speed, which are key parameters of erosion.
When a sharp change in operation conditions happens or an ESP is about to be worn out, the repair or
even addition of a new ESP needs to be carried out in time. Currently, CFD is widely used to predict
the erosion rate. The most commonly used DPM in simulations, however, has its own limitations.
Because one of the theoretical assumptions is to treat particles as mass points without volume, DPM is
not suitable for scenes with solid particles occupying a large volume fraction [34]. As the number of
particles increases, the problem of long calculation time worsens. Moreover, the current treatment of
fluid-solid boundaries in CFD commercial software is relatively simple, which may cause a certain
deviation from reality. A method of taking into account both accuracy and calculation efficiency
is expected.

4.3. Effects of Bubbles

In addition to oil, there is natural gas in the reservoir. When the pressure drops to the bubble
point pressure, gas in crude oil is separated out and gas-liquid two-phase flow appears in the ESP.
Head degradation due to the presence of gas is known but the mechanism is not yet clear for complex
flow dynamics in an impeller. Various research methods were proposed to study the effects of geometry,
GVF, and suction pressure. A visual experimental device was designed and utilized to observe the
flow pattern [30]. The velocity profile of each file was simultaneously obtained. The gas pocket on
the blade pressure side was believed to be the main reason for head degradation. The essence of
efficiency decline is flow heterogeneity. In the gas-blocked zones, the flow path of the fluid is reduced,
which causes an increase in the flow rate and slippage on the gas-fluid interface. Empirical models [64]
were put forward on the basis of large amounts of experimental data. However, slippage, suction
pressure, and specific speed were rarely taken into account. Numerical calculations [65,66] were
Energies 2020, 13, 5486 17 of 20

also conducted to predict pump performance. In order to cover various bubble sizes, a new model,
the population balance model (PBM), was employed [67]. However, the bubble size distribution in
this model is discrete rather than continuous. Due to simplified assumptions according to complex
equations of two-phase flow, the accuracy of the prediction results of the analytical model [31] still
needs to be further improved. The presence of bubbles also complicates the influence on erosion and
the mechanism [12,13]. More work is needed to completely understand two-phase or even three-phase
flow in ESPs.

5. Conclusion
We reviewed classic and fundamental researches and up-to-date contributions to the study of
viscosity effects [7–11] and two-phase flow [30–33] on ESP performance. In the oil and gas industry,
the non-Newtonian and particle-rich nature of crude oil makes it crucial for considering pump
performance degradation [14,15] when designing the operation process. There are three main research
methods for investigating these effects: experimental tests, analytical approaches, and CFD simulations.
Dimensionless [21,22] and normalized analyses [16] can be used to predict pump performance as a
simple tool, but they are relatively incapable of investigating complex conditions such as two-phase flow
or unconventional geometrical structures. Although widely utilized, results from CFD simulations are
usually considered less convincing unless compared with experimental results, as they usually contain
oversimplified hypotheses. For two-phase flow in ESPs, different methodologies were adopted for
specific properties of particles and bubbles. The Eulerian-Lagrangian model [34] applies to solid-liquid
flow and the Eulerian-Eulerian model is usually used for gas-liquid flow. The presence of solid particles
can reduce pump efficiency and, more importantly, cause wear. The vane edge area in impellers and
the throat area in diffusers are considered the most damaged areas [59]. Despite some exploration,
the mechanism of particle erosion is not completely clear, and empirical models are frequently used for
erosion prediction. A novel model for particle erosion in ESPs with physical mechanics, which covers
both accuracy and efficiency, is well anticipated. When the effects of bubbles are added, interactions
between phases are non-negligible but challenging to model, which deserve further investigation.

Author Contributions: Conceptualization, P.L. and Y.W.; Writing—original draft preparation, P.L., Y.W., J.G.;
writing—review and editing, F.Y., C.N., X.O., J.X.; All authors have read and agreed to the published version of
the manuscript.
Funding: We acknowledge support from the National Natural Science Foundation of China (Grant No.
51804319), the Science Foundation of China University of Petroleum, Beijing (Grant No. 2462018YJRC002
and No.2462020YXZZ045), and National Key R&D Program of China (Grant No. 2017YFC0805800).
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Sukarno, P.; Saepudin, D.; Dewi, S.; Soewono, E.; Sidarto, K.A.; Gunawan, A.Y. Optimization of gas injection
allocation in a dual gas lift well system. J. Energy Resour. Technol. 2009, 131. [CrossRef]
2. Robello Samuel, G.; Saveth, K.J. Optimal Design of Progressing Cavity Pumps (PCP). J. Energy Resour. Technol.
2006, 128, 275–279. [CrossRef]
3. Joel Romero, O.; Hupp, A. Subsea Electrical Submersible Pump Significance in Petroleum Offshore Production.
J. Energy Res. Technol. 2013, 136, 012902. [CrossRef]
4. Takacs, G. Electrical Submersible Pump Manual, Design, Operations and Maintenance; Elsevier: Amsterdam,
The Netherlands, 2009. Available online: https://g0.njuu.cf/extdomains/books.google.com.sg/books?
hl=zh-CN&lr=&id=isUpDwAAQBAJ&oi=fnd&pg=PP1&dq=Electrical+Submersible+Pump+Manual,
+Design,+Operations+and+Maintenance&ots=jvs5sBjmw9&sig=k0pVzQyCg4rMfMLNWiooGpXFLWs#
v=onepage&q=Electrical%20Submersible%20Pump%20Manual%2C%20Design%2C%20Operations%
20and%20Maintenance&f=false (accessed on 10 February 2020).
5. Carvajal Diaz, N. Effects of Sand on the Components and Performance of Electric Submersible Pumps.
Master’s Dissertation, Texas A&M University, College Station, TX, USA, 2012.
Energies 2020, 13, 5486 18 of 20

6. Yedidiah, S. A study in the use of CFD in the design of centrifugal pumps. Eng. Appl. Comput. Fluid Mech.
2008, 2, 331–343. [CrossRef]
7. Li, W.-G. Effects of viscosity of fluids on centrifugal pump performance and flow pattern in the impeller.
Int. J. Heat Fluid Flow 2000, 21, 207–212. [CrossRef]
8. Amaral, G.; Estevam, V.; Franca, F.A. On the influence of viscosity on ESP performance. SPE Prod. Oper.
2009, 24, 303–311. [CrossRef]
9. Stel, H.; Sirino, T.; Ponce, F.J.; Chiva, S.; Morales, R.E.M. Numerical investigation of the flow in a multistage
electric submersible pump. J. Pet. Sci. Eng. 2015, 136, 41–54. [CrossRef]
10. Ofuchi, E.M.; Stel, H.; Sirino, T.; Vieira, T.S.; Ponce, F.J.; Chiva, S.; Morales, R.E.M. Numerical investigation of
the effect of viscosity in a multistage electric submersible pump. Eng. Appl. Comput. Fluid Mech. 2017, 11,
258–272. [CrossRef]
11. Zhu, J.; Banjar, H.; Xia, Z.; Zhang, H.-Q. CFD simulation and experimental study of oil viscosity effect on
multi-stage electrical submersible pump (ESP) performance. J. Pet. Sci. Eng. 2016, 146, 735–745. [CrossRef]
12. Basaran, B. CFD Simulation for the Erosion on Electrical Submersible Pump due to Viscosity and Air presence.
Master’s Dissertation, Texas A&M University, College Station, TX, USA, 2017.
13. Gerald Morrison, Y.; Daniel, S.; Yiming, C.; Changrui, B.; Abhay, P. Effect of Gas Presence on Erosive Wear
of Split-Vane Electrical Submersible Pump. In Proceedings of the Turbomachinery and Pump Symposia,
Houston, TX, USA, 11–14 December 2017.
14. Stepanoff, A.J. Centrifugal and Axial Flow Pumps: Theory, Design, and Application. 1957. Available online:
https://www.sciencedirect.com/science/article/abs/pii/0016003257900431 (accessed on 10 February 2020).
15. Solano, E.A. Viscous Effects on the Performance of Electrical Submersible Pumps (ESPs). Master’s Thesis,
University of Tulsa, Tulsa, OK, USA, 2009.
16. Ofuchi, E.M.; Stel, H.; Vieira, T.S.; Ponce, F.J.; Chiva, S. Morales, Study of the effect of viscosity on the head
and flow rate degradation in different multistage electric submersible pumps using dimensional analysis.
J. Pet. Sci. Eng. 2017, 156, 442–450. [CrossRef]
17. Hydraulic, I. Tentative Standards of Hydraulic Institute: Charts for the Determination of Pump Performance When
Handling Viscous Liquids; The Institute: NewYork, NY, USA, 1948.
18. Timár, P. Dimensionless Characteristics of Centrifugal Pump. Chem. Papers-Slovak Acad. Sci. 2005, 59, 500.
19. Gülich, J.F. Pumping highly viscous fluids with centrifugal pumps—Part 2. World Pumps 1999, 1999, 39–42.
20. Zhu, J.; Zhu, H.; Cao, G.; Zhang, J.; Peng, J.; Banjar, H.; Zhang, H.-Q. A New Mechanistic Model to Predict
Boosting Pressure of Electrical Submersible Pumps Under High-Viscosity Fluid Flow with Validations by
Experimental Data. SPE J. 2019. [CrossRef]
21. Morrison, G.; Yin, W.; Agarwal, R.; Patil, A. Development of Modified Affinity Law for Centrifugal Pump to
Predict the Effect of Viscosity. J. Energy Res. Technol. 2018, 140, 092005. [CrossRef]
22. Patil, A.; Morrison, G. Affinity Law Modified to Predict the Pump Head Performance for Different Viscosities
Using the Morrison Number. J. Fluids Eng. 2019, 141, 141. [CrossRef]
23. Li, W.-G. Mechanism for onset of sudden-rising head effect in centrifugal pump when handling viscous oils.
J. Fluids Eng. 2014, 136, 074501. [CrossRef]
24. Ippen, A.T. The influence of viscosity on centrifugal pump performance. Trans. ASME 1946, 68, 823–848.
25. Divine, D.L.; Lannom, R.W.; Johnson, R.A. Determining Pump Wear and Remaining Life From Electric
Submersible Pump Test Curves. SPE Prod. Facil. 1993, 8, 217–221. [CrossRef]
26. Sun, D.; Prado, M.G. Single-phase model for electric submersible pump (ESP) head performance. SPE J.
2006, 11, 80–88. [CrossRef]
27. Vieira, T.S.; Siqueira, J.R.; Bueno, A.D.; Morales, R.E.M.; Estevam, V. Analytical study of pressure losses and
fluid viscosity effects on pump performance during monophase flow inside an ESP stage. J. Pet. Sci. Eng.
2015, 127, 245–258. [CrossRef]
28. Banjar, H.M. Experiments, CFD Simulation and Modeling of Oil Viscosity and Emulsion Effects on ESP
Performance. Ph.D. Thesis, The University of Tulsa, Tulsa, OK, USA, 2018.
29. Gülich, J.F. Pumping highly viscous fluids with centrifugal pumps—Part 1. World Pumps 1999, 1999, 30–34.
[CrossRef]
30. Barrios, L.; Prado, M.G. Experimental visualization of two-phase flow inside an electrical submersible pump
stage. J. Energy Resour. Technol. 2011, 133, 042901. [CrossRef]
Energies 2020, 13, 5486 19 of 20

31. Noghrehkar, G.; Kawaji, M.; Chan, A.; Nakamura, H.; Kukita, Y. Investigation of Centrifugal Pump
Performance under Two-Phase Flow Conditions. J. Fluids Eng. 1995, 117, 129–137. [CrossRef]
32. Zhu, H.; Zhang, J.; Zhu, J.; Rutter, R.; Zhang, H.-Q. A Numerical Study of Turbulence Model and Rebound
Model Effect on Erosion Simulations in an Electrical Submersible Pump (ESP). In Proceedings of the
ASME-JSME-KSME 2019 8th Joint Fluids Engineering Conference, San Francisco, CA, USA, 28 July–
1 August 2019.
33. Zhu, H.; Zhu, J.; Rutter, R.; Zhang, H.-Q. A Numerical Study on Erosion Model Selection and Effect of Pump
Type and Sand Characters in Electrical Submersible Pumps by Sandy Flow. J. Energy Res. Technol. 2019, 141,
1–58. [CrossRef]
34. Russell, R.; Marsis, E. A state-of-the-art computational fluid dynamics simulation for erosion rates prediction
in a bottom hole electrical submersible pump. In Proceedings of the SPE Heavy Oil Conference-Canada,
Alberta, Canada, 11–13 June 2013.
35. King, D.; Traylor, F.; Stewart, R. Abrasion Technology for Electric Submergible Pumps. In Proceedings of the
SPE Annual Technical Conference and Exhibition, San Francisco, CA, USA, 5–8 October 1983.
36. Morrison, G.; Carvajal, N.; Saleh, R.; Bai, C. The Measured Impact of Erosion on the Rotodynamic and
Performance Characteristics of a Mixed Flow ESP. In Proceedings of the 31st International Pump Users
Symposium, Houston, TX, USA, 14–17 September 2015.
37. Ahlert, K.R. Effects of Particle Impingement Angle and Surface Wetting on Solid Particle Erosion of AISI 1018 Steel;
University of Tulsa: Tulsa, OK, USA, 1994.
38. Haugen, K.; Kvernvold, O.; Ronold, A.; Sandberg, R. Sand erosion of wear-resistant materials: Erosion in
choke valves. Wear 1995, 186, 179–188. [CrossRef]
39. Oka, Y.; Okamura, K.; Yoshida, T. Practical estimation of erosion damage caused part 2: Mechanical properties
of materials directly associated with erosion damage. Wear 2005, 259, 102–109. [CrossRef]
40. Zhang, Y.; Reuterfors, E.; McLaury, B.S.; Shirazi, S.; Rybicki, E. Comparison of computed and measured
particle velocities and erosion in water and air flows. Wear 2007, 263, 330–338. [CrossRef]
41. Mansouri, A. Development of Erosion Equations for Slurry Flows; Advisory Board Report; Erosion/Corrosion
Research Center, the University of Tulsa: Tulsa, OK, USA, 2014.
42. DNV, G. Recommended Practice–DNVGL-RP-O501–Managing Sand Production and Erosion; DNV GL: Akershus,
Norway, 2015; pp. 3–60.
43. Gülich, J.F. Centrifugal Pumps; Springer: Berlin/Heidelberg, Germany, 2008; Volume 2.
44. Finnie, I. Erosion of surfaces by solid particles. Wear 1960, 3, 87–103. [CrossRef]
45. Finnie, I. Some observations on the erosion of ductile metals. Wear 1972, 19, 81–90. [CrossRef]
46. Bitter, J.G.A. A study of erosion phenomena part I. Wear 1963, 6, 5–21. [CrossRef]
47. Bitter, J. A study of erosion phenomena: Part II. Wear 1963, 6, 169–190. [CrossRef]
48. Neilson, J.; Gilchrist, A. Erosion by a stream of solid particles. Wear 1968, 11, 111–122. [CrossRef]
49. Jianmin, D.; Lyczkowski Robert, W. Three-dimensional kinetic theory modeling of hydrodynamics and
erosion in fluidized beds. Powder Technol. 1992, 73, 127–138.
50. Zhu, H.; Zhu, J.; Rutter, R. Sand Erosion Model Prediction, Selection and Comparison for Electrical
Submersible Pump (ESP) Using CFD Method. In Proceedings of the ASME 2018 5th Joint US-European
Fluids Engineering Division Summer Meeting, Montreal, QC, Canada, 15–20 July 2018.
51. Grant, G.; Tabakoff, W. Erosion prediction in turbomachinery resulting from environmental solid particles.
J. Aircr. 1975, 12, 471–478. [CrossRef]
52. Forder, A.; Thew, M.; Harrison, D. A numerical investigation of solid particle erosion experienced within
oilfield control valves. Wear 1998, 216, 184–193. [CrossRef]
53. Lyczkowski, R.W.; Bouillard, J. State-of-the-art review of erosion modeling in fluid/solids systems. Prog. Energy
Combust. 2002, 28, 543–602. [CrossRef]
54. Levy, A.V.; Chik, P. The effects of erodent composition and shape on the erosion of steel. Wear 1983, 89,
151–162. [CrossRef]
55. Khalid, Y.A.; Sapuan, S.M. Wear analysis of centrifugal slurry pump impellers. Ind. Lubr. Tribol. 2007, 59,
18–28. [CrossRef]
56. Batalović, V. Erosive Wear Model Slurry Pump Impeller. J. Tribol. 2010, 132, 021602. [CrossRef]
57. Zhong, Y.; Minemura, K. Measurement of erosion due to particle impingement and numerical prediction of
wear in pump casing. Wear 1996, 199, 36–44. [CrossRef]
Energies 2020, 13, 5486 20 of 20

58. Krüger, S.; Martin, N.; Dupont, P. Assessment of wear erosion in pump impellers. In Proceedings of the 26th
International Pump Users Symposium, Turbomachinery Laboratory, Texas A&M University, College Station,
TX, USA, 16–18 March 2010.
59. Noon, A.A.; Kim, M.-H. Erosion wear on centrifugal pump casing due to slurry flow. Wear 2016, 364–365,
103–111.
60. Xiao, Y.; Guo, B.; Ahn, S.-H.; Luo, Y.; Wang, Z.; Shi, G.; Li, Y. Slurry Flow and Erosion Prediction in a
Centrifugal Pump after Long-Term Operation. Energies 2019, 12, 523.
61. Pirouzpanah, S.; Patil, A.; Chen, Y.; Morrison, G. Predictive erosion model for mixed flow centrifugal pump.
J. Energy Res. Technol. 2019, 141. [CrossRef]
62. Shojaeefard, M.; Tahani, M.; Ehghaghi, M.; Fallahian, M.; Beglari, M. Numerical study of the effects of some
geometric characteristics of a centrifugal pump impeller that pumps a viscous fluid. Comput. Fluids 2012, 60,
61–70.
63. Shojaeefard, M.; Tahani, M.; Khalkhali, A.; Ehghaghi, M.; Fallah, H.; Beglari, M. A parametric study for
improving the centrifugal pump impeller for use in viscous fluid pumping. Heat Mass Transfer. 2013, 49,
197–206.
64. Duran, J. Pressure Effects on ESP Stages Air-Water Performance; University of Tulsa: Tulsa, OK, USA, 2003.
65. Minemura, K.; Uchiyama, T. Three-dimensional calculation of air-water two-phase flow in centrifugal pump
impeller based on a bubbly flow model. J. Fluids Eng. 1993, 115, 766–771.
66. Minemura, K.; Uchiyama, T.; Shoda, S.; Egashira, K. Prediction of air-water two-phase flow performance of a
centrifugal pump based on one-dimensional two-fluid model. J. Fluids Eng. 1998, 120, 327–334.
67. Yiming, C.; Abhay, P.; Yi, C.; Changrui, B.; Yintao, W.; Gerald, M. Numerical Study on the First Stage Head
Degradation in an Electrical Submersible Pump with Population Balance Model. J. Energy Res. Technol. 2018,
141, 022003.

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional
affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).