Review

A Review of Experiments and Modeling of Gas-Liquid

Flow in Electrical Submersible Pumps
Jianjun Zhu * and Hong-Quan Zhang
McDougall School of Petroleum Engineering, The University of Tulsa, Tulsa, OK 74104, USA;
[email protected]
* Correspondence: [email protected]; Tel.: +1-918-760-6666

Received: 18 October 2017; Accepted: 3 January 2018; Published: 11 January 2018

Abstract: As the second most widely used artificial lift method in petroleum production (and first
in produced amount), electrical submersible pump (ESP) maintains or increases flow rate by
converting kinetic energy to hydraulic pressure of hydrocarbon fluids. To facilitate its optimal
working conditions, an ESP has to be operated within a narrow application window. Issues like gas
involvement, changing production rate and high oil viscosity, greatly impede ESP boosting
pressure. Previous experimental studies showed that the presence of gas would cause ESP hydraulic
head degradation. The flow behaviors inside ESPs under gassy conditions, such as pressure surging
and gas pockets, further deteriorate ESP pressure boosting ability. Therefore, it is important to know
what parameters govern the gas-liquid flow structure inside a rotating ESP and how it can be
modeled. This paper presents a comprehensive review on the key factors that affect ESP
performance under gassy flow conditions. Furthermore, the empirical and mechanistic models for
predicting ESP pressure increment are discussed. The computational fluid dynamics (CFD)-based
modeling approach for studying the multiphase flow in a rotating ESP is explained as well.
The closure relationships that are critical to both mechanistic and numerical models are reviewed,
which are helpful for further development of more accurate models for predicting ESP gas-liquid
flow behaviors.

Keywords: electrical submersible pump (ESP); multiphase flow; mechanistic modeling; CFD;
flow pattern; viscosity effect; drag coefficient; bubble size; in-situ gas void fraction

1. Introduction
The electrical submersible pump (ESP), as a high-efficiency downhole equipment for converting
kinetic energy to hydraulic pressure head, has improved significantly since it was invented in the
1910’s by the Russian engineer Arutunoff. Globally, there are more than 100,000 ESP installations,
making ESP the second most widely used artificial lift method in oil production, but the first based
on the produced amount [1,2]. ESPs excel in producing crude oils at very high flow rate, but they
have to be operated within a narrow application window. Gas involvement, changing production
rate and high oil viscosity can greatly affect the ESP performance [3,4].
Previous studies showed that the presence of gas would cause ESP hydraulic head degradation.
The flow behaviors inside ESPs under gassy conditions, such as pressure surging and gas pockets,
further deteriorate ESP boosting pressure. Surging may result in vibrations and short service life, and
gas pockets can severely limit liquid production rates [5]. Although handling gas-liquid mixture has
gradually become common for ESPs, the physical mechanism of two-phase flow affecting ESP
hydraulic performance is not well understood. The gas bubble formation, coalescence and breakup
mechanisms inside ESPs, which affect the two-phase flow characteristics, are still unclear. Due to the

Energies 2018, 11, 180; doi:10.3390/en11010180 www.mdpi.com/journal/energies

Energies 2018, 11, 180 2 of 41

compact and complex geometries of multistage ESPs, the visualization of internal flow structures and
bubble movement is very difficult.
The main objective of this paper is to provide a comprehensive review of literature concerning
experimental studies and modeling approaches for predicting ESP pressure increment under
gas-liquid flow conditions. The review helps better understand the multiphase flow characteristics
inside a rotating ESP, in order to develop mechanistic model to predict ESP boosting pressure under
gassy flow conditions. Section 2 presents the fundamental concepts and definitions of a multistage
centrifugal pump. Section 3 discusses the previous knowledge of ESP hydraulic performance and
flow structures under gassy conditions observed from experiments. Section 4 is divided into four
subsections. It begins with the empirical correlations, followed by one-dimensional two-fluid modeling
approach. Then, the CFD-based (Computational Fluid Dynamics) modeling and mechanistic
approaches are discussed. Finally, the closure relationships to make models solvable are evaluated
in Section 5.

2. Boosting Pressure of Centrifugal Pump

ESPs are widely used in petroleum industry to increase the hydrocarbon fluid production rates,
especially for off-shore deep-water oil fields. Normally, multistage ESP is assembled in series (Figure 1a)
with each stage comprising the rotating impeller and stationary diffuser (Figure 1b,c). In oil fields,
the ESP system usually contains hundreds of stages to meet the boosting pressure requirement due
to the great depth of reservoir. The impeller is the dominant part of ESP since it forces the fluid flow
and adds kinetic energy to the fluids by spinning the blades. At the outlet of impeller, the main part
of kinetic energy in the fluid is converted to pressure potential by diffuser vanes, which are fixed and
work as the guiding channels for the impeller of the next stage.

(b)

(c)

(a)
Figure 1. Main components of an electrical submersible pump (ESP). (a) Multistage assembly (figure
courtesy of http://www.franklinwater.com); (b) impeller components; (c) diffuser components [3].
Energies 2018, 11, 180 3 of 41

As a type of centrifugal pumps, ESPs can be classified into three different categories: radial, axial
and mixed types based on the dimensionless specific speed (NS). The non-dimensional NS is given by:

Q 1 2
NS  (1)
gH 3 4
where Ω is the rotational speed (rad/s), Q is liquid flow rate (m3/s), g is local gravitational acceleration
(m/s2), H is pump head (m). The pump industry uses a more practical expression for the specific
speed NS as below:

N q
Nq  (2)
h 0.75
where N, q and h are rotational speed (rpm), flow rate (gpm) and pump head (ft), respectively. Based
on Equation (2), centrifugal pumps are categorized into radial, mixed and axial types. The radial
pumps usually fall in the range 500 < NS < 1800, while the mixed pumps can reach a maximum
NS = 4500.
For a centrifugal pump, there are three important variables to characterize its hydraulic and
mechanical performance, namely: pumping head (H), efficiency (η) and brake horsepower (BHP).
Figure 2 below shows the typical pump performance curves provided in the product brochure, which
are obtained experimentally using tap-water.

Figure 2. Typical ESP pump performance curves (figure courtesy of Wood Group ESP, Inc., Oklahoma
City, OK, USA).

With the liquid flow rate increase, the ESP performance curves exhibit different trends. The
horsepower and pumping head render a monotonic increasing and decreasing trend versus flow rate,
respectively, whereas the efficiency is in a semi-quadratic relationship with the pump capacity. The
best efficiency point (BEP) presented with the dashed line, corresponds to the highest efficiency of
68.9% at Q = 2700 bpd, N = 3500 rpm in Figure 2. It is an important parameter in characterizing the
ESP overall performance. In this study, the main objective is to review different models in predicting
ESP hydraulic boosting pressure under gas-liquid flow conditions.

2.1. Euler Head

For an ESP impeller, the velocity triangles at the intake and discharge are shown in Figure 3. All
the variables are in International System of units (SI). The absolute velocity C can be decoupled into
two components: relative velocity W and peripheral velocity U, which is calculated by = Ω × .
Energies 2018, 11, 180 4 of 41

The subscripts 1 and 2 denote the inlet and outlet of impeller, respectively. Here, W is relative to the
impeller and C is relative to the global coordinate and is equal to the vector summation of U and W,
e.g., = + . In Figure 3c, C’2 is an ideal absolute velocity assuming infinite number of impeller
blades, while C2 is the real absolute velocity at the impeller outlet.

(a) (b) (c)

Figure 3. Velocity triangles. (a) Schematic of impeller flow channel; (b) inlet velocity triangle; (c) outlet
velocity triangle.

According to the ideal conservation law of angular momentum in rotating centrifugal pump,
the Euler head (HE) can be expressed as [6]:
   
C2  U 2  C1  U1 U 2C2U  U1C1U
HE   (3)
g g
where C1U and C2U are the projection of absolute velocities at the impeller inlet and outlet to the
direction of peripheral velocities. Applying the velocity triangle relationship in Figure 3, one can obtain:

U 22  U 12 W12  W22 C 22  C12

HE    (4)
2g 2g 2g
and:

 2 r22  r12  Q  1 1 
HE     
2gh  tan  2 tan 1 
(5)
g
where r is the radius of impeller, h is the channel height, β is the blade angle. If the fluids enter the
impeller without pre-rotation, Equation (5) can be written as:

2 r22 Q
HE   (6)
g 2gh tan 2

2.2. Head Loss Mechanisms

As shown in Equation (6), the ideal Euler head HE is in linear relationship with liquid flow rate.
In reality, pressure losses in the impeller, the diffuser, and losses from the interaction between them
must be deducted from the ideal Euler head, including friction, shock and recirculation losses.
Therefore, the actual pump head can be calculated by:

H  H E  H frition  H shock  H leakage  H recirculat ion  H diffuser  H disk (7)

Figure 4a shows the change of the ideal Euler head with the outlet blade angle β. Figure 4b shows
the losses to the ideal head and the final H-Q curve of a centrifugal pump according to Equation (7).
Friction loss becomes prominent at high flow rates. In contrast, leakage loss is more at relatively low
Energies 2018, 11, 180 5 of 41

flow rates. Shock loss takes place when the liquid flow rate differs from the designed flow rate.
Tables 1–5 below discuss each pressure loss mechanism by presenting existing models that are
available in literature.

(a) (b)
Figure 4. Schematic of head curves. (a) Euler heads with different outlet blade angles; (b) actual head
after deducting losses [7].

Table 1 summarizes the studies on the friction losses inside a centrifugal pump impeller. Takacs [3]
pointed out that friction losses progressively increase with liquid rate and are due to fluid friction in
the impeller.

Table 1. Friction loss models in literature.

Reference Model Coefficient

Ito [8], Jones [9], Churchill [10], −
ℎ = =
Shah [11], Sun [12] 8
Wiesner [13], Sun and Prado [14], ( − )( + )
ℎ = b2—constant
Thin et al. [15] 8
+
Ito and Nanbu [16], Bing et al. [17] ℎ =
4
Zhu [1] ℎ = fFI—empirical constant
2

The mechanism of shock losses in centrifugal pump is still not well studied. Thus, only empirical
correlations are available in the literature. Table 2 lists the existing prediction models of shock loss as
implemented in the prediction models. Shock losses are negligible at the best efficiency point (BEP)
of the pump, but increase at lower and higher liquid flow rates. They occur at the entrance and the
exit of the impeller and are caused by sudden changes in the direction of flow [3].

Table 2. Shock loss models in literature.

Reference Model Coefficient

Stepanoff [6], Amaral [18]
ℎ = ( − ) kshock—empirical constant
Thin et al. [15]
Wiesner [13]
−
Sun and Prado [14] ℎ =
2
Thin et al. [15]

Zhu [1] ℎ = fTI—empirical constant

2
Energies 2018, 11, 180 6 of 41

As suggested by Tackas [3], the leakage losses represent the losses through the clearances
between the rotating and stationary parts of the pump stage (at the impeller eye, through balancing
holes, etc.), which decrease with increase of liquid rates. Table 3 below shows the calculation models
available in literature.

Table 3. Leakage loss models in literature.

Reference Model Coefficient

Bing et al. [17] ℎ =
2
3
Zhu [1] ℎ = + fLK—empirical constant
4 2

In a rotating centrifugal pump, recirculation loss is interpreted as the dissipated fluid energy
which is continuously obtained from the shaft due to adverse pressure gradient from the impeller
inlet to outlet. As liquid flow rate decreases, the adverse pressure gradient increases, resulting more
severe head loss due to fluid recirculation. Currently, the agreement has not been reached as for the
definition of recirculation, which in turn a fully understanding in terms of recirculation loss
mechanism is unavailable by far. However, the empirical correlations have been proposed in
literature, as summarized in Table 4 below.

Table 4. Recirculation loss models in literature.

Reference Model Coefficient

Gulich [19,20] ℎ =0
ℎ = ℎ(3.5 ) /
= arcsin
( )

Sun and Prado [14], Bing et al. [17] = 8 × 10

0.75 1
=1+ −1
1− +

ℎ =0 Q > QBEP
Tuzson [21], Thin et al. [15] .
ℎ = Q ≤ QBEP
−
Zhu [1] ℎ = C2E—effective outlet velocity
2

Table 5. Diffuser loss models in literature.

Reference Model Coefficient

−
Ito [8], Jones [9], Churchill [10], Shah [11], Sun [12] ℎ =−
8
( + )
Sun and Prado [14], Bing et al. [17] ℎ =
4
ℎ = − ( )
=
Amaral [18] .
= =
−

Sun and Prado [12,14,22] claimed that the diffuser loss is caused mainly by the friction on the
diffuser walls. Table 5 presents the calculation formulae of diffuser losses that were implemented in
the prediction models. Most equations are similar to the friction loss formulae in Table 1.
Disk friction losses, as Stepanoff [6] defined, are due to the contact between a rotating disk and
fluid. Thus, additional power is consumed to keep the disk rotating since the viscous shear forces act
on the disk surfaces [7]. Table 6 shows some of existing models of calculating disk friction losses in
literature.
By deducting the pressure losses from Euler head in Equation (7), the real pressure increment of
a centrifugal pump is obtained for single-phase flow. As can be seen in the Tables 1–6, different
pressure loss models are available in literature. A proper combination of these models may output
Energies 2018, 11, 180 7 of 41

an optimized estimation of pump performance. Yet, due to the empirical nature of these models, the
universal validity of such combination to deduce the most appropriate prediction model is questioned.
Thus, more and more researchers resort to numerical approaches to calculate the hydraulic performance
and understand the complicated flow structures inside rotating ESPs.

Table 6. Disk loss models in literature.

Reference Model Coefficient

Sun and Prado [14] −
ℎ =
Amaral [18] 2
Thin et al. [15] ℎ =
10
Van Esch [23] ℎ = cm—empirical constant
2
ℎ = 1−
. ⁄
Gulich [19,20] = + . ⁄( ) ℎ = 0.05
Ladouani [24] .
= −2 × 10
10

3. Experimental Studies
The ESP boosting pressure depends on many hydraulic factors including pump geometry, fluid
properties and multi-phase flow conditions. In recent years, with more and more installations of ESP
in oil production systems, the effects of fluid viscosity and gas entrainment on ESP’s pressure
boosting ability have been the focal areas of research interests.

3.1. Single-Phase Tests

For single-phase flow, ESP performance is sensitive to the fluid viscosity. With the increase of
oil viscosity, ESP boosting pressure becomes lower corresponding to the same flow rate. By the same
token, the flow rate through the ESP decreases corresponding to the same boosting pressure in ESPs.
Ippen [25] conducted over 200 performance tests for oil viscosities up to 10,000 Saybolt Second
Universal (SSU) on four variants of centrifugal pumps. The experimental results were summarized
by plotting the ratio of oil head to water head, brake horsepower and efficiency against a Reynolds-
type dimensionless number, based on which the general correction factors for specific speeds from
800 to 2200 were proposed to correlate pump’s boosting pressure under viscous fluid flow.
Hydraulic Institute provided a typical empirical approach with correction factors to estimate the
conventional centrifugal pump boosting pressure for viscous liquid flow if the water performance
were known. However, the accuracy of this approach was questioned by Gülich [19,20] and Li [26]
since the experiments carried out by Hydraulic Institute were within a narrow range of the pump
specific speeds. Unreasonable errors were found if extrapolation was beyond that range.
Stepanoff [6] proposed a similar Reynolds-type number by using only one correction factor to
get the new H-Q curves if the water performance were known. A more general model based on the
evaluation of viscous dissipation for disk and hydraulic frictions to predict the boosting pressure of
centrifugal pumps was proposed by Gülich. The friction losses on disk and in flow passage, as the
author claimed, were dominating factors impairing centrifugal pump’s ability to handle high viscosity
fluids. Compared with available data, Gülich also pointed out that friction losses were affected
significantly by pump geometry, fluid properties and thermal conditions. More recent experimental
studies conducted by Amaral et al. [27] and Solano [28] further revealed that the Hydraulic Institute
charts and empirical correlations available in literature were unable to give appropriate correction
factors to predict the ESP boosting pressure for viscous oil flow. The discrepancies between
experimental measurements and predictions by existing correlations or charts are discussed in
Barrios et al. [29] and Banjar et al. [30].
Energies 2018, 11, 180 8 of 41

3.2. Two-Phase Tests

The pioneering experimental studies conducted by Murakami and Minemura [31,32] investigated
centrifugal pump performance with gas entrainment. By employing a semi-opened impeller pump
with a transparent casing, they experimentally observed the behavior of entrained air bubbles.
The decreasing total head of the pump caused by air admission and the work consumed for air
delivery were reported. Since then, investigators have conducted more experimental measurements
and mathematical modeling on centrifugal pump performance under gassy flow conditions.
Experimental researches of ESP performance under gas-liquid flow conditions have been carried out
extensively at the Tulsa University Artificial Lift Project (TUALP) in the past decades. Table 7 lists a
summary of these studies.

Table 7. Experimental studies on ESP two-phase performance conducted at TUALP.

Study Content Pump Fluid

GN4000
Cirilo [33] Compare two-phase flow performance of three different ESPs Air/water
GN7000
ESP gas-liquid performance with an advanced gas handler
Romero [34] GN4000 Air/water
installed upstream
Measure stage-by-stage pump pressure increment of a multistage
Pessoa [35] GC6100 Air/water
ESP
Beltur [36] Investigate pressure surging in ESP and affecting factors GC6100 Air/water
Duran [37] Correlate experimental data of ESP two-phase performance GC6100 Air/water
Investigate pump rotational speed effect on ESP two-phase
Zapata [38] GC6100 Air/water
performance
Visualize the internal flow of a 2nd stage ESP under gas/liquid
Barrios [39] GC6100 Air/water
flow conditions
Visualize ESP two-phase flow pattern using a similar pump
Gamboa [40] GC6100 Air/water
prototype as Barrios [37]
Air/oil
Trevisan [41] Visualize ESP internal flow under air/viscous-liquid flow GC6100
visualization
Banjar [30] Investigate ESP performance with air/oil flow DN1750 Air/oil
Investigate ESP gas/liquid performance with various flow
Salehi [42] TE2700 Air/water
conditions
Oil/water
Croce [43] Investigate ESP performance with water/oil emulsion flow DN1750
Emulsion
Investigate ESP gas/liquid flow performance with/without Air/water
Zhu [1] TE2700
surfactant injections Surfactant

As can been seen in Table 7, the previous experimental studies, which were conducted by Cirilo [33],
Pessoa [35], Beltur [36], Duran [37], Zapata [38], Barrios [39], Gamboa [40], and Salehi [42] among
others, covers a wide range of flow conditions. Cirilo built the experimental flow loops for testing
ESP performance at the TUALP. Using water and air as the working fluids, he measured both the
water and air-water performances of three different types of ESPs as a function of GVFs (the
volumetric fraction of gas phase at the ESP inlet), intake pressure and rotational speeds. The obtained
data indicated that the mixed type pumps were able to handle much higher GVFs (>30%) than radial
type pumps (<10%). With necessary modifications of the same testing flow loops, Pessoa conducted
experimental investigations of two-phase performance of ESP using a 22-stage GC6100 pump. By
monitoring the stage-by-stage pressure increment, his results revealed that the ESP average behavior
was significantly different from that observed per stage. Also, phenomena like surging and gas
locking were observed and their boundaries were mapped. Additionally, a second region after
pressure surging was observed in mapping test curves, where the slope of pressure increment versus
flow rate changed again.
Using the same experimental flow loop of Pessoa, Beltur, Duran, Zapata and Salehi conducted
extensive experimental measurements of ESP performance under both liquid- and gas-liquid flow
conditions. Beltur focused on ESP performance deterioration with the presence of gas for varying
Energies 2018, 11, 180 9 of 41

intake GVFs and pressure. Data analysis revealed that the intake GVF is the most important factor
affecting ESP boosting pressure under gassy flow. A higher deterioration of pumping head occurs at
GVFs above 6%. Duran and Zapata developed empirical correlations for predicting the pressure
increment across the stage and the flow regime boundaries. Zapata also carried out further
measurements with a wide range of rotational speed to study its effects on the average efficiency of
ESP. Salehi did similar measurements using a 14-stage TE2700 ESP, on which the effects of the stage
number, intake pressure and fluid properties were investigated. It was found that ESP two-phase
boosting pressure varied stage by stage only when the GVF exceeded a certain value, below which
the deterioration was mild and independent of stage number. However, the degradation of pumping
head became more prominent and affected by the stage number if the GVF reached a critical value,
at which the pump boosting pressure of the downstream stage is better than upstream ones.
A typical testing result of ESP two-phase performance with pure water-air flow is shown in
Figure 5 [40], including surging (Figure 5a) and mapping tests (Figure 5b). For surging test, the liquid
flow rate is fixed, but gas flow rate or inlet gas volumetric fraction (GVF) varies. On the contrary, the
liquid flow rate changes while the gas flow rate is fixed in mapping test. In Figure 5a, different curves
correspond to different pump intake pressures. Obviously, the pump pressure increment declines
non-linearly with the increase of inlet GVF. For mapping test results in Figure 5b, each curve
corresponds to a different gas flow rate, which increases from the top to bottom curves. With a fixed
gas flow rate, a sudden drop of pump stage pressure increment is observed when flow rate is reduced
to a certain value.

(a) (b)
Figure 5. ESP two-phase pressure increment [40]. (a) Surging test; (b) mapping test.

A more recent experimental study regarding surfactant effect on ESP gas-handling ability and
boosting pressure under gassy flow was conducted by Zhu et al. [44]. Surfactants are molecules with
a hydrophobic and a hydrophilic part, and therefore they preferentially adsorb at the interface of
continuous/dispersed phases. In the process, they reduce the interfacial tension [45]. Studies of
surfactant effect on bubble/droplet formation in air/water or water/oil binary immiscible two-phase
flows were carried out by Hu et al. [46], Omer and Pal [47]. Surfactant effect on liquid loading in gas
well has been studied by van Nimwegen et al. [48,49], Ajani et al. [50,51]. Both reported a lower
critical gas velocity at which liquid loading occurs, meaning that the surfactant presence helps
prevent liquid loading in wells.
However, the studies on the surfactant effect on centrifugal pump performance under two-phase
flow conditions are few. Zhu et al. [44] measured the ESP two-phase pressure increment with/without
surfactant injection in the flow loop. The surfactant used in their study is isopropanol alcohol (IPA).
The decline trend of normalize pressure (Np) in Figure 6a versus the intake GVF improves with
surfactant injections, and the pressure surging phenomenon depicted by a sudden drop of ESP
boosting pressure disappears. Figure 6b compares ESP H-Q performance curves under air-water flow
with different IPA volumetric concentrations. A clear difference can be seen at low QL in terms of the
stage pressure increment. Without IPA injection, the pressure increment drops to zero if QL becomes
very low. But, the ESP performance improves significantly with surfactant presence.
Energies 2018, 11, 180 10 of 41

(a) (b)
Figure 6. Comparison of ESP two-phase pressure increment with/without surfactant injection [44].
(a) Constant liquid flow rate testing; (b) constant gas flow rate testing.

3.3. Flow Pattern Recognition

In multiphase pipe flow, flow pattern varies due to phase interaction and interfacial momentum
transfer changes, which depend on fluid properties, flow rates and pipe geometry. Many studies on
the recognition of flow patterns and their transition boundaries are available in literature [52–54].
Flow pattern is important to accurately predict the liquid holdup and pressure gradient in pipe flow.
Take the upward gas/liquid flow in a vertical pipe for example, the main flow patterns are bubble,
slug, churn and annular flows as illustrated in Figure 7a. Correspondingly, the flow pattern map
showing the transition boundaries between different flow patterns is plotted in Figure 7b with the
superficial velocities of gas and liquid as its coordinates.
The dispersed bubble flow prevails if the gas flow rate is low and the liquid flow rate is high, as
shown in Figure 7a. For this flow pattern, the turbulence forces are high enough to breakup gas-phase
into small bubbles and disperse them in the continuous liquid-phase [52].
When both gas flow rate and liquid flow rate are low, gas bubbles become larger and move
upward in a zigzag path. The bubble flow regime may or may not exist depending on the pipe
diameter. Confined by maximum lattice packing, the in-situ gas void fraction of bubble flow should
be below 0.52 [55]. However, for a practical application, the maximum gas void fraction below 0.25
is commonly used for vertical co-current pipe flow.

(a) (b)
Figure 7. Typical flow patterns and transition boundaries in vertical pipe, (a) main observed flow
patterns [54]; (b) transition boundaries and flow pattern map [52].
Energies 2018, 11, 180 11 of 41

Slug flow regime corresponds to the steady alternating flow of gas pockets (Taylor bubble) and
liquid slugs [53] at even higher gas flow rates (see Figure 7b), where Taylor bubbles occupy the entire
cross section of the pipe except for a thin liquid film on the wall. Slug flow can be considered as
steady-state flow pattern since its slug length, frequency, translational velocity etc. are time-
independent. However, a further increase of gas flow rates triggers the collapse of slugs and lead to
an unstable flow that is described as the churn flow regime due to the large degree of turbulence in
the flow [54].
Annular flow is characterized by extremely high gas flow rates and continuous gas core flowing
upward with thin liquid film on the pipe wall. Due to high interfacial shear and drag force effects,
the liquid droplets from the wavy interface between gas core and liquid film can be swept into the
gas core, and pushed upward by the high-speed gas phase.
Several visualization studies have been conducted to observe multiphase flow structures/patterns
inside a rotating centrifugal pump. Table 8 below lists the visualization experiments on flow patterns
inside a centrifugal pump in literature. Most of these studies used transparent casing to visually
observe the pump internal flow patterns. Starting with the original volute-type centrifugal pumps
[31,32,56–70], they cover several research topics including flow pattern recognition, bubble
movement visualization, mapping transition boundaries etc.

Table 8. Experimental studies on two-phase flow patterns inside a centrifugal pump.

Study Approach Observation

Murakami and Minemura 1st study that associated pump perfor-mance
Transparent casing
[31,32,56] with gas-liquid flow pattern inside an impeller
Small bubble flow and stationary large bubble
Patel & Runstadler [57] Transparent casing coinciding with significant reduction of pump
head
Transparent casing The evolvement of dispersed bubble to slug
Sekoguchi et al. [58]
electric resistivity probe flow and a large gas pocket
Kim et al. [59] Transparent casing Phase slippage leads to bubble agglomeration
Bubble flow to separated flow as gas flow rate
Sato et al. [60] Transparent casing
increases
Bubble accumulation coincided with large
Takemura et al. [61] Transparent casing
pressure gradient
Gas filled space starts to develop at the suction
Chisely [62] Transparent casing
pipe of the pump
Bubble accumulates at pressure side of impeller
Suryawijaya et al. [63] Transparent casing
blade
Estevam [64] Transparent casing 1st study of flow patterns in ESP
Gas bubbles, gas pocket and blockage of pump
Poullikkas [65] Transparent casing
flow
Larger accumulation of bubbles is observed at
Thum et al. [66] Transparent casing pressure side than suction side, four flow
patterns: bubble, plug, slug and stratified flow
Barrios [39,67] Transparent casing Bubbly flow and segregated flow are observed
Gamboa [40] Transparent casing 1st study of flow pattern map in ESP impeller
Measurement of gas void distribution in
Schäfer et al. [68] HireCT, non-intrusive
multiphase centrifugal pump
Measurement of gas void distribution in
Neumann et al. [69] HireCT, non-intrusive
multiphase centrifugal pump
Flow patterns in ESP: bubble, agglomerated
Verde et al. [70] Transparent casing
bubble, gas pocket, segregated flow

Murakami and Minemura [31,32] experimental study first investigated the gas entrainment
effect on the hydraulic pump head of a centrifugal pump visually, and associated the degradation of
pump performance with gas-liquid flow pattern inside the rotor. Using a transparent casing, Estevam [62]
conducted the first visualization experiment of flow pattern recognition inside a rotating ESP impeller.
Energies 2018, 11, 180 12 of 41

Later visualization experimental work, conducted by Gamboa [40] presented similar observations as
Estevam’s study, based on which the first flow pattern map with transition boundaries between
different flow patterns inside an ESP was proposed. His work offers significant insight regarding
how to characterize the flow patterns in an ESP quantitatively.
Among previous visualization experiments, two recent studies from Schäfer et al. [68] and
Neumann et al. [69] are different from the others. They both used a new non-intrusive technique,
the so called high resolution computed tomography (HireCT) to accurately measure the distribution
of in-situ gas void fractions. Compared to the intrusive technique, the new measurement approach,
as the authors claimed, can provide better image resolution of the gas void distribution and thus a
higher measurement accuracy.
By visualizing the ESP internal flow, Barrios [39], Barrios and Prado [67] observed that bubbles
enlarged when inlet GVFs increased and pump rotational speeds decreased. Such enlargement
corresponding to the poorer pump performance indicated that bubble behaviors played a significant
role in ESP’s ability of handling gas-liquid mixtures. In addition, visualization experiments also
showed different flow patterns prevailing inside ESP channels at higher GVFs.
In Figure 8, λG denotes the intake GVF. It is evident that flow behaviors inside ESP impeller
altered significantly with flow conditions. From Figure 8a,b, the GVF increase caused formation of
larger bubbles and gas pockets, which in turn choked the flow passage for liquids and decreased the
pump hydraulic head. Moreover, the flow patterns prevailing in ESP impeller at a specific value of
λG, are comparable to that of gas-liquid two-phase flow in pipelines.

(a) (b)

(c) (d)
Figure 8. Impeller channel flow behavior at N = 600 rpm and QL = 174 bpd with various gas flow rates.
(a) λG = 0.15%; (b) λG = 0.23%; (c) λG = 0.39%; (d) λG = 1.05% [39].
Energies 2018, 11, 180 13 of 41

Similar as the gas/liquid pipe flow, Estevam [64] and Estevam et al. [71] categorized the flow
patterns in a rotating ESP impeller as bubbly flow, transition flow, and elongated-bubble flow.
A more thorough experimental study on flow pattern recognition of two-phase flow in ESP impeller
was carried out by Gamboa [40], Gamboa and Prado [72]. Under a given flow condition, the typical
flow pattern map from their visualization experiments is shown in Figure 9, where the curves
denoted by different symbols represent the transition boundaries between flow patterns, including
homogenous flow, bubbly flow, gas-pocket formation, and segregated flow. Figure 9 is important to
understand the two-phase flow mechanisms in a rotating ESP. From the perspective of mechanistic
modeling, each flow regime corresponds to the specific governing equations for flow characteristics,
such as bubble size (db), in-situ gas void fraction (αG), slippage velocity (Vslip) between gas and liquid
phases. With the flow pattern prevailing inside a rotating ESP determined, the governing equations
based on mass and momentum conservations can be simplified.

1.2

Homogenous Boundary
1
Normalized Liquid Rate [fraction]

Homogenous Bubbly Flow

0.8
Regime Regime
Unstable Gas Pocket
0.6

0.4
Stable Gas Pocket
Regime

0.2
Segregated Flow Regime
0
0 0.05 0.1 0.15 0.2
Normalized Gas Rate [fraction]

Figure 9. Flow pattern map for GC-6100 ESP at 2400 rpm and 150 psig [39].

Verde et al. [70] conducted visualization experiments on flow pattern recognition inside a
rotating ESP impeller using high speed and resolution imaging technique. As shown in Figure 10,
four different flow patterns were classified, including bubble flow (Figure 10a), agglomerated bubble
flow (Figure 10b), gas pocket flow (Figure 10c) and segregated flow (Figure 10d). They observed that
the intensity of pump performance degradation is directly influenced by the flow pattern within the
impeller. The occurrence of the gas pocket flow pattern is linked to the intensification of the
deterioration of pump performance and the appearance of operation instabilities. Moreover, the
segregated flow patterns correspond to the severe performance degradation which makes the pump
incapable of generating pressure.
Summarized by Verde et al. [70], the schematic representations of each flow pattern are shown
in Figure 11 with intake GVF increasing from left to right. Due to the relatively small intake GVF, the
homogenous flow regime is featured by tiny and evenly-dispersed bubbles inside impeller channels,
as shown in Figure 11a. Under this regime, the bubbles are deemed to move together with liquid
phase. Slippage between gas and liquid is small, meaning the in-situ αG is almost the same as λG.
As the intake GVF increases, the tiny bubbles are prone to collide and aggregate to form bigger
ones, resulting in bubbly flow regime. In contrast to homogenous flow regime, the phase slippage
between gas and liquid, shown by Figure 11b can no longer be neglected. Thus, depending on the
intake GVF, the in-situ αG under bubbly flow becomes higher than λG. A further increase of GVF
induces more severe collision and aggregation of bubbles so that the large gas pocket forms. This
Energies 2018, 11, 180 14 of 41

flow pattern is similar as slug/churn flow patterns in pipelines, which are featured by a gas core/pocket
followed by a liquid slug. As shown in Figure 11c, the Taylor-bubble-like gas pocket forms near the
suction side of ESP impeller, which occupies a significant portion of the impeller flow passage.

(a) (b)

(c) (d)
Figure 10. Visualized flow patterns in a rotating ESP impeller [68]. (a) Bubble flow pattern;
(b) agglomerated bubble flow pattern; (c) gas-pocket flow pattern; (d) segregated flow pattern.

(a) (b) (c) (d)

Figure 11. Schematic representations of flow patterns inside a rotating ESP impeller [70], (a) bubble
flow; (b) agglomerated bubble flow; (c) gas-pocket flow; (d) segregated flow.

A relatively high gas flow rate and low liquid flow rate lead to the segregated flow pattern,
similar to the annular/stratified flows in pipe. As it can be seen in Figure 11d, the elongated bubble
expands and occupies the entire impeller length. For practical applications, the segregated flow
pattern corresponds to gas-locking, an operational failure which gives null pump head and null flow
rate [70].
As a direct observation of flow patterns inside the ESP impeller, visualization experiments can
help reveal gas-liquid flow behaviors intuitively. However, the experimental facility needs special
designs associated with necessary modifications on pump geometries, such as the removal of
Energies 2018, 11, 180 15 of 41

impeller hub and attaching the Pyrex glass on its top for visualized observation. Barrios [39] pointed
out that the visualization of two-phase flow structures in a multistage ESP assembled in series was
much more difficult. Thus, how to characterize the multiphase flow in ESPs has become a challenging
topic. Although several technologies [68,69,73–75] to visualize the internal flows inside a centrifugal
pump were discussed, they required some modifications on pump geometry and the implementation
was difficult to carry out. It required mounting HireCT into the apparatus so that the internal flow
structures could be visualized regardless of the opaque pump casing or volute. In addition, the data
processing involved time-averaged rotation-synchronized CT scanning techniques, adding further
complexity in analyzing the obtained experimental results.
Figure 12 shows the measurements using the HireCT technique by Schäfer et al. [68]. The
horizontal axis denotes the inlet GVF and the vertical axis is the volumetric averaged in-situ αG in the
rotating pump impeller. Clearly, there is a sharp αG increase at a GVF corresponding to the severe
gas-pocket formation and pumping head degradation. The step change is between inlet GVF = 2.5%
and 3.0%. As confirmed by Schäfer et al. observation, such change is due to the rapid flow pattern
change from bubbly flow to intermittent flow.

Figure 12. In-situ gas void fractions at various intake GVFs measured by HireCT [68].

The aforementioned experimental studies on ESP gas-liquid performance used tap water as the
working fluid, while the gas phase was compressed air or nitrogen. Several recent experimental
studies focusing on ESP gas-handling ability under viscous fluid flow were conducted by Trevisan [41],
Trevisan and Prado [76], Banjar et al. [30], and Paternost et al. [77]. Using a visualization prototype
built from original ESP components and with minimal geometrical modifications, Trevisan [41],
Trevisan and Prado [76] conducted experiments to investigate the viscosity effect on liquid/gas
two-phase flow through ESP. The authors identified four liquid/air flow patterns inside the impeller
channels: agglomerated bubble, gas pocket, segregated gas and intermittent gas flows. It was
concluded that the agglomerated bubble flow is responsible for pressure surging phenomenon,
which is the initiation of pump head deterioration due to gas entrainment. The authors also observed
that the increase in viscosity caused surging to occur at relatively lower inlet GVFs. Similar
experimental observations were made by Banjar et al. [30]. Paternost et al. [77] further investigated
the performance of a centrifugal pump handling single-phase viscous liquids and analyzed the
impact of free gas entrainment. They concluded that the degradation of pump head was due to the
stagnation of large gas-pocket formation, which became worse with the increase of liquid viscosity.
An empirical correlation similar to the dimension analysis procedure in Solano [28] was proposed to
correlate ESP pressure increment under two-phase flow conditions accounting the inlet GVF and
fluid viscosity.
Energies 2018, 11, 180 16 of 41

4. Modeling Approaches
In this section, several empirical correlations for calculating centrifugal pump performance
under gas/liquid two-phase flow are reviewed. Although empirical correlations are much easier to
implement compared to other numerical or mechanistic models, the application range is rather
limited. With this consideration, the more sophisticated one-dimensional two fluid models coupled
with interfacial momentum transfer mechanisms are discussed. The 3D CFD simulations, based on
the conservation laws of mass and momentum, can provide detailed numerical results in terms of
flow structure, phase distribution and slippage etc. However, CFD simulations for multiphase flow,
which usually demand high computational cost, are not as reliable or robust as for single-phase flow.
Furthermore, both one-dimensional two-fluid model and CFD-based numerical model are extremely
sensitive to the closure relationships, such as bubble size, drag force model and other interphase
momentum transfer terms, etc. Thus, the mechanistic models rooted in physics, including flow
pattern, mass and momentum conservation equations, have become a new direction for developing
the best prediction model of boosting pressure in ESPs with gas involvement.

4.1. Emperical Correlations

The first empirical correlation based on experimental data of ESP pressure degradation due to
free gas entrainment was proposed by Turpin et al. [78]:

 q   346430  qg 410  
H m  H exp  g     
 ql   C1 pin 2  ql C1 pin  
(8)
  
where, Hm is the head of ESP with gas/liquid flow (ft); H is the single-phase pump head (ft); qg and ql
are gas and liquid flow rates (bpd), respectively; pin is pump intake pressure (psi); C1 = 0.145 is a unit
conversion constant. As the authors claimed, a critical constraint of stable flow condition inside ESPs
should be applied to this model so as to achieve a reasonable prediction accuracy. Thus, another
empirical correlation to distinguish the stable/unstable flow boundaries are given as below:

 qg ql 
  2000  (9)
 3 pin 
The stable flow condition corresponds to  < 1, which is the recommended application range
of Equation (8).
Romero [34] established a multiphase head model for the type of mixed flow pumps based on
Cirilo’s [33] experimental data:

qld   qld  

2
Hm  q
 1   a   ld  1 (10)
H max  qd max   qd max  qd max 
 
where Hmax is the shut-in pump head in ESP; qld is the dimensionless liquid flow rate, a ratio of the
intake liquid flow rate to the open flow rate (OFR). The application range for Equation (10), as pointed
out by the authors, is dispersed bubble flow or low degradation of ESP performance:

a  2.902G  0.2751 (11)

qd max  1  2.2035G (12)

To extend the application range of above two empirical correlations, Duran and Prado [79]
presented models for the mild and severe head degradations in ESPs, which correspond to bubbly
and elongated bubble flow regimes:
Energies 2018, 11, 180 17 of 41

 ql qg
 1   l H 1    g H 
pm  
 q 1     (13)
 0.47075 0.21626ln g 
  qmax 
where Δpm is the pressure increment of single stage ESP with gas/liquid flow (psi), α is the in-situ gas
void fraction. Equation (13) is applied to bubbly flow and elongated bubble flow, respectively.
A closure relationship is presented below to solve for α in above equations:

 m  ql 1    
1.622

  0.843  0.85 
qg 1     l  qmax 
 (14)
qmax  ql 1   
0.435
 1.6213qmax

Similarly, Equation (14) corresponds to Equation (13), which can be used to calculate the in-situ
gas void fraction and then the pressure increment in ESP with gas presence can be obtained
accordingly. Although the application range of Equations (13) has been extended compared to
Equation (10), its formulation is only based on the experimental data from a single ESP geometry
without any validation against different pump models. Thus, its general validity is questioned due
to its empirical nature.

4.2. One-Dimension Two-Fluid Modeling

Zakem [80] first developed a mathematical model using a one-dimensional control volume
method to analyze gas bubbles and liquid interaction for straight blade impellers. Furuya [81]
developed a similar analytical model by incorporating the pump geometry, void fraction, flow
slippage, and the flow regime but neglecting the compressibility and condensation effects. The basic
formula in Furuya study is given by:

m l2 1  1 d 1 d dn  m g2 1  1 d 1 d dn 

     
l dn2 1  1 ds dn ds  g dn2  2   ds dn ds 

  m    1 d 1 d dn 
2 2
 1 d 1 d dn   l
m
 l  g          
2  gdn
(15)
  ds dn ds   l 1 dn  1 ds dn ds 
 
2
3C  l
m m 
 l  g 1r2 sin cos  d l   g 
8 rb  l 1dn gdn
and are the mass flow rates of liquid and gas. s is the streamline coordinate. n is the coordinate
normal to the streamline coordinate s. β is the angle between relative flow velocity and
circumferential direction. γ is the angle between the radial direction and the stream surface. Cd is the
drag coefficient. rb is the bubble radius. Comparing with experimental data in literature, the model
predictions were within the relative average error band of ±30% for GVF <20%, and ±50% for GVF
>30%.
Sachdeva et al. [82,83] conducted a comprehensive investigation of two-phase flow in ESPs with
air/water and diesel/CO2 mixtures. A dynamic five-equation, one-dimensional, two-fluid model
accounting for pump geometry, intake pressure and GVF, fluid properties, was developed and
verified to predict ESP boosting pressure. The basic equations of Sachdeva model are presented
below.
Energies 2018, 11, 180 18 of 41

The mass balance equations are:

dGGWGr sin 
0 (16)
ds

d 1G LWLr sin 

0 (17)
ds
where r is radius in ESP, WG and WL are the mass flow rates of gas and liquid, respectively.
The momentum balance equations are:

VR,G P r
GGVR,G  G GGr2  FW,G  Fi  Fv (18)
s s s

VR,L P r
1G LVR,L  1G   1G Lr2  FW,L  Fi  Fv (19)
s s s
where VR,G and VR,L are the radial components of gas and liquid absolute velocities, respectively. FW,G
and FW,L are the friction forces of gas and liquid against the channel walls per unit volume of fluids.
Fi is the interfacial friction force between gas and liquid per unit volume. Fv is the virtual mass force
per unit volume.
Minemura et al. [84] also studied the performance of centrifugal pumps in the nuclear industry
under air-water two phase flow conditions with a low inlet GVF (<10%). Based on energy change in
the flow from the rotating impeller to the stationary volute, a 1D, two-fluid model considering fluid
viscosity and air-phase compressibility in a rotating impeller was proposed. This model can be solved
numerically with a prediction error of ±20% of the related flow capacity. However, both Sachdeva et al.
and Minemura et al. models are valid only under a narrow application range or specific experimental
conditions. Compared to Sachdeva model, the major difference in Minemura et al. model is that the
momentum balance equations are based on the relative velocities of gas/liquid rather than the radial
components of the absolute velocities, which are given as:

WG P r
G GWG  G  G G r 2  FW ,G  Fi  Fv (20)
s s s

1  G  LWL WL  1  G  P  1  G  Lr2 r  FW ,L  Fi  Fv (21)

s s s
Based on Sachdeva et al. and Minemura et al. one-dimensional two-fluid models, Sun [12,22]
developed a new two-phase model including a set of one-dimensional mass and momentum balance
equations for predicting ESP performance. He also improved analytical models for wall frictional
losses and shock loss, as well as new correlations for the drag coefficient. The general momentum
balance equation along the streamlines is given by:

dp  dWp  dp  ds M p,s ds 
   pWp   p2r      (22)
dr streamline  dr  ds  f , p dr  p dr  streamline
where the subscript p is g or l for gas or liquid phase, respectively. And Mp,s is the interfacial
momentum transfer term. Eliminating the pressure increment term at the left hand side in Equation (24),
the combined momentum balance equation can be expressed as

  dp   dp   1
lWl dWl   gWg dWg  l  g 2rdr        ds  Ml ,s ds (23)
  ds  f ,l  ds  f , g  l g
Energies 2018, 11, 180 19 of 41

Applying a finite differencing algorithm, a preferred numerical scheme [85], to solving this
model, a good agreement of pump performance curve, αG distribution, surging and gas lock
conditions against correspondent experimental measurements was obtained. Although many studies
for modeling ESP performance under gassy conditions were conducted, mechanistic modeling is still
needed due to the over-simplification and assumptions or narrow application ranges of the existing
models.

4.3. Three-Dimension Computational Fluid Dynamics (CFD)

With the advances of computer technology, computational fluid dynamics (CFD) becomes a
more and more powerful tool to study centrifugal-pump performance under single-phase and
multiphase flow conditions. Due to complicated ESP geometries, it is difficult to investigate the
internal velocity and pressure fields experimentally. However, CFD offers an alternative way to
simulate the complex internal flow structures. A centrifugal pump consists of an impeller rotating at
a set angular velocity and a volute which is stationary. For an ESP, the rotating and stationary parts
are the impeller and diffuser, which are accommodated in the rotating and stationary computational
domains in CFD, respectively.
Using a 3D CFD code with the frozen-rotor interface model, the internal flow inside centrifugal
pumps can be simulated, including velocity and pressure fields [86–88] as well as flow recirculation
and separation [89,90]. The frozen-rotor model is considered as steady-state simulation because it
holds the rotating and stationary parts in two separate reference frames. Some transient simulations
were conducted using the sliding-mesh technique to investigate the dynamic flow structures in
centrifugal pumps [91–94]. Gonzalez et al. [91] performed numerical simulations of unsteady flow in
a single-phase centrifugal pump, considering impeller-volute interaction. By solving viscous,
incompressible Navier-Stokes (N-S) equations with the sliding mesh technique, the unsteady flow
behavior inside a centrifugal pump due to impeller-volute interaction was captured. A relationship
between the global variables, such as torque, as a function of impeller relative position, secondary
flow in volute, etc., was obtained numerically. This approach was successful in predicting the
dynamic interaction between impeller and volute. Huang et al. [93] studied unsteady flow and
pressure fluctuations due to interaction between impeller and diffuser vanes by the sliding mesh
technique. His study confirmed that the global variables are primarily affected by impeller blade
passing frequency.
In addition to designing turbomachinery, CFD can help engineers study the viscosity effects on
centrifugal pump performance. Shojaeefard et al. [95,96] conducted both experimental study and
numerical simulation of a centrifugal pump handling viscous fluids. The authors stated that a good
agreement between simulation and experimental data was obtained by solving the steady state
Reynolds-averaged Navier–Stokes (RANS) equations with Shear Stress Transport (SST) k-ω
turbulence model. Using the same pump geometry, Sirino et al. [97] and Stel et al. [98] performed
numerical investigations of viscosity effects on single-stage and three-stage ESPs, respectively.
Similar numerical methodologies were used in their work including SST turbulence model with
transient rotor-stator techniques. Both simulation results matched experiments well under a wide
range of fluid viscosity. Stel et al. [98,99] pointed out that CFD simulated boosting pressure with
multistage ESP geometry agreed with experimental results better than that with a single-stage
geometry. The phenomenon of rising head with moderate increase of fluid viscosity was studied by
Li [100]. By implementing the standard k-ε turbulence model and non-equilibrium wall function into
RANS equations, the author confirmed that the rising pump head was due to the transition from
hydraulically rough regime to hydraulically smooth regime. Zhu et al. [84] solved a set of 3D, steady-
state RANS equations with standard SST turbulence model using the commercial CFD solver ANSYS
CFX by employing the frozen-rotor technique. Their results matched correspondent experimental
data well.
CFD has also been used to simulate pump performance with gas involvement, such as cavitation
phenomenon [101,102], and free-gas entrainment flow [103–105]. Unlike single-phase simulations,
the two-phase simulation requires the solution of conservation equations of mass, momentum and
Energies 2018, 11, 180 20 of 41

energy for the continuous and dispersed phases. Meanwhile, another set of constitutive equations,
for describing the interphase interactions, such as interfacial momentum/heat transfer, need to be
solved simultaneously.
Minemura and Uchiyama [106] solved 3D momentum equations with a finite-element method
to predict gas/liquid flow behavior in a rotating centrifugal pump impeller. Tremante et al. [107]
numerically studied gas/liquid flow through a cascade axial pump by CFD simulation. Coupled with
a modified k-ε turbulence model by considering the viscosity of the liquid phase and the
compressibility of the gas phase, the gas/liquid distributions versus different attack angles were
obtained. Caridad et al. [108,109] studied ESP impeller performance handling water/air mixtures
using CFD simulation. Applying two fluid models in 3D CFD simulations, the pressure and velocity
fields, as well as the gas phase distributions (Figure 13) were obtained. The gas pocket near the impeller
blade was also identified and compared with experimental observations.

Figure 13. Typical multiphase CFD simulation results of gas/liquid flow in ESPs [108].

The sensitivity analysis on GVF and bubble diameter indicated that ESP performance
deteriorated with GVF and bubble diameter increase. Barrios [39], Barrios et al. [103] conducted
multiphase CFD simulations on a single-stage ESP impeller with the new models of bubble size and
drag coefficient predictions. Their simulations agreed with laboratory visualization images of
streamlines and gas-accumulation zones. Qi et al. [110] designed ESPs for geothermal application
with high-temperature gas-liquid two-phase flow. Using CFD simulations, the designed mix-type
centrifugal impeller and diffuser were optimized for better gas-handling and higher efficiency within
a wide range of production rate. Zhu and Zhang [104] conducted multiphase CFD simulations on a
three-stage ESP with each stage comprising of an impeller and a diffuser. Comparing to experimental
measurements, their work revealed that bubble size was a critical factor affecting ESP performance
under gassy conditions. A new bubble size prediction model was later proposed in Zhu and Zhang [105]
based on CFD simulations. They also predicted the in-situ gas void fraction (αG) with a mechanistic
model and validated the results with numerically simulated values (Zhu and Zhang [90]). A typical
CFD simulation result of gas/liquid flow inside a rotating ESP from their studies is shown in
Figure 14 below.
Multiphase flow phenomena in ESP are transient in nature, such as fluctuations of local pressure,
gas void fractions, and breakup or coalescence of bubbles. To better simulate the hydrodynamics of
gas-liquid two-phase flow in a rotating centrifugal pump, the unsteady CFD simulation code coupled
with multiphase flow model and transient rotor-stator algorithm to account for the interactions
between impeller and diffuser should be adopted. However, the computational cost for transient
CFD simulations is far more than that for steady-state simulations. Marsis et al. [111] performed
transient two-phase CFD simulations on eight multi-vane ESP designs. The predicted pump
performance was confirmed by experimental results. The final design was achieved with the stage
Energies 2018, 11, 180 21 of 41

pressure increased by 4% for single-phase flow and 23% for two-phase flow at inlet GVF = 20% by
optimizing the meridional profile and number of blades.
Yu et al. [112,113] conducted unsteady numerical simulation on gas-liquid flow in a multiphase
centrifugal pump. Considering multiple interfacial momentum transfer components, such as drag
force, lift force, virtual mass force (VMF) and turbulent dispersion force (TDF), they concluded that
the two-fluid multiphase model was able to capture the transport process more accurately than the
homogeneous model. Compared to turbulent dispersion force, the drag force plays a more dominant
role, as shown in Figure 15a. In Figure 15, z is in the streamwise direction, while L is the maximum
streamline as the fluid flows from the intake to discharge of the computational domain. As bubble
size (Db) may vary in a rotating centrifugal pump, its effect on the local drag force is limited, which
can be verified in Figure 15b.

(a) (b)
Figure 14. CFD predicted gas void fraction distribution and the corresponding pressure contour [90],
(a) gas void fraction color map; (b) pressure contour on half impeller span.

(a) (b)
Figure 15. Magnitude analysis of interface forces by CFD simulations [112], (a) magnitude ratios of
non-drag forces to drag force; (b) bubble size effect on drag force magnitude.
Energies 2018, 11, 180 22 of 41

Huang et al. [94] investigated the transient inhaling process of gas initially filling a section of
pipe into centrifugal pump impeller by sliding mesh and transient-frozen-rotor methods. The phase
distribution, pressure and velocity versus time were computed and analyzed. Pineda et al. [114]
presented an alternative approach to obtain the distribution and concentration of the dispersed phase
in a rotating centrifugal pump by solving the realizable k–ɛ turbulence model coupled with the
volume of fluid (VOF) multiphase model. They observed that the numerically simulated results of
in-situ αG could be correlated by the Lockhart-Martinelli parameter. Zhang et al. [115,116] compared
the unsteady CFD simulation results of multiphase flow patterns in a three-stage centrifugal pump
with the corresponding visualization experiments. Good agreement on the positions and shapes of
the gas pocket was observed. Ye et al. [117] combined the advanced transient CFD multiphase
simulation with finite-element-analysis (FEA) to optimize the 3D metal printing process of hybrid
stage prototype and ESPs for high-gas wells. A design-validation tool was developed and a prototype
ESP was manufactured which can pump up to 75% GVF gas/liquid mixture without gas locking.

4.4. Mechanistic Modeling

Unlike multiphase pipe flow, whose flow structures and characteristics are easier to visualize
and quantify, the mechanistic modeling of ESP performance under gas/liquid flow is more difficult
due to its highly twisted flow channel and compact assembly. At present, only limited studies are
available in the literature related to this topic, such as Barrios [39], Gamboa [40] and Zhu [1]. Similar
to the mechanistic model of multiphase pipe flow, the flow patterns in ESP two-phase flow should
be identified first, based on which the corresponding flow models can be developed.
Based on Sachdeva [82] and Sun [12] momentum equations in Equation (24) along the streamline
inside a rotating centrifugal pump, Zhu [1] derived the combined momentum equation of gas and
liquid phases by assuming the Taylor bubble flow pattern prevailing in rotating ESP two-phase flow:

 L vT  vF vS  vF   C vT  vC vS  vC   F SF

 
lF H LF A
(24)
 1 1 
 I SI      L  C 2 RI  0.
 H LF A 1  H LF A 
streamline

where the subscripts L, F, S, C, LF, and I denote liquid, film, slug, gas core, liquid film and interface,
respectively. vT is the translational velocity. Compared to the combined momentum equation for slug
flow in pipe [115–116,118–120], the difference in Equation (24) is the body force term. All the velocities
are the relative velocities to the ESP channel. By applying the flow pattern and necessary assumptions,
Equation (24) can be reduced to the simplified flow model, under which the flow characteristics and
pump performance can be solved.

4.4.1. Flow Pattern Map and Transition Boundary

As aforementioned, the visualization of flow patterns inside ESP with gas/liquid flow either
involves difficult experimentation to install transparent glasses so as to enable the visualization of
internal flow fields [39,70] or requires the complicated and expensive non-intrusive instrumentation [66].
However, the identification of flow patterns and transition boundaries inside a rotating ESP impeller
is still not well understood. Gamboa and Prado [72] proposed an alternative approach to recognize
the flow patterns of ESP multiphase flow by relating the inflection points on pump performance
curves. Figure 16 below illustrates the identification process schematically.
Energies 2018, 11, 180 23 of 41

Figure 16. Flow pattern recognition by interpreting performance curves [72].

Gamboa and Prado [72] ascribed the inflections on each performance curve to the flow pattern
transitions inside an ESP with gas/liquid flow. As the authors pointed out that such analogy can
merely achieve a rough accordance, it did offer significant convenience to obtain the transition
boundary among flow patterns. In Figure 16, the horizontal axis is normalized liquid flow rate, and
the vertical axis is the measured pump pressure increment. Compared to single-phase performance
curve with no inflections, it is found that the two-phase performance curve suffers from significant
pressure degradation with inflections at varying loci as qld drops. Different markers representing the
inflection points on each mapping test performance curve correspond to different flow pattern
transitions. The white circles denote the transition boundary between dispersed bubble flow and
bubbly flow; green diamond shows the transition from bubbly flow to intermittent flow; green square
corresponds to the transition from intermittent flow to segregated flow.
Transition from Dispersed Bubble Flow to Bubbly Flow (1st Transition Boundary)
As discussed above, the dispersed bubble flow is considered as no-slip flow. On the contrary,
bubbly flow, featured by phase slippage, corresponds to the uneven dispersion of gas bubbles in
liquids. For a rotating ESP, the initiation of pressure surging is closely related to the 1st transition
boundary [44,72]. Due to the lack of a valid mechanistic model to predict the initiation of pressure
surging inside ESPs, the empirical correlations have been developed from multi-variable regression
of experimental data. Table 9 below summarizes several existing pertinent studies.
Figure 17 shows the comparison of surging initiation models in ESPs. The same flow conditions
with rotational speed N = 3500 rpm and separator pressure at Psep = 1034 kpa are applied to all
calculations. As can be seen, the predictions of ESP surging initiation from existing models vary
significantly. Turpin et al. [78] and Cirilo [33] suggested that pressure surging should be independent
on liquid flow rates (QL). Other studies [37,38,72] claimed that the critical gas volumetric fraction (λC)
at which ESP pressure surging initiates is in a close relationship with QL. Duran [37] and Zapata [38]
correlations predict λC in monotonic-increase trend with respect to QL. Gamboa and Prado correlated
λC and QL with a concave quadratic function, and claimed that this model was validated for
⁄ > 0.2, beyond which the prediction error might occur.

Table 9. Correlations of surging initiation (1st transition boundary) in literature.

Study Correlation Application Range

 2000  QG
Turpin et al. [78]     —empirical constant

 3Pi  QL
Pi < 2.8 mPa

Pi < 3.4 mPa

Cirilo [33] C  0.0187 Pi 0.4342 QL < 0.02 m3/s
Estevam [64] C  31.92  32.151  G 
Energies 2018, 11, 180 24 of 41

1.421
QG    Q  Pi < 2.4 mPa
Duran [37]   5.58 G  0.098 L  QG < 0.02 m3/s
Qmax  L  Qmax  QL < 0.013 m3/s

QL
0.027 Pi < 1.4 mPa
QG Qmax
Zapata [38]  QG < 0.02 m3/s
Qmax Q
0.9001  L QL < 0.016 m3/s
Qmax
0.2 0.4 4.4682
QG   G   D 2    Q 
Gamboa and
     0.102 exp L  
Qmax   L   Q  Pi < 1.7 mPa
Prado [72] 
     max  
1
 0.4 2
2 
  L   G  R I 
2

Zhu et al. [44] C  3 2 1


   5  PQ L  5  L 5
10 .056      
  L    LV   G 

In Figure 17, the red-asteroid curve is obtained based on the mechanistic model predictions with
a dome-shape with the maximum λC at QL close to BEP (grey dashed line). At low QL, turbulent kinetic
energy in ESP impeller is low. It is easy for bubbles to coalesce and form large gas pocket. The flow
pattern transition and pressure surging will occur earlier at smaller λC. At high QL, the hydraulic head
of ESP is close to zero due to open flow conditions. Thus, the turbulent kinetic energy of fluids is also
low according to Padron [121] postulation = ∆ ⁄( ), where k is a constant value obtained
from experiments. Then, bubbles are more likely to collide and coalesce to generate bigger ones,
leading to further decrease of pump boosting pressure if the gas entrainment rate increases. Thus, the
dome-shape of λC versus QL is reasonable.

Figure 17. Comparison of surging initiation models for ESP gas/liquid flow [122].

Transition from Bubbly Flow to Intermittent Flow (2nd Transition Boundary)

For this flow pattern transition, it associates with the severe collision, breakup and coalescence
of free bubbles, which triggers the formation of large gas pocket and unstable flow characteristics.
Unfortunately, quite few modeling studies are available in literature on this topic. Similar to slug
flow in pipes, the intermittent flow inside a rotating ESP involves transient flow coupled with
Energies 2018, 11, 180 25 of 41

complex pump geometry, which brings in great difficulties to the quantitative analysis. Zhu [1]
claimed a critical in-situ gas void fraction (αCrit) based on the maximum packing theory in a 3D cubic
that initiates the transition of bubbly flow to intermittent flow. In multiphase pipe flow, αCrit = 0.25 is
mostly used as the critical value. To account for the rotation effect in ESP, a new value was proposed
by Zhu [1] as below:

  1   N 
n

 Crit      exp      (25)
6  6 4   N REF  
 
where NREF is the rotational speed at best efficiency point, which is 3500 rpm for most ESPs. If N = 0
rpm (no-rotation), Equation (25) is reduced to αCrit = 0.25, the same value as used in prediction of the
transition boundary from bubbly flow to slug flow in two-phase pipe flow. If N = +∞, the equation
reduces to αCrit = 0.52, corresponding the maximum packing of bubbles. The index n (n > 0, usually
bounded by 4.0) is an empirical constant determined by experimental data [1].
Transition from Bubbly Flow to Intermittent Flow (3rd Transition Boundary)
When the transition from slug flow to annular flow occurs, the momentum exchange term can
be neglected. Given the superficial gas velocity vSG, and making a guess for the superficial liquid
velocity vSL, the critical liquid holdup of the film can be obtained by iterations.
The combined momentum balance equation (Equation (26)) can be reduced to:

 F SF  1 1 
   I S I      L  C 2 RI 0 (26)
H LF A H
 LF A 1  H LF  A 
streamline

With the calculated HLF (liquid holdup in film), vF (film velocity), HLC (liquid holdup in gas core),
and vC (gas core velocity), Equation (26) can be solved to get a new vF. Several iterations are required
to reach convergence. Finally, the superficial gas/liquid velocity can be calculated.
Figure 18 shows a typical calculation result for predicting the transition boundary, where the
regimes I, II, III and IV correspond to the dispersed bubble flow, bubbly flow, intermittent flow and
segregated flow, respectively. The legend labels of DB, BF, SF, and AF denote dispersed bubble flow,
bubbly flow, slug flow and annual flow. The colored markers represent the flow patterns from
experimental results. As can be seen, the flow patterns predicted by mechanistic model are comparable
to that detected from experimental performance curves. Compared to Gamboa [40] visualization
experiments, similar flow pattern transition boundaries are well predicted. At low gas flow rate (QG),
the dispersed bubble flow encompassed by the blue line or bubbly flow bounded by the red line
prevails. With the increase of gas flow rate or decrease of liquid flow rate (QL), slug flow takes place
with the boundary of the black dot-dash line. The segregated flow (regime IV) corresponds to
extremely low QL.
Energies 2018, 11, 180 26 of 41

Figure 18. Validation of mechanistic model for predicting the flow pattern transition boundary in ESP
under gas/liquid flow [1].

4.4.2. Flow Model

With the analytical understanding of flow patterns prevailing inside a rotating ESP and the
transition boundaries, the assumptions that simplify the governing equation (Equation (26)) can be
made. Then the flow characteristics of each flow pattern can be solved by iterations. However, due
to the limited studies on ESP flow models with gas/liquid flow in literature, the material is insufficient
to conduct the corresponding review. Thus, this part is not discussed in detail.

4.4.3. Mechanistic Model Calculation

For the mechanistic models of two-phase flow in a rotating ESP discussed above, a calculation
flow chart is given below in Figure 19.
From our previous studies [1,90,122], the inputs of the mechanistic model contain pump
geometrical parameters such as radius, blade angles, channel volume and cross sectional area etc., as
well as the fluid properties, e.g., gas/liquid density, viscosities, surface tension. The ESP operation
conditions (rotational speed, flow rates etc.) are also required. The flow pattern transition boundaries
are similar to two-phase pipe flow. The flow map can be divided into four different flow regimes,
namely dispersed bubble flow, bubbly flow, intermittent flow, and segregated flow. Based on the
limited visualization results, the intermittent and segregated flows in a rotating ESP are slug flow
and co-current annular flow, respectively.
With the flow pattern determined, the respective flow model can be called to solve the governing
equations to obtain the flow structure and other hydraulic parameters, among which the in-situ gas
void fraction (αG) inside the rotating impeller is the most important parameter to estimate ESP
boosting pressure under gassy flow conditions.
Energies 2018, 11, 180 27 of 41

Figure 19. Flow chart for gas-liquid flow calculation inside a rotating ESP impeller.

5. Closure Relationship
In mechanistic or numerical models, the closure relationships are needed on top of the
conservation equations. The closure relationships in modeling centrifugal pump two-phase performance
include bubble size, drag force coefficient, in-situ gas void fractions, friction factors etc.

5.1. Bubble Size

The bubble size prediction is a critical closure relationship in mechanistic modeling of ESP
performance under gassy conditions. However, a generally validated mechanistic model for
predicting bubble size in centrifugal pumps is not available. Several proposed bubble size models for
centrifugal pump are either empirical or semi-empirical. They were verified with specific pumps and
air/water as working fluids [31,32,39,40]. The generality of these models is questionable. The bubble
characteristics inside the rotating impeller are affected by many factors, including fluid properties
(density, viscosity, surface tension), pump geometry and operating parameters (rotational speed,
flow rates).
By photographing the bubble dispersion in a pump with a transparent Plexiglas casing,
Murakami and Minemura [31,32] correlated the observed bubble sizes with a linear relationship of
Sauter Mean Diameter (d32) versus GVF (Equation (29)):
3

 N  4
d m  21.82  0.618 4.273  (27)
 6.862 
Although Equation (27) is based on rotational speed (N) and inlet GVF, the empirical nature
limited its applicability (λ < 8%, 1000 < N < 1500 rpm). Another model of bubble size in a centrifugal
pump was proposed by Estevam [64] based on analogy to gas-liquid two-phase pipe flow. Applying
Hinze [123] theory for droplet breakup mechanism in turbulent flow to bubble size prediction in
centrifugal pump, Estevam obtained the following equation to calculate the maximum dispersed
bubble size:
3 2

  5  2 f  5
d max  1.17    ,  (28)
 l   DH 
Energies 2018, 11, 180 28 of 41

where the constant of 1.17 is a parameter accounting for the curvature of impeller flow passages; σ,
ρl and DH are interfacial tension, liquid density and hydraulic diameter, respectively; the friction
factor (fβ,ω) is determined by analogy to fluid flow in pipeline and considering rotational speed effect.
Following a similar methodology of modeling bubble size inside a centrifugal pump, Barrios [39]
proposed a bubble size prediction model based on the visualization experimental data inside ESP
taken by high-speed camera. Barrios pointed out that the experimental results were necessary to
determine the relationship of the maximum bubble size, the critical bubble size and inlet GVF. The
critical Weber number (Wecrit) is a parameter dominating gas bubble breakup. Hence, Equation (30)
explicitly relates the bubble size with the rotational speed and liquid properties:
3/ 5
  1
db _ surg  0.0348N   
0.8809 1 / 4
(29)
 l  N r 
3 2 2/5
1

where r1 is the impeller radius. Gamboa [40] employed the Levich [124] model for maximum stable
bubble size in pipe flow and proposed an alternative way of modeling bubble size inside an ESP.
Gamboa improved the bubble size model by introducing dispersed gas phase density and Wecrit based
on Kouba [125] droplet breakup studies:
3/ 5 2 / 5 1/ 5
   D 4   l 
d max  14.27  
  
  
 
1  191.7  0.2
(30)
 l     g
where ρg is in-situ gas density, v is liquid kinetic viscosity and Ω, D are impeller rotational speed and
diameter, respectively.
A new correlation for computing the representative bubble sizes inside or rotating ESP impeller
was recently proposed by Zhu and Zhang [105], as given by Equation (31). The model is based on
CFD simulations of pump pressure increment under gassy flows, where the bubble size plays a
dominant role to make the numerically simulated ESP heads comparable to the corresponding
experimental data. The proposed correlation was verified by further comparisons of the CFD
simulation results with the incorporation of Equation (33) against the experimental pump pressure
increment, which shows a good match:
35 2 5 15
   Pq   c 
d 32  6.034 G       (31)
 c    cV   d 
where d32 is the Sauter mean diameter; λG is the inlet gas volumetric fraction; subscripts of c and d
denote continuous and dispersed phases, respectively. ΔP is the pressure increment of ESP under
single-phase flow; V is the volume of the entire impeller flow domain.
The comparison of existing bubble size models in literature is shown in Figure 20, where the
horizontal axis corresponds to λG and the vertical axis shows the calculated bubble diameters. As can
be seen, the predicted bubble sizes in the rotating ESP impeller by different models vary significantly.
The reason might be due to the derivation of prediction model that is based on either empirical
constant or partial regression of experimental data. The complex physical reality of multiphase flow
through a rotating ESP impeller further contributes to the difficulty of reaching a generally validated
prediction model of bubble size. Meanwhile, the mutual validation among different models seems to
be unavailable due to the lack of relevant experimental measurement of in-situ bubble sizes as well
as geometrical details (channel volume etc.). A similar trend of db versus λG is used by Barrios [39]
and Gamboa [40] to correlate the experimental measurement of bubble diameters. However, the
linear relationship of db versus λG is applied by Murakami and Minemura [31,32], and Zhu and Zhang
[105] studies, where a good agreement between the model predicted db against either visualization
experiments or CFD simulation results can be reached.
Energies 2018, 11, 180 29 of 41

Figure 20. Comparison of bubble size prediction models.

Although several models are available in literature for bubble size prediction in a centrifugal
pump with rotating turbulent flow, their empirical/semi-empirical natures limited the range of
applications. As discussed by Gamboa [40], the challenge in modeling bubble size inside a rotating
ESP is the mechanism that dominates bubble formation, including coalescence and breakup. Therefore,
investigation is needed to better understand the bubble dispersion mechanism and develop better
model for bubble size prediction.

5.2. Drag Coefficient

In gas-liquid two-phase flow, the drag force is the interfacial momentum transfer due to velocity
difference between gas and liquid phases [102]:

 f U  v U  v
FD  C D Ap (32)
2
where U and v are velocities of liquid and gas phases, respectively. The drag coefficient (CD) for
bubbles in no-rotating flow fields without shearing was studied by Schiller and Naumann [126],
Clift et al. [127], Ishii and Zuber [128], Mei et al. [129] among others. Schiller and Naumann proposed
an empirical correlation of CD for 0.1 < Re < 800. Clift et al. developed a more accurate expression
which is valid for higher Re up to 3 × 105 [130]:

24 0.42
CD 
Re
1  0.15 Re0.687  
1  4.25 104 Re1.16
(33)

Ishii and Zuber incorporated a correlation term (1 + αd)−γ to account for effects of bubble
volumetric fraction and flow regime, where γ is an empirical constant determined by fluid properties
and particle shapes. Mei et al. [129] studied the behavior of clean bubbles (no contaminations or
surfactants involved) in a uniform flow and proposed an empirical drag CD for 0.1 < Re < 100:

16   8 1  
1
 
1
CD  1  1  3.315 Re 2    (34)
Re   Re 2  
 
 
For shear-induced flow, the viscous drag force exerted on bubbles is increased by broadening
the near wake [131]. Thus, the dimensionless shear rate, Strouhal number (Sr) as an indicator of shear
strength, was employed by Legendre and Magnaudet [132] to calculate CD in shear flow for moderate-
to-large Re (≥50):
Energies 2018, 11, 180 30 of 41

d b
Sr  (35)
U v
and:

 0.55 
CD ,sr  CD ,0 1  2  (36)
 Sr 
where CD,0 is the drag coefficient calculated without shear effect [131]. Rastello et al. [133,134] revised
Equation (36) to get a better fitting of their experimental data for low, moderate and high Reynolds
numbers with a broad range of Sr:

 0.3 
CD ,sr  CD ,0 1  2.5  (37)
 Sr 
Barrios [39] measured bubble sizes inside a single-stage ESP with a visualization experimental
system and calculated the drag coefficients on stagnant bubbles in rotating flow field. Then, the drag
coefficients were correlated by:

24
CD  1  f Re,Y  (38)
Re Y
where Y and f (Re,Y) are given by:

5.48
f Re,Y   ReY 0.427  0.36 ReY  (39)
24 24
and:

Re
Y  0.00983 389.9 (40)
N2
Figure 21 shows the comparison of drag coefficients predicted by the empirical models discussed
above. As can be seen, the evident discrepancy is observed under relatively low Reynolds number
with/without considering the rotating effects in flow fields. At higher Reynolds number, the similar
drag coefficients, which are close to 0.44, can be obtained from the existing empirical models. The
drag coefficients predicted by Clift et al. [127] and Mei et al. [128] is close to each other due to
neglecting rotation flow. However, when accounting for fluid rotating effect, the model predictions
of CD vary significantly. Barrios [39] correlation outputs CD that is several times larger than that
calculated by Legendre and Magnaudet [132] model. A possible explanation is that Barrios model is
specifically on basis of the bubble flow inside a rotating ESP impeller, while the Legendre and
Magnaudet model is regressed from a single bubble flowing inside a rotating cylinder.
Energies 2018, 11, 180 31 of 41

Figure 21. Comparison of drag coefficient models.

5.3. In-Situ Gas Void Fraction (αG)

The in-situ gas void fraction (αG) is an important variable in two-phase flow related to the
velocity slippage between two immiscible phases and the gas accumulation. Moreover, αG is needed
to determine the in-situ gas/liquid mixture density by = + (1 − ) , which in turn affects
the pump boosting pressure significantly according to ∆ = . ρM is the mixture density, and
HM is the hydraulic head of ESP with gas/liquid flow. However, due to the complicated pump
geometries and fluid flow dynamics in the ESP impeller, it is challenging to develop a mechanistic
model to predict αG with general validity. Very few studies on mechanistic modeling of local gas void
fraction in ESPs can be found in literature so far.
The simplest model for predicting αG in multiphase centrifugal pump flows is the homogeneous
model, which assumes no slippage between gas and liquid phases:
QG
 G  G  (41)
QG  QL
The homogenous model is valid for very low inlet GVFs when the slippage between gas and
liquid is minimal. Errors will result from applying the homogeneous model to flow conditions with
high inlet GVFs when the slippage is not negligible. Accounting for the phase slippage, several
empirical correlations were proposed by Chisely [62], Estevam [64], Zapata [38] and Pineda et al. [114],
among others.
Chisely studied loss of coolant accident with a volute-type centrifugal pump in the nuclear
industry. The pressure distribution and flow regimes were determined with the experimental
measurement and high-speed photographing. A model for predicting the centrifugal pump pressure
increment under two-phase flow conditions was proposed. To make this model solvable, the in-situ
αG was correlated as:
1
  1  x    G   L  
0.64 0.36 0.07

G  1  0.28       (42)
  x    L   G  

where χ is the gas quality, μG and μL are the gas and liquid viscosities.
Zapata experimentally studied the rotational speed effect on ESP two-phase performance. Using
least-square regression, a new correlation of αG was presented to predict the pump boosting pressure.
The empirical correlation as function of gas and liquid flow rates and the rotational speed is given as:
Energies 2018, 11, 180 32 of 41

1
 q  qL qmax  1.277  0.034 N d N d  0.921 0.068 N d N d
 G   G  Nd  (43)
 qmax  0.598  0.223 N d 
where qG and qL are the gas and liquid flow rates, is normalized rotational speed, and qmax is the
maximum single-phase liquid flow rate.
Pineda et al. [114] proposed an empirical correlation based on the Lockhart-Martinelli type
parameter Xtt by non-linear regression of CFD simulated values of in-situ αG:

 G  7 .119 X tt0.8778  0 .002138 (44)

where:
0.1 0.9 0.5
   1 x   G 
X tt   L      (45)
 G   x   L 
Zhu and Zhang [90] proposed a mechanistic model to compute the in-situ gas void fraction. The
model prediction was validated by the 3D CFD simulation. A better match over the existing empirical
models was achieved. The model is based on the balance of centrifugal buoyancy force and drag force
on a stable gas bubble in ESP, which incorporates the gas-liquid interaction mechanism and is
applicable to a wide flow conditions. The prediction model is given as,

RS  1  1  RS 2  4 RS G
G  (46)
2 RS
where:

VSR2RI  ZI TB YI
RS  (47)
Q QLK
VSR is the slippage velocity in radial direction, RI is the representative impeller radius, ZI is the
impeller number, TB is the blade thickness, YI is the impeller channel height, Q and QLK is the liquid
flow rate and its corresponding leakage flow rate. The drag coefficient is calculated by

C 1  0.55Sr2  Re  50
CD   D,0
CD,0 1  0.3Sr  Re  50
2.5 (48)

where Sr is Strouhal number and defined by = Ω⁄| − |. U and v are phase velocities. CD,0 is
the drag coefficient proposed by Clift et al. [127] without consideration of shear effect.
Figure 22 shows the comparison of existing models in predicting in-situ gas void fraction in a
rotating pump impeller. As can be seen, Zhu and Zhang [90] model predicts in-situ αG with an error
less than 25%, and most predictions are bounded by 10% error line. Nevertheless, αG predicted by
previous empirical correlations deviates from CFD simulated αG increasingly as inlet λG increases.
Since most of the correlations were based on homogeneous assumption rather than velocity slippage
between gas and liquid phases, the predictions from empirical correlations are acceptable only at low
λG. For relatively higher λG, the phase slippage dominates bubble behaviors, which results in the
failure of empirical correlations.
Energies 2018, 11, 180 33 of 41

+25%

-10%

(a)

+25%

-10%

(b)
Figure 22. Comparison of model predictions of in-situ gas void fraction against CFD simulation
results [90], (a) N = 3500 rpm; (b) N = 1800 rpm.

6. Conclusions and Future Work

In this paper, the experimental studies and modeling approaches for investigating gas/liquid
two-phase flow inside a rotating electrical submersible pump have been reviewed in detail. Different
empirical and mechanistic models for predicting ESP boosting pressure under gassy flows were
discussed. The state-of-art CFD-based numerical simulation as a complemental method for ESP two-
phase performance prediction was elaborated.
The review of the published experimental studies for gas-liquid flow in centrifugal pumps has
highlighted the complications behind the accurate prediction of two-phase performance inside a
rotating ESP. The turbulent flow with transient behaviors make it difficult to visualize the internal
flow structures and flow patterns. This is not helped by the twisted blade/vane geometries and
multistage assembly. Due to these observation difficulties, the multiphase flow mechanisms that
dominate the ESP head degradation were not well understood. This in turn delayed the development
of mechanistic models. Although CFD simulation serves as an alternative way to study multiphase
flow, its validity and reliability, especially coupled with the strong shearing effect and phase
interactions are questioned.
The ESP performance under gassy flow is affected by many factors. There is room for further
improvements on the mechanistic modeling and prediction:
Energies 2018, 11, 180 34 of 41

 Experimental visualization of internal two-phase flow structures in a rotating ESP is inadequate,

which restricts the flow pattern assessment and the formulations of governing equations.
Moreover, the measurement of in-situ gas phase is insufficient, resulting in further difficulty to
characterize pump performance quantitatively. Future experimental work should take these
challenges into consideration, thus providing in-depth knowledge of multiphase flow
mechanisms in ESPs.
 The computational fluid dynamics techniques need to be further developed in order to capture
the complex behaviors and movement of gas bubbles inside ESPs and thus better understand
the interactions between the gas and liquid phases in different flow patterns.
 The mechanistic models should focus more on the physics of the two-phase flow in rotating ESPs
to propose better closure relationships. From the one-dimensional two-fluid model, the
simplified mass and momentum conservation equations are solvable with the help of the
improved closure relationships in future, ending up with a more accurate and reliable
mechanistic model to predict ESP two-phase flow performance.

Acknowledgments: The authors appreciate the technical and financial support of the Tulsa University Artificial
Lift Projects (TUALP) member companies. The work is supported by the National Natural Science Foundation-
Outstanding Youth Foundation (51622405).

Conflicts of Interest: The authors declare no conflict of interest.

Nomenclature
A area, L2, m2
a channel length, L, m
b blade thickness, L, m; or channel width, L, m
BEP best efficiency point
BHP brake horsepower, ML2/T3, kg∙m2/s3
C absolute velocity, L/T, m/s
CD drag coefficient
Cm disk friction coefficient
Cp diffuser loss coefficient
d bubble diameter, L, m
D1t diameter at tip of inlet, L, m
Df diffusion factor
DH hydraulic diameter, L, m
Di impeller diameter, L, m
f friction factor
fdisk disk friction loss coefficient
fE liquid entrainment factor
shape factor dependent on the impeller diameters and the angle between the sidewall of the
fgeo
rotor and the pump shaft
interfacial force vector, M/(LT2), Pa
gravity acceleration vector, L/(T2), m/s2
GVF gas volumetric fraction
h channel height, L, m or hydraulic head, L, m
H hydraulic head, L, m or holdup
k turbulent kinetic energy, L2/(T2), m2/s2
kRR friction factor
mass flow rate, M/T, kg/s
M momentum transfer term per unit volume, M/(L2T2), Pa/m
n phase number
N rotational speed, 1/L, rpm
p pressure, M/(LT2), Pa
ΔP stage pressure increment, M/(LT2), Pa
P pressure, M/(LT2), Pa
q flow rate, L3/T, m3/s
Q mass flow rate, M/T, kg/s
Energies 2018, 11, 180 35 of 41

r radius, L, m
rH hydraulic radius, L, m
Re Reynolds number
s streamline, L, m
Sr Strouhal number
t time, T, s
T torque, (ML2)/T2, kg∙m2/s2
phase velocity vector, L/T, m/s
U peripheral velocity, L/T, m/s
v velocity, L/T, m/s
v’ velocity fluctuation, L/T, m/s
V velocity, L/T, m/s
Vol volume, L3, m3
W relative velocity in ESP, L/T, m/s
We Weber number
x mass fraction or mole fraction
Y channel height, L, m
Z blade number
Greek Symbols
gas void fraction, or absolute flow angle formed between the absolute
α
velocity and its tangential component
β tangential blade angle, deg
λG no-slip gas void fraction (GVF)
μ dynamic viscosity, M/(LT), Pa∙s
μt turbulent viscosity, M/(LT), Pa∙s
Ω angular speed, 1/T, rad/s
ω shaft or impeller blades angular velocity, 1/T, rad/s
ρ fluid density, M/L3, kg/m3
σ surface tension, M/T2, N/m or slip factor
stress-strain tensor, M/(LT2), Pa
ε turbulent energy dissipation rate per unit mass, L2/T3, m2/s3
θ component of the radial coordinate system
Subscripts
1 inlet
2 outlet
32 Sauter mean diameter
B bubble or blade
b_surg bubble at pressure surging
c continuous phase
C gas core or critical
CRIT critical
d diffuser or dispersed phase or dimensionless parameter
D diffuser
eff effective
E Euler
F film
FI fluid in ESP impeller
FD fluid in ESP diffuser
G gas phase
H hydraulic parameter
I interface or impeller
L liquid phase
LF liquid film
M meridional direction or mixture
m mixture
R radius direction
sphere sphere
streamline projection on streamline
Energies 2018, 11, 180 36 of 41

S specific speed or shear or slug body

SG superficial gas
SI impeller channel wall
SL superficial liquid
SR shear in the radial direction
U peripheral direction
vm virtual mass
w water
W wall
Units
bpd barrel per day, 1 bpd ≈ 1.8942 × 10−6 m3/s
psi pound per square inch, 1 psi ≈ 6894.76 pa
psia pound per square inch for absolute pressure
psig pound per square inch for gauge pressure

References
1. Zhu, J. Experiments, CFD Simulation and Modeling of ESP Performance under Gassy Conditions.
Ph.D. Thesis, The University of Tulsa, Tulsa, OK, USA, 2017.
2. Barrios, L.; Rojas, M.; Monteiro, G.; Sleight, N. Brazil field experience of ESP performance with viscous
emulsions and high gas using multi-vane pump MVP and high power ESPs. In Proceedings of the SPE
Electric Submersible Pump Symposium, The Woodlands, TX, USA, 24–28 April 2017.
3. Takacs, G. Electrical Submersible Pumps Manual: Design, Operations, and Maintenance; Gulf Professional Publishing:
Burlington, NY, USA, 2009.
4. Oliva, G.B.; Galvão, H.L.; dos Santos, D.P.; Silva, R.E.; Maitelli, A.L.; Costa, R.O.; Maitelli, C.W. Gas effect
in electrical-submersible-pump-system stage-by-stage analysis. SPE Prod. Oper. 2017, 32, 294–304.
5. Zhou, D.; Sachdeva, R. Simple model of electric submersible pump in gassy well. J. Pet. Sci. Eng. 2010, 70,
204–213.
6. Stepanoff, A.J. Centrifugal and Axial Flow Pumps: Theory, Design and Application, 2nd ed.; John Wiley & Sons:
New York, NY, USA, 1957.
7. 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.
8. Ito, H. Friction factors for turbulent flow in curved pipes. J. Basic Eng. 1959, 81, 123–134.
9. Jones, O.C. An improvement in the calculation of turbulent friction in rectangular ducts. J. Fluids Eng. 1976,
98, 173–180.
10. Churchill, S.W. Friction-factor equation spans all fluid-flow regimes. Chem. Eng. 1977, 84, 91–92.
11. Shah, R.K. A correlation for laminar hydrodynamic entry length solutions for circular and noncircular
ducts. J. Fluids Eng. 1978, 100, 177–179.
12. Sun, D. Modeling Gas-Liquid Head Performance of Electrical Submersible Pumps. Ph.D. Thesis, The University
of Tulsa, Tulsa, OK, USA, 2003.
13. Wiesner, F.J. A review of slip factors for centrifugal impellers. J. Eng. Power 1967, 89, 558–566.
14. Sun, D.; Prado, M.G. Single-phase model for electric submersible pump (ESP) head performance. SPE J.
2006, 11, 80–88.
15. Thin, K.C.; Khaing, M.M.; Aye, K.M. Design and performance analysis of centrifugal pump. World Acad.
Sci. Eng. Technol. 2008, 46, 422–429.
16. Ito, H.; Nanbu, K. Flow in rotating straight pipes of circular cross section. J. Basic Eng. 1971, 93, 383–394.
17. Bing, H.; Tan, L.; Cao, S.; Lu, L. Prediction method of impeller performance and analysis of loss mechanism
for mixed-flow pump. Sci. China Technol. Sci. 2012, 55, 1988–1998.
18. Amaral, G. Single-Phase Flow Modeling of an ESP Operating with Viscous Fluids. Master’s Thesis,
State University of Campinas, Campinas, São Paulo, Brazil, 2007.
19. Gülich, J.F. Pumping highly viscous fluids with centrifugal pumps—Part 1. World Pumps 1999, 395, 30–34.
20. Gülich, J.F. Pumping highly viscous fluids with centrifugal pumps—Part 2. World Pumps 1999, 396, 39–42.
21. Tuzson, J. Centrifugal Pump Design; John Wiley & Sons: New York, NY, USA, 2000.
22. Sun, D.; Prado, M.G. Modeling gas-liquid head performance of electric submersible pumps. J. Press. Vessel
Technol. 2005, 127, 31–38.
Energies 2018, 11, 180 37 of 41

23. Van Esch, B.P.M. Simulation of Three-Dimensional Unsteady Flow in Hydraulic Pumps. Master’s Thesis,
University of Twente, Enschede, The Netherlands, 1997.
24. Ladouani, A.; Nemdili, A. Influence of Reynolds number on net positive suction head of centrifugal pumps
in relation to disc friction losses. Forsch. Ing. 2009, 73, 173.
25. Ippen, A.T. The Influence of Viscosity on Centrifugal Pump Performance; Issue 199 of Fritz Engineering
Laboratory Report, ASME Paper No. A-45-57; The American Society of Mechanical Engineers: New York,
NY, USA, 1945.
26. Li, W.G. Experimental investigation of performance of commercial centrifugal oil pump. World Pumps 2002,
425, 26–28.
27. Amaral, G.; Estevam, V.; França, F.A. On the influence of viscosity on ESP performance. SPE Prod. Oper.
2009, 24, 303–310.
28. Solano, E.A. Viscous Effects on the Performance of Electro Submersible Pumps (ESP’s). Master’s Thesis,
The University of Tulsa, Tulsa, OK, USA, 2009.
29. Barrios, L.J.; Scott, S.L.; Sheth, K.K. ESP technology maturation: Subsea boosting system with high GOR
and viscous fluid. In Proceedings of the SPE Annual Technical Conference and Exhibition, San Antonio,
TX, USA, 8–10 October 2012.
30. Banjar, H.M.; Gamboa, J.; Zhang, H.-Q. Experimental study of liquid viscosity effect on two-phase stage
performance of electrical submersible pumps. In Proceedings of the SPE Annual Technical Conference and
Exhibition, New Orleans, LA, USA, 30 September–2 October 2013.
31. Murakami, M.; Minemura, K. Effects of entrained air on the performance of a centrifugal pump (1st report,
performance and flow conditions). Bull. JSME 1974, 17, 1047–1055.
32. Murakami, M.; Minemura, K. Effects of entrained air on the performance of a centrifugal pump (2nd report,
effects of number of blades). Bull. JSME 1974, 17, 1286–1295.
33. Cirilo, R. Air-Water Flow through Electric Submersible Pumps. Master’s Thesis, The University of Tulsa,
Tulsa, OK, USA, 1998.
34. Romero, M. An Evaluation of an Electric Submersible Pumping System for High GOR Wells. Master’s Thesis,
University of Tulsa, Tulsa, OK, USA, 1999.
35. Pessoa, R. Experimental Investigation of Two-Phase Flow Performance of Electrical Submersible Pump
Stages. Master’s Thesis, The University of Tulsa, Tulsa, OK, USA, 2001.
36. Beltur, R. Experimental Investigation of Two-Phase Flow Performance of ESP Stages. Master’s Thesis,
The University of Tulsa, Tulsa, OK, USA, 2003.
37. Duran, J. Pressure Effects on ESP Stages’ Air-Water Performance. Master’s Thesis, The University of Tulsa,
Tulsa, OK, USA, 2003.
38. Zapata, L. Rotational Speed Effects on ESP Two-Phase Performance. Master’s Thesis, The University of Tulsa,
Tulsa, OK, USA, 2003.
39. Barrios, L. Visualization and Modeling of Multiphase Performance inside an Electrical Submersible Pump.
Ph.D. Thesis, The University of Tulsa, Tulsa, OK, USA, 2007.
40. Gamboa, J. Prediction of the Transition in Two-Phase Performance of an Electrical Submersible Pump.
Ph.D. Thesis, The University of Tulsa, Tulsa, OK, USA, 2008.
41. Trevisan, F.E. Modeling and Visualization of Air and Viscous Liquid in Electrical Submersible Pump.
Ph.D. Thesis, The University of Tulsa, Tulsa, OK, USA, 2009.
42. Salehi, E. ESP Performance in Two-Phase Flow through Mapping and Surging Tests at Various Rotational
Speeds and Intake Pressures. Master’s Thesis, The University of Tulsa, Tulsa, OK, USA, 2012.
43. Croce Mundo, D.D. Study of Water/Oil Flow and Emulsion Formation in Electrical Submersible Pumps.
Master’s Thesis, The University of Tulsa, Tulsa, OK, USA, 2014.
44. Zhu, J.; Zhu, H.; Zhang, J.; Zhang, H.-Q. An experimental study of surfactant effect on gas tolerance in
electrical submersible pump (ESP). In Proceedings of the ASME 2017 International Mechanical Engineering
Congress and Exposition, Tampa, FL, USA, 3–9 November 2017.
45. DeGennes, P.-G.; Brochard-Wyart, F.; Quere, D. Capillarity and Wetting Phenomena; Springer: New York,
NY, USA, 2004.
46. Hu, B.; Nienow, A.W.; Stitt, E.H.; Pacek, A.W. Bubble sizes in agitated solvent/reactant mixtures used in
heterogeneous catalytic hydrogenation of 2-butyne-1,4-diol. Chem. Eng. Sci. 2006, 61, 6765–6774.
47. Omer, A.; Pal, R. Effects of surfactant and water concentrations on pipeline flow of emulsions. Ind. Eng.
Chem. Res. 2013, 52, 9099–9105.
Energies 2018, 11, 180 38 of 41

48. Van Nimwegen, A.T.; Portela, L.M.; Henkes, R.A. The effect of surfactants on vertical air/water flow for
prevention of liquid loading. SPE J. 2015, 4, 488–500.
49. Van Nimwegen, A.T.; Portela, L.M.; Henkes, R.A.W.M. The effect of surfactants on upward air-water pipe
flow at various inclinations. Int. J. Multiph. Flow 2016, 78, 132–147.
50. Ajani, A.; Kelkar, M.; Sarica, C.; Pereyra, E. Effect of surfactants on liquid loading in vertical wells. Int. J.
Multiph. Flow. 2016, 83, 183–201.
51. Ajani, A.; Kelkar, M.; Sarica, C.; Pereyra, E. Foam flow in vertical gas wells under liquid loading: Critical
velocity and pressure drop prediction. Int. J. Multiph. Flow 2016, 87, 124–135.
52. Shoham, O. Mechanistic Modeling of Gas-Liquid Two-Phase Flow in Pipes; Society of Petroleum Engineers:
Richardson, TX, USA, 2006.
53. Parsi, M.; Najmi, K.; Najafifard, F.; Hassani, S.; McLaury, B.S.; Shirazi, S.A. A comprehensive review of
solid particle erosion modeling for oil and gas wells and pipelines applications. J. Nat. Gas Sci. Eng. 2014,
21, 850–873.
54. Wu, B.; Firouzi, M.; Mitchell, T.; Rufford, T.E.; Leonardi, C.; Towler, B. A critical review of flow maps for
gas-liquid flows in vertical pipes and annuli. Chem. Eng. J. 2017, 326, 350–377.
55. Taitel, Y.; Bornea, D.; Dukler, A.E. Modelling flow pattern transitions for steady upward gas-liquid flow in
vertical tubes. AIChE J. 1980, 26, 345–354.
56. Minumura, K.; Murakami, M. A theoretical study on air bubble motion in a centrifugal pump impeller. J.
Fluids Eng. 1980, 102, 446–455.
57. Patel, B.R. Investigations into the two-phase flow behavior of centrifugal pumps. In Polyphase Flow in
Turbomachinery, Proceedings of the ASME Winter Annual Meeting of the American Society of Mechanical
Engineering, San Francisco, CA, USA, 10–15 December 1978; The American Society of Mechanical Engineers:
New York, NY, USA; pp. 79–100.
58. Sekoguchi, K.; Takada, S.; Kanemori, Y. Study of air-water two-phase centrifugal pump by means of electric
resistivity probe technique for void fraction measurement: 1st report, measurement of void fraction
distribution in a radial flow impeller. Bull. JSME 1984, 27, 931–938.
59. Kim, J.H.; Duffey, R.B.; Belloni, P. On centrifugal pump head degradation in two-phase flow. In Design
Methods for Two-Phase Flow in Turbomachinery; The American Society of Mechanical Engineers: New York,
NY, USA, 1985.
60. Sato, S.; Furukawa, A.; Takamatsu, Y. Air-water two-phase flow performance of centrifugal pump
impellers with various blade angles. JSME Int. J. Ser. B Fluids Therm. Eng. 1996, 39, 223–229.
61. Takemura, T.; Kato, H.; Kanno, H.; Okamoto, H.; Aoki, M.; Goto, A.; Egashira, K.; Shoda, S. Development
of rotordynamic multiphase pump: The first report. In Proceedings of the International Conference on
Offshore Mechanics and Arctic Engineering, Vancouver, BC, Canada, 13–17 April 1997.
62. Chisely, E.A. Two Phase Flow Centrifugal Pump Performance. Ph.D. Thesis, Idaho State University,
Pocatello, ID, USA, 1997.
63. Suryawijaya, P.; Kosyna, G. Unsteady measurement of static pressure on the impeller blade surfaces and
optical observation on centrifugal pumps under varying liquid/gas two-phase flow condition. J. Comput.
Appl. Mech. 2001, 6759, D9–D18.
64. Estevam, V. A Phenomenological Analysis about Centrifugal Pump in Two Phase Flow Operation.
Ph.D. Thesis, Universidade Estadual de Campinas, Campinas, Brazil, 2002.
65. Poullikkas, A. Two phase flow performance of nuclear reactor cooling pumps. Prog. Nucl. Energy 2000, 36,
123–130.
66. Thum, D.; Hellmann, D.H.; Sauer, M. Influence of the patterns of liquid-gas flows on multiphase-pumping
of radial centrifugal pumps. In Proceedings of the 5th North American Conference on Multiphase Technology,
Banff, AB, Canada, 31 May–2 June 2006.
67. Barrios, L.; Prado, M.G. Experimental visualization of two-phase flow inside an electrical submersible
pump stage. J. Energy Resour. Technol. 2011, 133, doi:10.1115/1.4004966.
68. Schäfer, T.; Bieberle, A.; Neumann, M.; Hampel, U. Application of gamma-ray computed Tomography for
the analysis of gas holdup distributions in centrifugal pumps. Flow Meas. Instrum. 2015, 46, 262–267.
69. Neumann, M.; Schäfer, T.; Bieberle, A.; Hampel, U. An experimental study on the gas entrainment in horizontally
and vertically installed centrifugal pumps. J. Fluids Eng. 2016, 138, 091301.
70. Verde, W.M.; Biazussi, J.L.; Sassim, N.A.; Bannwart, A.C. Experimental study of gas-liquid two-phase flow
patterns within centrifugal pumps impellers. Exp. Therm. Fluid Sci. 2017, 85, 37–51.
Energies 2018, 11, 180 39 of 41

71. Estevam, V.; França, F.A.; Alhanati, F.J. Mapping the performance of centrifugal pumps under two-phase
conditions. In Proceedings of the 17th International Congress of Mechanical Engineering, Sao Paulo, Brazil,
10–14 November 2003.
72. Gamboa, J.; Prado, M. Review of electrical-submersible-pump surging correlation and models. SPE Prod. Oper.
2011, 26, 314–324.
73. Schäfer, T.; Neumann, M.; Bieberle, A.; Hampel, U. Experimental investigations on a common centrifugal
pump operating under gas entrainment conditions. Nucl. Eng. Des. 2017, 316, 1–8.
74. Bieberle, A.; Schlottke, J.; Spies, A.; Schultheiss, G.; Kühnel, W.; Hampel, U. Hydrodynamics analysis in
micro-channels of a viscous coupling using gamma-ray computed tomography. Flow Meas. Instrum. 2015,
45, 288–297.
75. Wagner, M.; Bieberle, A.; Bieberle, M.; Hampel, U. Dynamic bias error correction in gamma-ray computed
tomography. Flow Meas. Instrum. 2017, 53, 141–146.
76. Trevisan, F.; Prado, M. Experimental investigation of the viscous effect on two-phase-flow patterns and
hydraulic performance of electrical submersible pumps. J. Can. Pet. Technol. 2011, 50, 45–52.
77. Paternost, G.M.; Bannwart, A.C.; Estevam, V. Experimental study of a centrifugal pump handling viscous
fluid and two-phase flow. SPE Prod. Oper. 2015, 30, 146–155.
78. Turpin, J.L.; Lea, J.F.; Bearden, J.L. Gas-liquid through centrifugal pumps-correlation of Data. In
Proceedings of the Third International Pump Symposium, Houston, TX, USA, 20–22 May 1986.
79. Duran, J.; Prado, M.G. ESP stages air-water two-phase performance—Modeling and experimental data.
SPE 120628. Presented at the 2004 SPE ESP Workshop, Houston, TX, USA, 28–30 April 2004.
80. Zakem, S. Determination of gas accumulation and two-phase slip velocity ratio in a rotating impeller.
In Polyphase Flow and Transportation Technology, Proceedings of the ASME Century 2 Emerging Technology
Conference, San Francisco, CA, USA, 13–15 August 1980; The American Society of Mechanical Engineers:
New York, NY, USA; Volume 12, pp. 167–173.
81. Furuya, O. Analytical model for prediction of two-phase (noncondensable) flow pump performance.
J. Fluids Eng. 1985, 107, 139–147.
82. Sachdeva, R. Two-Phase Flow through Electric Submersible Pumps. Ph.D. Thesis, The University of Tulsa,
Tulsa, OK, USA, 1988.
83. Sachdeva, R.; Doty, D.R.; Schmidt, Z. Performance of electric submersible pumps in gassy wells. SPE Prod. Facil.
1994, 2, 55–60.
84. Minemura, K.; Uchiyama, T.; Shoda, S.; Kazuyuki, E. 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.
85. Gao, C.; Gray, K.E. The development of a coupled geomechanics and reservoir simulator using a staggered
grid finite difference approach. In Proceedings of the SPE Annual Technical Conference and Exhibition,
San Antonio, TX, USA, 7–12 October 2017.
86. Asuaje, M.; Bakir, F.; Kouidri, S.; Kenyery, F.; Rey, R. Numerical modelization of the flow in centrifugal
pump: Volute influence in velocity and pressure fields. Int. J. Rotating Mach. 2005, 3, 244–255.
87. Maitelli, C.; Bezerra, V.; Da Mata, W. Simulation of flow in a centrifugal pump of ESP systems using
computational fluid dynamics. Braz. J. Pet. Gas 2010, 4, 1–9.
88. Rajendran, S.; Purushothaman, K. Analysis of a centrifugal pump impeller using ANSYS-CFX. Int. J. Eng.
Res. Technol. 2012, 1, 1–6.
89. Cheah, K.W.; Lee, T.S.; Winoto, S.H.; Zhao, Z.M. Numerical flow simulation in a centrifugal pump at design
and off-design conditions. Int. J. Rotating Mach. 2007, doi:10.1155/2007/83641.
90. Zhu, J.; Zhang, H.-Q. Mechanistic modeling and numerical simulation of in-situ gas void fraction inside
ESP impeller. J. Nat. Gas Sci. Eng. 2016, 36, 144–154.
91. Gonzalez, J.; Fernandez, J.; Blanco, E.; Santolaria, C. Numerical simulation of the dynamic effects due to
impeller-volute interaction in a centrifugal pump. J. Fluids Eng. 2002, 124, 348–355.
92. Gonzalez, J.; Santolaria, C. Unsteady flow structure and global variables in a centrifugal pump. J. Fluids Eng.
2006, 128, 937–946.
93. Huang, S.; Mohamad, A.A.; Nandakumar, K.; Ruan, Z.Y.; Sang, D.K. Numerical simulation of unsteady
flow in a multistage centrifugal pump using sliding mesh technique. Prog. Comput. Fluid Dyn. 2010, 10,
239–245.
94. Huang, S.; Su, X.; Guo, J.; Yue, L. Unsteady numerical simulation for gas-liquid two-phase flow in
self-priming process of centrifugal pump. Energy Convers. Manag. 2014, 85, 694–700.
Energies 2018, 11, 180 40 of 41

95. Shojaeefard, M.H.; Tahani, M.; Ehghaghi, M.B.; Fallahian, M.A.; 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.
96. Shojaeefard, M.H.; Boyaghchi, F.A.; Ehghaghi, M.B. Experimental study and three-dimensional numerical
flow simulation in a centrifugal pump when handling viscous fluids. IUST Int. J. Eng. Sci. 2006, 17, 53–60.
97. Sirino, T.; Stel, H.; Morales, R.E. Numerical study of the influence of viscosity on the performance of an
electrical submersible pump. In Proceedings of the ASME 2013 Fluids Engineering Division Summer
Meeting, Incline Village, NV, USA, 7–11 July 2013.
98. Stel, H.; Sirino, T.; Prohmann, P.R.; Ponce, F.; Chiva, S.; Morales, R.E. CFD investigation of the effect of
viscosity on a three-stage electric submersible pump. In Proceedings of the ASME 2014 4th Joint US-
European Fluids Engineering Division Summer Meeting Collocated with the ASME 2014 12th International
Conference on Nanochannels, Microchannels, and Minichannels, Chicago, IL, USA, 3–7 August, 2014.
99. 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.
100. Li, W.G. Mechanism for onset of sudden-rising head effect in centrifugal pump when handling viscous oils.
J. Fluids Eng. 2014, 136, doi:10.1115/1.4026882.
101. Flores, N.G.; Goncalves, E.; Rolland, J.; Rebattet, C. Head drop of a spatial turbopump inducer. J. Fluids Eng.
2008, 130, doi:10.1115/1.2969272.
102. Jeanty, F.; De Andrade, J.; Asuaje, M.; Kenyery, F.; Vásquez, A.; Aguillón, O.; Tremante, A. Numerical
simulation of cavitation phenomena in a centrifugal pump. In Proceedings of the ASME 2009 Fluids
Engineering Division Summer Meeting, Vail, CO, USA, 2–6 August 2009; pp. 331–338
103. Barrios, L.; Prado, M.G.; Kenyery, F. CFD modeling inside an electrical submersible pump in two-Phase
flow condition. In Proceedings of the ASME 2009 Fluids Engineering Division Summer Meeting, Vail, CO,
USA, 2–6 August 2009; Volume 1, pp. 457–469.
104. Zhu, J.; Zhang, H.-Q. CFD simulation of ESP performance and bubble size estimation under gassy
conditions. In Proceedings of the SPE Annual Technical Conference and Exhibition, Amsterdam,
The Netherlands, 29 October 2014.
105. Zhu, J.; Zhang, H.-Q. Numerical study on electrical-submersible-pump two-phase performance and
bubble-size modeling. SPE Prod. Oper. 2017, 32, 267–278.
106. 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.
107. Tremante, A.; Moreno, N.; Rey, R.; Noguera, R. Numerical turbulent simulation of the two-phase flow
(liquid/gas) through a cascade of an axial pump. J. Fluids Eng. 2002, 124, 371–376.
108. Caridad, J.; Kenyery, F. CFD analysis of electric submersible pumps (ESP) handling two-phase mixtures.
J. Energy Resour. Technol. 2004, 126, 99–104.
109. Caridad, J.; Asuaje, M.; Kenyery, F.; Tremante, A.; Aguillón, O. Characterization of a centrifugal pump
impeller under two-phase flow conditions. J. Pet. Sci. Eng. 2008, 63, 18–22.
110. Qi, X.; Turnquist, N.; Ghasripoor, F. Advanced electric submersible pump design tool for geothermal
applications. GRC Trans. 2012, 36, 543–548.
111. Marsis, E.; Pirouzpanah, S.; Morrison, G. CFD-based design improvement for single-phase and two-phase
flows inside an electrical submersible pump. In Proceedings of the ASME 2013 Fluids Engineering Division
Summer Meeting, Incline Village, NV, USA, 7–11 July 2013.
112. Yu, Z.; Zhu, B.; Cao, S. Interphase force analysis for air-water bubbly flow in a multiphase rotodynamic pump.
Eng. Comput. 2015, 32, 2166–2180.
113. Yu, Z.Y.; Zhu, B.S.; Cao, S.L.; Wang, G.Y. Application of two-fluid model in the unsteady flow simulation
for a multiphase rotodynamic pump. In Proceedings of the IOP Conference Series: Materials Science and
Engineering, Kuala Lumpur, Malaysia, 2–4 July 2013; p. 52.
114. Pineda, H.; Biazussi, J.; López, F.; Oliveira, B.; Carvalho, R.D.; Bannwart, A.C.; Ratkovich, N. Phase
distribution analysis in an Electrical Submersible Pump (ESP) inlet handling water-air two-phase flow
using Computational Fluid Dynamics (CFD). J. Pet. Sci. Eng. 2016, 139, 49–61.
115. Zhang, H.-Q.; Wang, Q.; Sarica, C.; Brill, J.P. Unified model for gas-liquid pipe flow via slug dynamics—
Part 1: Model development. J. Energy Resour. Technol. 2003, 125, 266–273.
116. Zhang, H.-Q.; Wang, Q.; Sarica, C.; Brill, J.P. Unified model for gas-liquid pipe flow via slug dynamics—
Part 2: Model validation. J. Energy Resour. Technol. 2003, 125, 274–283.
Energies 2018, 11, 180 41 of 41

117. Ye, Z.; Rutter, R.; Martinez, I.; Marsis, E. CFD and FEA-Based, 3D metal printing hybrid stage prototype
on electric submersible pump ESP system for high-gas wells. In Proceedings of the SPE North America
Artificial Lift Conference and Exhibition. Society of Petroleum Engineers, Woodlands, TX, USA,
25–27 October 2016.
118. Zhang, H.-Q.; Wang, Q.; Sarica, C.; Brill, J.P. A unified mechanistic model for slug liquid holdup and
transition between slug and dispersed bubble flows. Int. J. Multiph. Flow 2003, 29, 97–107.
119. Zhang, H.-Q.; Wang, Q.; Sarica, C.; Brill, J.P. Unified model of heat transfer in gas-liquid pipe flow.
In Proceedings of the SPE Annual Technical Conference and Exhibition, Houston, TX, USA, 26–29
September 2004.
120. Zhang, H.-Q.; Sarica, C. Unified modeling of gas/oil/water pipe flow-basic approaches and preliminary
validation. SPE Proj. Facil. Constr. 2006, 1, 1–7.
121. Padron, G.A. Effect of Surfactants on Drop Size Distributions in a Batch, Rotor-Stator Mixer. Ph.D. Thesis,
University of Maryland, College Park, MD, USA, 2004.
122. Zhu, J.; Guo, X.; Liang, F.; Zhang, H.-Q. Experimental study and mechanistic modeling of pressure surging
in electrical submersible pump. J. Nat. Gas Sci. Eng. 2017, 45, 625–636.
123. Hinze, J.O. Fundamentals of the hydrodynamic mechanism of splitting in dispersion processes. AIChE J.
1955, 1, 289–295.
124. Levich, V. Physicochemical Hydrodynamics, 1st ed.; Prentice Hall: Upper Saddle River, NJ, USA, 1962.
125. Kouba, G.E. Mechanistic models for droplet formation and breakup. In Proceedings of the ASME/JSME
2003 4th Joint Fluids Summer Engineering Conference, Honolulu, HI, USA, 6–10 July 2003; Volume 1,
pp. 1607–1615.
126. Schiller, L.; Naumann, A. Fundamental calculations in gravitational processing. Z. Ver. Dtsch. Ing. 1933, 77,
318–320.
127. Clift, R.; Grace, J.; Weber, M. Bubbles, Drops, and Particles; Academic Press: New York, NY, USA, 1978.
128. Ishii, M.; Zuber, N. Drag coefficient and relative velocity in bubbly, droplet or particulate flows. AIChE J.
1979, 25, 843–855.
129. Mei, R.; Klausner, J.F.; Lawrence, C.J. A note on the history force on a spherical bubble at finite Reynolds
number. Phys. Fluids 1994, 6, 418–420.
130. Tran-Cong, S.; Gay, M.; Michaelides, E.E. Drag coefficients of irregularly shaped particles. Powder Technol.
2004, 139, 21–32.
131. Van Nierop, E.A.; Luther, S.; Bluemink, J.J.; Magnaudet, J.; Prosperetti, A.; Lohse, D. Drag and lift forces on
bubbles in a rotating flow. J. Fluid Mech. 2007, 571, 439–454.
132. Legendre, D.; Magnaudet, J. The lift force on a spherical bubble in a viscous linear shear flow. J. Fluid Mech.
1998, 368, 81–126.
133. Rastello, M.; Marié, J.L.; Grosjean, N.; Lance, M. Drag and lift forces on interface-contaminated bubbles
spinning in a rotating flow. J. Fluid Mech. 2009, 624, 159–178.
134. Rastello, M.; Marié, J.L.; Lance, M. Drag and lift forces on clean spherical and ellipsoidal bubbles in a solid-
body rotating flow. J. Fluid Mech. 2011, 682, 434–459.

© 2018 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/).