You are on page 1of 10

Original Article

Flow simulation in radial pump impellers


and evaluation of slip factor

Proc IMechE Part A:


J Power and Energy
0(0) 110
! IMechE 2015
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0957650915594953
pia.sagepub.com

Mohamed G Khalafallah, Hassan A Elsheshtawy, Abdel-Naby


M Ahmed and Ahmed I Abd El-Rahman

Abstract
This work is concerned with the study of the slip phenomenon in centrifugal pumps and the evaluation of its dependence
on the flow rate for a four-bladed pump. The finite volume method is used, and the impeller domain is represented by a
structured grid topology. The calculations assume a rotationally periodic boundary condition, while the frozen rotor
technique is used to model the interaction between the pump impeller and its surrounding volute casing. The simulation
uses an implicit time integration of the dynamic equations and is carried out using the commercial ANSYS CFX-solver.
Results from the simulation are found in reasonable agreement with the pump performance curve with a maximum
relative error of 4% in the range of flow coefficient from 0.8 to 1.2. The calculated values of the slip factor, as a function of
the flow rate, show good agreement with the Qius mathematical model while retaining the default value of the defined
shape factor F 0.52. In this particular study, the results show that although the slip factor improves with the increase of
either the number of blades or splitter length, the corresponding predicted hydraulic efficiency decreases due to the
increasing friction loss.
Keywords
Centrifugal pumps, turbomachinery flow, pump performance/efficiency, slip behavior/factor
Date received: 16 December 2014; accepted: 5 June 2015

Introduction
The slip phenomenon takes place in radial machines
as a result of the induced relative ow circulation, and
consequently, the uid becomes unable to faithfully
follow the guiding blades. The slip factor is a measure
of such ow deviation and is dened in terms of the
exit whirl velocity. Slip leads the ow to leave
the impeller with a mean relative angle 2 less than
the blade exit angle 20 , as shown in Figure 1. This
results in a signicant reduction in the work done on
the uid, and consequently, the pump head is dramatically inuenced
HE

Cu20 U2 U22
U2 Q


cot 20
gD2 b2
g
g

A pump theoretical head HE is typically represented


by the Euler equation (equation (1)), where all terms
except the ow rate Q are constant. Here, U2 is the
impeller peripheral speed, while D2 and b2 refer to the
impeller diameter and the passage width, respectively,
all measured at the exit section. However, due to slip
and other losses, the actual head and ow rate (HQ)
characteristics exhibit a signicant deviation from the
corresponding Euler behavior. Keeping that in mind,
the slip in radial impellers depends on various

parameters, such as the number of blades and the


blade geometry, and possibly varies with the pump
ow rate. Although increasing the number of blades
helps reduce the ow deviation in the impeller exit, it
also promotes blockage in the ow channels.
Therefore, a careful study on the eect of the
number of blades on the slip factor, as well as the
developed head and the pump eciency is considered
in this work.
Stodola and Loewenstein1 was the rst to develop a
mathematical expression for the slip factor in centrifugal pumps. He considered a circular control volume
lling the blade passage near the impeller exit and
assumed that the slip velocity is caused by the relative
eddy motion. He derived his equation for the slip velocity CSL in terms of the exit blade angle 20 and the

Department of The Mechanical Power Engineering, Cairo University,


Giza, Egypt
Corresponding author:
Ahmed I Abd El-Rahman, Department of The Mechanical Power
Engineering, Cairo University, Cairo University Rd., Giza 12613, Egypt.
Email: aiarahman@eng.cu.edu.eg

Downloaded from pia.sagepub.com by guest on September 16, 2015

Proc IMechE Part A: J Power and Energy 0(0)

Figure 1. The slip-influenced (solid lines) and the theoretical


blade-geometry (dashed lines) exit velocity triangles. CSL is the
slip velocity, considered by Stodola and Loewenstein1 in his
definition for the slip factor  1  CSL =U2.

number of blades Z, as follows


CSL

 sin 20
U2
Z

Using the potential ow theory, Busemann2 developed a theoretical framework for the estimation of
the slip velocity and the calculation of the slip factor
for several blade angles and number of blades. His
results were plotted as a function of the inlet-tooutlet radius ratio and indicated constant slip behavior zone at small inlet-to-outlet radius ratios followed
by a sharp reduction in slip at higher values of the
inlet-to-outlet radius.
Later, Wiesner3 presented a general review of the
various prediction methods, developed for the calculation of basic slip factors, applicable for centrifugal
impellers. He concluded the rst part of his work by
supporting the validity of the classical theoretical
method of Busemann.2 Furthermore, he carefully
explored the Busemann experimental results and proposed the following simpler empirical expression for
slip factor estimation, while considering the impeller
exit velocity diagram, illustrated in Figure 1
p
sin 20
 1
Z0:7

Although limited by the impeller inlet-to-outlet radius


ratio, Wiesners values showed a more accurate t to
the Busemann test data. Wiesner then carried out
extensive comparisons of slip factors, with test data,
reported in the literature for more than 60 dierent
impeller geometries that further demonstrated its high
potential in describing the slip phenomenon. Thus,
Wiesners slip model is currently widely accepted for
centrifugal pump design.
Von Backstrom and Theodor4 derived a unied
correlation for the slip factor assuming a single relative eddy centered on the rotor axis in his uid
dynamic model. He ignored other mechanisms

aecting the slip phenomenon such as the blade turning angle and the ow-induced wakes, and proposed a
model to calculate the magnitude of the recirculating
ow caused by the relative eddy. He argued that the
eddy-induced slip velocity is dependent on the blade
solidity (c=s2 , where c and s2 correspond to the chord
and pitch at the blade exit, respectively) and dened
the slip factor in terms of the normalized slip velocity,
following the work of Wiesner,3 as introduced in
equation (4). In his trial to unify the previously
derived formulas for the slip factor, Backstrom compared his results with other attempts to demonstrate
its feasible replacement; however, his model does not
show any dependence on the pump ow rate
 1  1=1 5cos 20 0:5 c=s2 

Memardezfouli and Nourbakhsh5 experimentally studied the slip factor in centrifugal pumps at dierent
ow rates. They compared their results using ve different industrial pumps with the theoretical slip factors modeled by Wiesner3 and Stodola and
Loewenstein.1 They found good agreement at the
pump best eciency point (BEP), whereas a signicant divergence was found at o-design conditions,
specically at low-ow rates operating regimes. They
further dened the local slip factor and illustrated its
nonuniform distribution in the blade-to-blade passage. A relative decrease in the local slip factor was
noticed, while moving across the streamlines from the
blade pressure side (PS) to the blade suction side (SS).
Although inconsistent, their values for the mean slip
factors at the impeller exit showed clear dependence
on the ow rates and the number of blades.
Further, Caridad and Kenyery6 carried out a 3D
computational uid dynamics (CFD) simulation
using the commercial CFX-solver package on ve different impellers of known geometries and specic
speeds. They were able to calculate the slip factor
for both single- and two-phase ow and found a similar linearly decreasing behavior for all slip curves with
the increase of the pump ow coecient. The predicted slip factor was reported to decrease as the
gas-void fraction increases from 10 to 17%. They further compared their slip results with values produced
by Wiesner3 and Stodola and Loewenstein1 correlations and showed discrepancies as large as 52%.
Qiu et al.7 were the rst to consider the inuence of
the variable pump ow rate in their derivation for a
unied slip factor. Qiu et al. distinguished the mechanism controlling the ow behavior, and thus the slip
factor, within the impeller at the impeller exit into
three components. The rst component represented
the radial Coriolis eect, evident for typical radial
impellers. The second contribution accounted for the
blade-turning rate (d=dm, where m represents the
meridional distance on the ZR plane) characterizing
the extra loading from the streamline curvature, while
the third component described the weak eect of the

Downloaded from pia.sagepub.com by guest on September 16, 2015

Khalafallah et al.

passage width variation on the back ow and wakes


generation (db=dm, where  is the ow density). Qiu
et al.7 dened the slip factor in terms of the slip velocity normalized by the rim rotor speed
 
F cos 2 sin 2
Fs2 2 d

Z
4 cos 2 dm 2


F2 s2 sin 2 db

dm 2
42 b2

 1

Here, F,  2, and 2 refer to the shape factor, as dened


by Qiu et al.,7 the meridional inclination angle, and
the exit ow coecient, respectively. Furthermore,
their model was validated for several case studies
including an industrial pump with an actual bladeturning rate d=dm of 4.92 degrees/m. An adjustable
tting parameter F 0.6 was needed to obtain a good
match with the test data. Qiu et al. also compared
their slip results with corresponding values, calculated
using other prediction methods, but specically indicated a close agreement with the Wiesner3 model.
Recently, a CFD simulation, using the commercial
software package of STAR-CD, of the ow eld in
centrifugal compressors was reported by Huang et al.8
They considered the eect of the mass ow rate, the
blade exit angle 2, the blade wrap angle  (see Figure
2), and the number of blades Z. Considering the
Busemann mathematical model,2 they dened the
slip factor in terms of exit ow angle as
 1  tan 2  tan 20

Cm2

U2

For simulation and comparison purposes, they considered the geometry of Eckardt9 rotor A impeller.

(a)

Huang et al. showed that the onset of slip occurs close


to the exit section at a normalized camberline distance
at which the deviation of the ow angle from the
blade angle begins to increase. The calculated total
pressure ratio and slip factor were successfully compared to Eckardt9 results. Surprisingly, their slip
factor exhibited a slight rise with the increase of the
mass ow rate.
Another recent numerical attempt has been
reported by Li10 who performed a CFD analysis of
the slip factor in a centrifugal pump. The slip factor
was calculated from the velocity triangles at the impeller exit and from the impeller theoretical head. He
concluded that the numerical results from the two
methods were inconsistent. Also, the eect of ow
rate on the slip factor was dramatic for blades
having small exit angles, as expected, while the viscosity had minimal eect on the slip factor.
To the best of our knowledge, no simulation has
been developed which provides accurate prediction of
the slip characteristics of a typical centrifugal pump
and shows plausible comparison with actual performance and empirical relations. The goal of the present
study is to develop a nite volume model which helps
to understand the ow behavior within the blade-toblade passage of the chosen centrifugal pump and
reveals the slip variation with the pump ow rate.
Also the study aims to verify the relation suggested
by Qiu et al.7 as the authors believe that their developed expression could predict the slip factor at dierent ow rates reasonably well. A parametric study
about the eect of varying the number of blades and
splitters on both the slip behavior and the pump performance is presented.
The next section elaborates the generation of the
pump geometry along with the meshing details. This is

(b)

Figure 2. The 2D schematic drawing (a) and the full 3D model (b) of the pump geometry.

Downloaded from pia.sagepub.com by guest on September 16, 2015

Proc IMechE Part A: J Power and Energy 0(0)

followed by the Setup of the numerical model section, in which the implementation of the periodic
boundary condition is rst introduced. This section
also presents the model validation by comparing the
numerical values with the corresponding pump characteristic curve. In the Results section, an illustration of the predicted ow eld is presented, followed
by the slip factor evaluation and comparison with different mathematical models. The dependence of the
slip factor and the pumps hydraulic eciency on
the number of blades and splitter blades is nally
considered.

Pump geometry and domain meshing


In this study, an industrial centrifugal pump with four
blades is chosen for computational modeling and
simulation purposes. To replicate the pump geometry,
the CFturbo11 software is used to draw the contours
of both the impeller blade and the volute casing. The
impeller blade assumes a shape of a fourth-order polynomial prole that leads to an exit blade angle 2 of
17.5 and inlet blade angle 1 of 20.0 , measured with
respect to the tangential direction. Here, the angles 
and  are measured with respect to the radial direction, as illustrated in Figure 2(a). The volute has a
tongue angle  of 29 and an inner diameter D3 of
397 mm, approximately. The main dimensions of the
pump geometry are summarized in Table 1.
The impeller and volute data are then passed to the
ANSYS BladeGen12 software, to produce the 3D
model of the centrifugal pump, as shown in

Figure 2(b). The used splitters have three dierent


lengths of 30, 50, and 70% of the full-blade chord.
The ANSYS Turbogrid13 program is used to generate
the 3D nite-volume hexahedral elements for accurate
and fast CFD analysis of the ow behavior within the
radial impeller. On the other hand, an unstructured
mesh is applied to the volute computational domain,
as shown in Figure 3(a). The general grid interface is
used to couple the structured impeller grid and
unstructured volute grid while the frame change
frozen rotor model is used to model the rotor rotating
domain relative to the stator stationary domain, as
described in detail in the ANSYS CFX14 User
Documentation.
In this work, the entire domain is rst considered
for validation purpose by including the volute eect in
the simulation to allow comparison of the HQ characteristics. The study then proceeds to predict the slip
behavior within the impeller passage and at the impeller exit. For this objective, it is unreasonable to
numerically model the pump entire domain since it
contains a periodically repeating ow eld between
each two successive blades, instead, a representative
control volume is considered and appropriate boundary conditions are imposed at its boundaries, as indicated in Figure 3(b). To ensure that the enclosed
domain behaves as a representative section of the
entire impeller, rotationally periodic boundary conditions are enforced on the control volume surfaces in
the peripheral direction. This approach also helps to
reduce the computational eort.

Setup of the numerical model


Table 1. The pump main dimensions (mm, degrees).
D1

D2

D3

b2

177

360

397

29.5

124

29

In the present model, the atmospheric pressure is


imposed at the impeller inlet, while the outlet boundary condition is set as a constant mass ow rate. The
no-slip boundary condition is applied at the impeller

Figure 3. (a) illustrates the impeller structured mesh versus the unstructured mesh of the volute casing, while plot (b) shows the
meshing and the imposed boundary conditions of the periodic domain of the impeller.

Downloaded from pia.sagepub.com by guest on September 16, 2015

Khalafallah et al.

and volute casing walls. A turbulence intensity of 5%


is imposed at the inlet section.
The present model solves the quasi-steady
Reynolds-averaged NavierStokes equations using
the two-equation ke turbulence model and a coupled
solver. It is worth noting here that the ke model does
not yield the best accuracy for cases when either stall
or reversed ow is present inside the domain.15 The
convergence criteria settings include the specication
of the maximum number of iterations to 1000 and the
residuals to 104, while the imbalances in the governing equations are set to less than 1%. The evolution of
the total pressure is also monitored as the simulation
proceeds to verify and record the attained steady-state
solution. The transient algorithm is employed with a
specied time increment of 60/N, where N refers to the
impellers number of revolutions per minute. Here,
N 1450 r/min.
Figure 4(a) shows the results of the grid independence study applied to the considered periodic domain.
It is clear that a grid size of 366,485 elements is sucient and is thus used in the rest of this study. This
simulation run is carried out at the pumps BEP, at
which the actual ow rate and head are 260 m3/h and
38 m, respectively. Furthermore, to validate the
numerical model, the pump characteristic curve is predicted and compared with the available actual curve.
Computations are carried out using the ke turbulence model.
Figure 4(b) compares the present numerical results
of the HQ curve with the corresponding actual
values. Here, both the pump head H and the pump
ow rate Q are normalized with respect to their values
at the BEP. Good agreement is found in the range of
normalized ow rate between 0.8 and 1.2 with a maximum relative error of 4%. However, in relatively
high and low ow rate regimes, the dierence between
the numerical and test results increases up to about
13%. Such a discrepancy may be attributed to the
inability of the present CFD model to properly
account for the expected separated ow regimes,
under high heads, far from the design point and the
eect of leakage under high ow rates.

(a)

Results
In this section, the numerical slip results are presented
and compared with a few mathematical models, along
with a description of the ow behavior within the
impeller passage. In the following simulation runs,
spatial variations in the ow static pressures and
meridional velocities are presented, followed by a
detailed discussion of the eect of increasing the
number of blades and adding splitters on the evaluated slip as a function of the pump ow rate.

Investigation of the flow field


Figure 5 shows the variation of both the meridional
velocity and the static head along the blade pitch at
three dierent radial locations, namely those corresponding to 25, 50, and 81% of the passage radial
length. The results in Figure 5 refer to the mean
ow behavior (i.e. along a plane located midspan
between the hub and the shroud). Here, the fourbladed impeller operates at its design condition.
With such low number of blades, the impeller-discharged ow is expected to become nonuniform
over the blade pitch. Figure 5(b) shows that at the
25% section, the meridional velocity prole is nearly
parabolic along 2/3 of the blade pitch, measured from
the SS, but then deforms as the ow moves outward
to become strongly nonuniform over the blade pitch
at the impeller exit.
It is clear that in the stream-wise ow direction, the
velocity gradually decreases as the ow moves outward because of the widening cross-sectional area of
the ow passage, while a corresponding increase in the
static head is indicated in Figure 5(d). The evolvement
of the recirculating ows nearby the leading edge on
top of the blade PS is shown in Figure 5(a). This is
typically generated because of the large blade-turning
rate that leads to boundary layer separation. The
inuence of wakes on the meridional velocity is
observed along the remaining 1/3 of the blade pitch,
noting that it becomes less signicant at the impeller
exit, as shown in Figure 5(b).

(b)

Figure 4. The pump head grid size dependence is presented in (a), while (b) refers to the model validation. The solid dots refer to
the present simulation results, while the solid line represents the pump characteristic curve.

Downloaded from pia.sagepub.com by guest on September 16, 2015

Proc IMechE Part A: J Power and Energy 0(0)

To help to illustrate the induced separation zone


within the impeller passage, the velocity vectors are
plotted in Figure 6. The incoming ow is shown to
separate closely at the leading edge on the blade PS.
The portion occupied by the wakes increases to block
about 30% of the blade pitch and then gradually
shrinks allowing the ow to eventually rell the
entire domain.

Slip factor evaluation

To further illustrate the signicant eect of the pump


ow rate on the slip factor, Figure 7(a) shows the
variation of the slip factor with the exit ow coecient 2, in comparison with the Wiesner and Qiu
et al. models. The present simulation shows a linear
reduction in the slip factor in the range of the ow
coecient from 0.075 to 0.105. Consistent results, but
underpredicted by 11.6%, are estimated by the Qiu
et al. model in the same range of the exit ow coecient. The eect of ow rate is represented by the third
term of equation (5), in which the blade-turning rate

In the present study, a separate postprocessing


MatLab program is developed to calculate the slip
velocity using the data of the tangential and radial
velocity components (Cu2 and Cm2), together with
the relative exit blade angles 20 , according to equation (7). This is then followed by the direct estimation
of the slip factor at the impeller exit. A comparison
between the numerical slip result and available empirical formulas, at the BEP, is shown in Table 2. The
mean slip factor at the impeller exit is calculated to be
0.58. The present result is seen to be in close agreement with both the Wiesner3 and the Qiu et al.7
models. The adjustable parameter is selected to
equal 0.6, as suggested by Qiu et al. in their model
CSL U2  Cu2 

Cm2
tan 20

Figure 6. The figure shows a contour plot of the flow absolute velocity, overlapped by the velocity vectors at BEP.

Figure 5. Contour plots of the meridional-velocity and the static-head distributions between two successive blades at three different radial locations. The normalized arc distance is the ratio of the circumferential distance to the local blade pitch. (a) Contours of
the meridional velocity, (b) meridional velocity profile, (c) contours of the static head, (d) static head profile.

Downloaded from pia.sagepub.com by guest on September 16, 2015

Khalafallah et al.

causes extra loadings due to streamline curvature. The


larger the amount of ow rate, the less controlled the
ow guidance and the smaller the slip factor. In this
particular run, the slip factor varies from 0.6 at
2 0.075 to 0.41 at 2 0.127. The similarity
between the present CFD results and the calculated
results according to the Qiu et al. model suggests that,
if the tting parameter F in equation (5) is retained in
its default value 0.52 instead of 0.6, a better agreement
with Qius prediction is achieved, as demonstrated in
Figure 7(a). Such validation has not been reported
before.
The local slip factor is calculated, using the local
velocity components Cu2i and Cm2i, and evaluated at
three dierent span locations over the blade pitch at
the impeller exit, namely the 0% span (hub), 50%
span, and 100% span (shroud). The variation of the
local slip factor as a function of the normalized arc
distance at the impeller exit is introduced in Figure
7(b). Consistent with our understanding of the ow
behavior near the solid boundaries, the local slip
factor exhibits a remarkable increase at the hub section as compared to those values estimated at the
mean and shroud sections because of the blade load
distribution along with the secondary ow eect.
Moreover, the comparison reveals a relatively lower
slip factor at 100% spanwise position, corresponding
to the impeller top section. This is due to the fact that,
in our particular model, the pump is unshrouded. This
result supports our knowledge that a shrouded impeller helps improving the slip characteristics and that an
unshrouded impeller induces a variation of the slip
factor with the ow rate.
Next, the number of blades is varied to study the
eect of their variation on the slip phenomenon, while

retaining the original impeller geometry. Specically,


impellers having six, eight, and 12 blades are considered. In principle, using more blades would result
in larger slip factors because of the improved guidance
the uid experiences through the impeller. As anticipated, the present numerical result shows a remarkable increase in the slip factor using more blades, as
indicated in Figure 8(a) and (b). Figure 8(b) shows
that as the number of blades increases the slip factor
asymptotically approaches a maximum of 0.69. It is
important to note that equation (5) denes the three
main factors aecting the slip in centrifugal impellers.
The contribution by the number of blades is obvious
in the second term of the right-hand side, thus, justifying the above eect of the number of blades on the
slip behavior.
Two dierent approaches were reported to enable
the investigation of the energy saving due to splitter
blades. Golcu et al.16 considered standard impellers
having dierent numbers of main blades, tted with
splitter blades having dierent lengths. The overall
number of blades (main and splitter) varies in their
results. In a recent study, Cavazzini et al.17 compared
the performance of centrifugal pump both with and
without splitters. In their work, the overall number of
blades (main and splitter) is preserved and set to eight.
Here, we pursue Cavazzinis approach.
The usefulness of using splitters in a conventional
pump is demonstrated through the comparison presented in Figure 8(c). With respect to the performance
of the eight-bladed standard impeller, the present
CFD simulation captures a relative improvement in
the slip behavior through the replacement of full original blades with splitter blades having dierent
lengths. It is clear that as the splitter length is

Table 2. The present slip factor versus reported correlations.


Z

Stodola

Backstrom

Eckardt

Qiu et al.

Wiesner

Present

0.7638

0.7275

0.8146

0.515

0.6288

0.58264

(a)

(b)

Figure 7. The figure shows the slip factor variation with the exit flow coefficient as compared to Wiesner and Qiu models (a) and
the change of the local slip factor over the blade pitch (b) at the impeller exit. Here, Z 4.

Downloaded from pia.sagepub.com by guest on September 16, 2015

Proc IMechE Part A: J Power and Energy 0(0)

(a)

(b)

(c)

Figure 8. The variation of the slip factor as a function of the pump flow rate using a different number of blades is shown in (a),
whereas its dependence on the number of blades at the best efficiency point is presented in (b). Plot (c) shows the influence of the
addition of splitters while preserving the overall number of blades (main and splitter). The blue squares refer to the standard eightbladed impeller having no splitters.

increased from 30 to 50%, the original slip factor is


fairly recovered. However, no further improvement is
noticed using a 70% splitter. Therefore, it is concluded that the 50% splitter is sucient to achieve
slip behavior comparable to that obtained using the
standard impeller.

Pump head and hydraulic efficiency


This section presents the pump performance represented by the pump head and the pump hydraulic eciency at dierent exit ow coecient 2. The
numerical values extracted from ANSYS CFX solver
enable us to calculate the pump head and eciency.
Contrary to the reduced slip, Figure 9(a) and (b)
shows a remarkable decrease in the pump head with
less impact on the corresponding pump hydraulic eciency by increasing the number of blades. This is possibly because of the built-up of frictional losses that
overcomes the energy harvested due to slip reduction.
The eect of replacing half of the original full
blades by splitter blades is further examined in
Figure 9(c) and (d). The gure shows the eect of
splitters on the pump head and hydraulic eciency
while preserving the overall number of blades (main
and splitter). A remarkable increase in the calculated
pump head along with a limited increase in the corresponding hydraulic eciency, particularly below the
BEP, is obtained as the splitter length is increased
from 30 to 50%. This anomaly in the dierence

between the insertion of a whole blade and the insertion of a splitter might be due to the blockage and the
skin friction eects that further deteriorates the performance of a whole blade as compared to a splitter.
Thus, we conclude that for the pump under consideration replacing half the full original blades with 50%
span splitters helps increase the pump head by at least
20% with minimal inuence in the pump hydraulic
eciency at the pumps BEP.

Conclusion
Numerical simulation of ow through a centrifugal
pump is carried out to study the behavior of the
ow through the impeller and to evaluate the slip
factor as function of the ow rate, the number of
blades, and the splitter length. It is shown that the
computational results are in good agreement with
the pumps performance curve in the neighborhood
of the BEP but is only in fair agreement with it near
the neighborhood of the low and high ow rates. The
overall slip factor is found to change linearly with the
ow coecient in a similar way as the expression by
Qiu et al.7 Reducing the tting parameter F given by
Qui et al. from 0.6 to its default value 0.52 gives very
good agreement. This indicates that the Qui et al. relation for estimating slip, although it is well accepted,
needs further investigation to evaluate the tting parameters correctly in its relations to the impeller
geometry.

Downloaded from pia.sagepub.com by guest on September 16, 2015

Khalafallah et al.

(a)

(b)

(c)

(d)

Figure 9. The pump head and hydraulic efficiency are plotted for a different number of blades and splitters lengths. In subplots (c)
and (d), the blue dots refer to the eight-bladed standard impeller having no splitters. The overall number of blades (main and splitter) is
preserved.

The eect of increasing the number of blades to up


to 10 for the tested pump, is found to increase the slip
factor to reach saturation at about 0.69. Also, the
insertion of 30 and 50% splitters increases the slip
factor. However, the 70% span splitter does not
show any further improvement in the slip behavior.
The application of splitters and the increase of the
number of blades are also examined in their inuence
on both the head and the hydraulic eciency. The
splitter blades are found to increase the developed
head by about 20% at the BEP, while a decrease in
the pump head is noticed when using more blades. It
is clear that there is a contrast when applying these
modications although both of them perform the
same function of improving the ow guidance. The
reason for this anomaly is that improving ow guidance increases the slip factor which in turn increases
the developed head; however, the higher incurred friction losses and blockage eect due to whole blades
insertion oset the improvement due to better guidance. Using splitters, improvement in the hydraulic
eciency is found with ow rate up to
Q=QBEP  1:0, which is then followed by slight
decrease due to the increased friction losses.
References
1. Stodola A and Loewenstein LC. Steam and gas turbines:
with a supplement on the prospects of the thermal prime
mover. New York: Peter Smith, 1945.

2. Busemann
A.
Das
Forderverhaltnis
radialer
Kreiselpumpen mit logarithmisch-spiraligen Schaufeln.
ZAMM 1928; 8: 372384.
3. Wiesner FJ. A review of slip factors for centrifugal impellers. J Eng Gas Turbines Power 1967; 89:
558566.
4. von Backstrom and Theodor W. A unified correlation
for slip factor in centrifugal impellers. J Turbomach
2005; 128: 110.
5. Memardezfouli M and Nourbakhsh A. Experimental
investigation of slip factors in centrifugal pumps. Exp
Thermal Fluid Sci 2009; 33: 938945.
6. Caridad JA and Kenyery F. Slip factor for centrifugal
impellers under single and two-phase flow conditions.
J Fluids Eng 2005; 127: 317321.
7. Qiu X, Japikse D, Zhao J, et al. Analysis and validation
of a unified slip factor model for impellers at design and
off-design conditions. J Turbomach 2011; 133: 19.
8. Huang J, Luo K, Chen C, et al. Numerical investigations of slip phenomena in centrifugal compressor
impellers. Int J Turbo Jet-Engines 2013; 30: 123132.
9. Eckardt D. Flow field analysis of radial and backswept
centrifugal impellers, part I. In: Proceedings of 25th
ASME gas turbine conference and 22nd annual fluids
engineering conference, New Orleans, 1980, New York,
NY: American Society of Mechanical Engineers,
1979; pp.7786.
10. Li W. Effects of flow rate and viscosity on slip factor of
centrifugal pump handling viscous oils. Int J Rotating
Mach 2013; 2013: 112.
11. CFturbo Turbomachinery Design System, Release 9.0,
Users Guide. Software & Engineering GmbH, Inc.

Downloaded from pia.sagepub.com by guest on September 16, 2015

10

Proc IMechE Part A: J Power and Energy 0(0)

12. ANSYS BladeModeler, Release 14.5, components


System, ANSYS Bladegen Users Guide. ANSYS, Inc.
13. ANSYS Turbogrid, Release 14.5, components System,
ANSYS Turbogrid Users Guide. ANSYS, Inc.
14. ANSYS CFX, Release 14.5, Help System, CFX
Documentation. ANSYS, Inc.
15. Shojaeefard MH, Tahani M, Ehghaghi MB, et al.
Numerical study of the effects of some geometric characteristics of a centrifugal pump impeller that pumps a
viscous fluid. Comput Fluids 2012; 60: 6170.

16. Golcu M, Pancar Y and Sekmen Y. Energy saving in a


deep well pump with splitter blade. Energy Convers
Manag 2006; 47: 638651.
17. Cavazzini G, Pavesi G, Santolin A, et al. Using splitter
blades to improve suction performance of centrifugal
impeller pumps. Proc IMechE, Part A: J Power and
Energy 2015; 229: 309323.

Downloaded from pia.sagepub.com by guest on September 16, 2015

You might also like