-
This article was downloaded by: [120.164.45.33]On: 09 June 2015,
At: 07:58Publisher: Taylor & FrancisInforma Ltd Registered in
England and Wales Registered Number: 1072954 Registered office:
MortimerHouse, 37-41 Mortimer Street, London W1T 3JH, UK
Engineering Applications of Computational
FluidMechanicsPublication details, including instructions for
authors and subscription
information:http://www.tandfonline.com/loi/tcfm20
A Study in The Use of CFD In The Design ofCentrifugal PumpsS.
Yedidiahaa aaa 89 Oakridge Rd, West Orange, NJ 07052, USA
E-Mail:Published online: 19 Nov 2014.
To cite this article: S. Yedidiah (2008) A Study in The Use of
CFD In The Design of Centrifugal Pumps, EngineeringApplications of
Computational Fluid Mechanics, 2:3, 331-343, DOI:
10.1080/19942060.2008.11015233
To link to this article:
http://dx.doi.org/10.1080/19942060.2008.11015233
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy
of all the information (the Content) containedin the publications
on our platform. However, Taylor & Francis, our agents, and our
licensors make norepresentations or warranties whatsoever as to the
accuracy, completeness, or suitability for any purpose ofthe
Content. Any opinions and views expressed in this publication are
the opinions and views of the authors,and are not the views of or
endorsed by Taylor & Francis. The accuracy of the Content
should not be reliedupon and should be independently verified with
primary sources of information. Taylor and Francis shallnot be
liable for any losses, actions, claims, proceedings, demands,
costs, expenses, damages, and otherliabilities whatsoever or
howsoever caused arising directly or indirectly in connection with,
in relation to orarising out of the use of the Content.
This article may be used for research, teaching, and private
study purposes. Any substantial or systematicreproduction,
redistribution, reselling, loan, sub-licensing, systematic supply,
or distribution in anyform to anyone is expressly forbidden. Terms
& Conditions of access and use can be found at
http://www.tandfonline.com/page/terms-and-conditions
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3, pp. 331343 (2008)
A STUDY IN THE USE OF CFD IN THE DESIGN OF CENTRIFUGAL PUMPS
S. Yedidiah
89 Oakridge Rd, West Orange, NJ 07052, USA E-Mail:
[email protected]
ABSTRACT: CFD has the potential of assisting an engineer in
arriving at improved designs. However, to be effective, this
requires a much closer cooperation and mutual understanding between
the pump specialist and the expert in CFD, than it is presently in
existence. This conclusion is based on actual case histories from
past experience, as well as considerations of the physical meaning
of certain mathematical expressions. The presented discussion
relates, primarily, to the design of centrifugal pumps. However,
there are strong indications that analogical situations exist also
in other fields of fluids engineering.
Keywords: centrifugal pumps, CFD, flow pattern, impeller,
volute, noise, pump design
1. INTRODUCTION
Successes in predicting the flow patterns within the passages of
centrifugal pumps indicate that CFD might be capable of assisting a
pump engineer in arriving at improved designs. Success, however, is
still very elusive. This paper discusses the principal cause of
that state of affair, and what can be done about it. For a
predicted flow pattern to be of practical use, it has to inform the
engineer whether it will allow the pump to perform at its best,
whether there exists a different flow pattern which will allow the
pump to perform even better, and what geometry of the waterways
will generate such a flow pattern. The nearest which a predicted
flow pattern came to answering some of the above questions is based
on the assumption that the presence of vortices have an adverse
effect on performance. This has been confirmed in practice
(Yedidiah, 1996: p. 191). However, cases in which the presence of a
pair of forced vortices has even improved the performance of a
centrifugal pump are also known. Below we shall discuss such a
case.
2. EFFECT OF A PAIR OF FORCED VORTICES AT THE PUMP INLET ON ITS
PERFORMANCE
Prang and Oates (1971) presented results of tests, which were
expected to shed light on the effect of the distribution of the
inlet-velocities on the performance of a centrifugal pump. The
objective of these tests was to determine the optimum
geometry of a side-inlet suction-nozzle (Fig. 1) of a
centrifugal pump. Suction nozzles of different geometries have been
prepared in transparent epoxy (Prang and Oates, 1971). The flow
patterns within their passages were observed visually by
introducing into the flowing liquid small amounts of glitter. In
addition to the above, the distributions of the axial and of the
tangential velocity components were measured at the outlet of these
suction nozzles along the diameter M-M, as shown in Fig. 2. (During
the performance tests, these outlets of the suction nozzles became
the inlets to the tested pump). These distributions were taken at a
number of discrete flow rates. For comparison, the distributions of
the same velocity components were also taken at the same flow
rates, when the pump was tested as an end-suction pump with a
straight, concentric suction pipe. In the latter case, the
measurements were taken along a vertical diameter, 5.5 upstream of
the suction flange of the pump. Fig. 3 presents the results of
these measurements for the side-suction inlet-nozzle shown in Fig.
2, as well as for the straight, concentric suction pipe. The
measurements shown in Fig. 3 were taken at the flow rate of 56
m3/hour. After completing these observations and measurements, each
of the experimental side-suction nozzles was assembled with the
same 328 end-suction centrifugal pump, thus converting it into a
side-suction unit. Afterwards, it was tested for performance. For
comparison, the same pump was also tested as a straight end-suction
pump with a straight, concentric suction pipe.
Received: 1 Dec. 2007; Revised: 5 Feb. 2008; Accepted: 4 Mar.
2008
331
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3 (2008)
Fig. 1 Centrifugal pump with side-inlet suction
nozzle.
Velocitym/sec
RADIUS (inches)
Fig. 4 presents the results of the performance tests carried out
with the suction-nozzle shown in Fig. 2, as well as with the
end-suction pipe. It shows that the performance with the
side-suction nozzle (which has produced a pair of forced vortices)
is superior to its performance with the straight, end-suction
pipe.
Fig. 2 Suction chamber used in the tests reported in
Fig. 3 and Fig. 4.
Fig. 3 Distribution of the axial and of the tangential
velocity components at the inlet of the tested pump at 56 m3/hr,
when tested as a straight end-suction unit and with the suction
chamber shown in Fig. 2.
At the first sight, these results seem to present an
impossible-to-solve enigma. However, a closer look at the physical
meaning of Eulers equations of motion of an inviscid incompressible
liquid demonstrates that there is nothing mysterious about the test
results shown in Fig. 4.
332
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3 (2008)
Fig. 4 Performance of the experimental pump, when
tested as a straight end-suction unit and as a side-inlet pump
with the suction nozzle shown in Fig. 2.
3. THE PHYSICAL MEANING OF EULERS EQUATIONS
A translation of Eulers equations of motion into their physical
meaning provides a feasible explanation of the cause of the test
results shown in Fig. 4. These equations are frequently expressed
in the following forms:
zV
yV
xV
xP
xxx
+
+
=
zyxx VVV1F
zV
yV
xV
yP
yyy
+
+
=
zyxy VVV1F (1)
zV
yV
xVP
zxz
+
+
=
zyxz VVVz1F
Let us have a look at the term x
Vx
xV .
As velocity means the derivative of distance with respect to
time , therefore, the above term can be reduced to:
)T/xV( x =
TV
xV
Tx
xV xxx
=
=
xV (2) H m
This means that the studied term expresses acceleration (rate of
change of velocity with respect to time). In a similar manner, it
is possible to prove that all other terms on the right side of
Eulers equations represent acceleration. To be more specific, each
of these equations expresses an adaptation of Newtons Second Law of
Motion to an inviscid, incompressible liquid. The accelerations in
such a liquid are the result of the combined action of external
forces plus the forces generated by the pressure gradients which
exist within the acted-upon liquid. These pressure gradients are
the generating forces (Yedidiah, 2004) which accelerate the liquid
in the direction opposite to the direction in which these pressures
are increasing. (This accounts for the minus sign of the terms
which express the pressure gradients).
100 m3/hour
EFF
When applied to the flow through an impeller, these equations of
motion tell us how the geometry and the motion of the
impeller-passages will alter the velocity distributions of the
incoming liquid. This has been confirmed by the experiments
reported in Hureau et al. (1993). 100 m3/hourLet us have a look at
the shroud-to-shroud distribution of the normal velocity components
measured along the axis 10 (upstream of the inlet edges of the
blades) and along the axis 6 (upstream of the outlet edges of the
blades) of the tested impeller (Fig. 5). Figs. 6 and 7 show these
distributions at two flow rates: at the design flow, and at 0.41 of
that flow.
10
6
Fig. 5 Axes, along which the normal velocity
components shown in Figs. 6, 7 and 8 were measured.
333
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3 (2008)
Fig. 6 Shroud-to-shroud distribution of the normal
velocity components upstream of the inlet edges of the blades
along axis 10 shown in Fig. 5 at optimum flow and at 0.41 of the
optimum flow.
Fig. 7 Shroud-to-shroud distribution of the normal
velocity components upstream of the outlet edges of the blades
along axis 6 shown in Fig. 5 at optimum flow and at 0.41 of the
optimum flow.
Fig. 8 Shroud-to-shroud distribution of the normal
velocity components along axis 10 shown in Fig. 5 at different
fractions of the optimum flow rate.
We note that at both flow rates, the geometry and the motion of
the impeller-passages have a similar effect on the distribution of
these velocity components: they have increased their relative
magnitudes towards the suction-shroud, and reduced them towards the
hub-shroud. However, due to the initial differences in these
distributions, they have converted a fairly uniform distribution of
the inlet velocities (at the design flow) into a very non-uniform
one upstream of the outlet edge. Also, they have converted a very
non-uniform distribution of the velocities at the inlet (at 0.41 of
the design flow) into a fairly uniform one, upstream of the outlet
edge.
nq 20Cn/U1 [ - ]
S/So [ % ]
It is possible that the improvements in performance shown in
Fig. 4 are due to the fact that the presence of the pair of the
forced vortices upstream of the impeller-eye have resulted in a
more advantageous distribution at the outlet of the blades. Also,
the advantages of that improvement in the velocity distribution at
the impeller-outlet had a more beneficial effect on performance
than the possible adverse effects of the presence of the pair of
forced vortices which existed upstream of the blades.
Cn/U2 [ - ]
This is yet not all. A glance at Fig. 6 tells us that this
colossal difference in the distribution of the inlet velocities was
caused by recirculation. This effect, as has been explained in
Yedidiah (1996 and 2005a), develops below a certain flow rate
(compare Fig. 8). The flow rate, below which recirculation starts
to develop, depends among others upon the geometry of the
blades.
S/So [ % ]
4. EFFECTS OF BLADE GEOMETRY ON INLET RECIRCULATION
Fig. 9 shows the effects of reducing the projected area of those
inlet-parts of the blades, which extend into the eye of the
impeller. It shows that a reduction in the magnitude of that area
also reduces the intensity and the flow rate at which recirculation
sets in (Yedidiah, 2005a). The blades of that particular impeller
were purely cylindrical. This makes it possible to establish a
simplified qualitative model of the manner, in which the projected
area of these parts of the blades is affecting recirculation.
nq 20Cn/U1 [ - ]
With reference to Fig. 10, let us consider the effect of a strip
of radial width Ro and of length Lo, which extends into the
impeller-eye. The volume of liquid directly affected by such a
strip of the blades is equal to (Yedidiah, 2003):
S/So [ % ]
oouoobo RL)CU(Q = (3)
334
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3 (2008)
335
And the head added to that acted-upon volume of liquid is equal
to (Yedidiah, 2003):
gCuo
oo Uh = (4)
Eq. (5) tells us that Ho increases with Qbo. The latter, as
follows from Eq. (3), increases with the product (Lo Ro). This
explains the effects shown in Fig. 9.
Consequently, the total head added to the total volume of the
pumped liquid by that strip of the blade is equal to (Yedidiah,
2003):
q
bogQQCuooo UH = (5)
The above discussions demonstrate that when applying CFD to the
study of the effects of the impeller-passages on the performance of
a centrifugal pump, it is mandatory to take into account a
significantly larger amount of parameters than those which are
presently included in any known CFD programs. In particular, that
recirculation is only one of the numerous factors which affect the
flow through an impeller (Ref. 18). The above conclusion is also
confirmed by the following case history.
This increase in the head of the liquid will generate a pressure
gradient which will accelerate the liquid in the direction opposite
to the direction in which the increase of that pressure takes
place. The magnitude of that acceleration will, of course, increase
with the increase in Ho.
Fig. 9 Effect of the blade parts which project into the eye of
the impeller on recirculation.
H mm Hg
Q/Qopt
Fig. 10 A simplified model of the effect of a strip of a blade
which projects into the eye of an impeller on the intensity of
recirculation.
36r
63r
R bR a51
r
OrignialImpeller
Orignial Impeller
OriginalImpeller
Ro Ro
L o
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3 (2008)
5. EFFECT OF THE STRENGTH OF FORCED INLET-VORTICES ON THE NPSH
REQUIREMENTS OF A PUMP
Fig. 11 presents the distribution of the axial and of the
tangential velocity components at the outlet of two different
side-suction nozzles (Prang and Oates, 1971). Both nozzles
resembled the shape shown in Fig. 2. However, the radius R (Fig. 2)
in shape C was larger, and its center was located nearer to the
diameter M-M than in shape D. Both geometries produced a pair of
forced vortices. However, the geometry D has produced stronger
vortices than the geometry C at all flow rates. These differences,
as can be seen in Fig. 12, had no effect on either the QH-curve or
on the efficiencies of the tested pump. However, the suction nozzle
which has generated stronger vortices has allowed the pump to
operate at lower available NPSH-values. The reason for this effect
is simply as follows.
Fig. 11 Distribution of the tangential and of the axial
velocity components at the outlet of two different
side-inlet-suction nozzles (nozzle C and nozzle D) at two different
flow rates.
Fig. 12 Performance of the same pump when tested
with suction nozzle C and suction nozzle D.
The impeller of the test pump used in Prang and Oates (1971) had
purely radial blades, which projected only a short distance into
the suction-eye. The pairs of the forced vortices have increased
the pressures at the outermost radii of the pump inlet. This means,
near the inlet-tips of the impeller blades. This is the location
where cavitation is most likely to start at reduced available
NPSH-values. Stronger vortices produce larger increases in
pressure. Consequently, the stronger vortices made it possible for
the pump to operate at lower available NPSH. CFD is a very powerful
and versatile logical tool, which is capable of providing engineers
with enormous assistance in arriving at better results. However,
versatility implies that it can be used in many different ways,
depending on the task it has to accomplish. To be of practical use,
CFD has to be applied in a manner specifically adapted to handle
the problem in question. This requires an in-depth knowledge and
understanding of the problem(s) to be handled. The present
explosion of information makes it prohibitive for the expert in CFD
even to try to master the design of centrifugal pumps. Similarly,
the pump specialist cannot afford to be distracted from his field
of activity in order to attempt to master CFD. The only feasible
solution to such a state of affair is teamwork. However, for
336
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3 (2008)
teamwork to be successful, each member of the team has to know
and to understand what the others are doing. This includes the
knowledge of how, where and when the use of CFD is the best choice
for a given task. Below, we shall discuss two design problems,
where CFD might be capable of providing extremely useful assistance
to pump designers.
6. A DESIGN PROBLEM IN WHICH CFD MIGHT PROVE EXTREMELY
USEFUL
When a liquid is flowing between the blades without separation,
the law of continuity mandates that the flow should satisfy the
following condition (Yedidiah, 2007) (see Fig. 13):
= ba fe dm dnwBdsCzR RB)2( (6) In Eq. (6), BR signifies the
width of the impeller at any given radius R, ds signifies the
length of an element of a given arc of radius R, Bd signifies the
width of the impeller at the center of an inscribed circle of
diameter d (Fig. 13).
Fig. 13 Schematic flow between impeller blades of a
centrifugal pump (hypothetical).
This means that the flow rate across any arc a-b of radius R is
the same, as the flow rate across any normal line e-f. If we assume
that the liquid is flowing through a set of congruent passages
(Fig. 13) and that Cm is constant across any arc of a given radius
R, Eq. (6) can be re-written as:
== fe d dnwBzRzQ mRCB)2( (7)
In a correctly designed impeller, the magnitude of the relative
velocities w, within the passages created by the overlapping
portions of the blades, is, in most cases, selected to vary
linearly with the radius. In that case, Eq. (7) can be approximated
by the following expression:
dBwCB)2( dcdmR =zR (8)
In Eq. (8), BR is the width of the impeller at any given radius
R, wd is the calculated relative velocity at the center of any
inscribed circle of diameter d, and Bdc is the width of the
impeller at the center of the same inscribed circle. At the first
sight, Eq. (8) seems to have no relevance to the actual pump
performance. After all, in a real pump, the liquid never flows in
congruent partial passages. Also, the magnitudes
are never the same across the total length of any arc a-b of a
given radius R. Still, a study of how far the geometry of the
passages deviates from the condition expressed by Eq. (8) provides
the pump designer with useful information.
of Cm
Let us see what will happen when our original assumption that
the magnitudes of Cm across any arc of a given radius R are
identical is not correct. Let the actual magnitudes of Cm, within
certain partial passages, be equal to Cm (1+c). For a given flow
rate, this means that there may exist an equal number of passages,
in which the magnitudes of Cm are equal to Cm (1c). The magnitudes
of the relative velocities are equal to:
f b e a d
n
.SinwCm = This means that if Eq. (8) is satisfied in the case
when Cm = constant along any given arc of radius R, it will also be
valid in many other cases. In summary, this leads to the following
conclusions.
R
If an examination of the passages reveals that Eq. (8) is
satisfied along their total length, there exists a possibility that
the flow through the passages will really be free of any major
disturbances. However, if we shall find significant differences
between the flow rates calculated by each side of Eq. (8), this
signifies the probability of major disturbances or even reversals
of flow.
B
In practice, certain deviations from the condition imposed by
Eq. (8) are allowable or even desirable (Ref. 18: sub-section B-3).
However, large differences between the flow rates calculated by
means of each side of Eq. (8) will unavoidably lead to undesirable
results. This conclusion has also been confirmed by the test
results reported in Yu, Cua and Leo (2001).
337
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3 (2008)
Fig. 14 Effect of the width of the passages between
the overlapping portions of the blades on the flow of the liquid
within an impeller (Yu, Cua and Leo, 2001).
Fig. 14 presents the distribution of the velocities within one
of the impellers studied in Yu, Cua and Leo (2001). For the tested
flow rate, the calculated correct distance between the blades at
the inlet of the passages (Ref. 18: sub-section 30-2.1) is equal to
the diameter of the circle inscribed (in Fig. 14) in one of the
passages marked I. We see here that the distances between the
blades at the inlets of the passage marked II are considerably
larger than the calculated ones. This has generated huge zones of
separation. It even gave rise to a significant reversal of flow
back into the suction-eye of the impeller. In a radial-flow
impeller, it is very easy to verify whether the geometry of the
passages deviates too much from the condition expressed by Eq. (8).
However, in a mixed-flow impeller, the task is much more
complicated. In particular, in such an impeller, the distribution
of the velocity components is usually drastically altered by the
condition of radial equilibrium (Ref. 18: sub-section 16; Yedidiah,
2005b). Here, CFD has the potential of becoming the best or maybe
even the only tool for optimizing the design. Of course, this is
not the only problem which could be best handled with the aid of
CFD. Below, we shall bring up a second case history in which only
CFD is capable of making it possible to arrive at the optimum
results.
7. IMPROVING THE EFFICIENCY OF A CERTAIN CLASS OF CENTRIFUGAL
PUMPS
Decades ago, the author has developed a very successful method
for designing sewage handling
pumps. From a commercial point of view, all pumps designed with
the aid of that method are regarded as having the highest quality.
However, when compared to the efficiencies attainable with
conventional pumps of the same size and specific speed (Stepanoff,
1957: Fig. 5.1), there were significant differences between the
attainable values and the results of tests (Yedidiah, 1997).
Particularly, baffling were the extremely high efficiencies of the
pump (Yedidiah, 1997: #6 in Table II). The expected peak efficiency
for a conventional pump of that size and that specific speed is
84%. The test results (Fig. 15) for that pump show considerably
higher magnitudes (over 89%). It took the author more than three
decades to find an explanation for that mystery.
II
I I
II II
I I
2 1
1 m/s II 3 4
Quadrant 1 4
Impeller C
Hm
m3/hour
Fig. 15 Performance of a 5 sewage disposal pump, designed in
1958.
It has been demonstrated in Yedidiah (2003) and confirmed by
results of tests discussed in Yedidiah (2001) that the head
developed by a blade of an impeller can be expressed by the
following equation:
)21(
))((CE 1n
nnq
nnvnqunvn
rRQ
RCEQQKCQ
+
+=
(9)
where nmun CotCUC = . The above equations imply that the head
developed by an impeller-blade is determined by its geometry (not
by its tip angles). This conclusion has also been corroborated by
the results of independent tests carried out thousands of miles
apart (Acosta and Bowerman, 1957; Saalfield, 1966). An impeller of
a sewage disposal pump is provided with a pair of heavy blades,
whose thickness varies along their total length. This
338
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3 (2008)
means that each side of a blade is determined by a different
curve (Fig. 16). This fact, as is implied in Eqs. (9) and (10),
causes each side of a blade to develop a different head.
Calculations based on these equations have demonstrated that
smaller differences between the heads developed by each side of the
blades are always producing higher efficiencies (Yedidiah, 1997),
as compared to what is expected from a conventional pump of the
same size and specific speed.
Fig. 16 Typical shape of the impeller blades of a
sewage disposal pump.
The above findings provide us with a very important clue to
improve the efficiencies of sewage disposal pumps. Also, with
certain adaptations, they may become capable of assisting a pump
designer in improving the efficiencies of any pump which is
equipped with mixed-flow blades. However, the need to take into
account many additional design parameters (as discussed in Ref. 18)
makes it almost impossible to know if the solution is really the
best possible one. Here, the use of CFD might be capable of
providing extremely useful assistance. While CFD may be the only
logical tool capable of optimizing certain design requirements of a
centrifugal pump, it does not yet mean that its use will produce a
pump which will perform better than any other existing ones. There
exists a huge number and variety of factors which are capable of
affecting the performance of a pump (Yedidiah, 1996 & 2006).
Below, we shall discuss one of such factors.
8. EFFECT OF THE DISTANCE BETWEEN THE BASE-CIRCLE OF THE VOLUTE
AND THE IMPELLER-RIM
Fig. 17 is a schematic presentation of the flow within a volute.
When the distance between the impeller-rim and the base circle of
the volute is equal to m, the liquid which is issued from the
impeller at the angular position from the vertical centerline is
barely capable to enter the discharge nozzle of the pump (Yedidiah,
2002). Now, let us see what will happen to the flow of the liquid
which exits the impeller at the same angular position , if we shall
reduce this distance m by, say, n. In that case, the liquid which
exits the impeller at the same angular position would have been
forced to re-enter the impeller at the position C (Fig. 17). This
study implies that pumps of higher specific speeds need greater
distances between the impeller-rim and the base circle of the
volute (as the ratio between the meridian velocity components and
the peripheral velocity components, at the impeller outlet,
increases with specific speed). This has been confirmed in practice
(Stepanoff, 1957: Fig. 7.8). However, the distance between the
impeller-rim and the base circle of the volute is not the only
parameter which affects the flow within an impeller. It is evident
that each of the volute shapes shown in Fig. 19 will have a
different effect on that distribution. This has been confirmed by
the tests reported in Benra et al. (2007). Instead of a double
volute, a diffuser ring has been used in those tests. The radial
distance between the impeller-rim and the blades of the diffuser
was very small. Nevertheless, these experiments have shown no
effect of the circumferential position of the impeller blades in
relation to the diffuser blades. The differences between the test
results shown in Fig. 19 and the results reported in Benra et al.
(2007) have a simple logical explanation. The discussion regarding
Fig. 17 leads to the conclusion that the effects observed in Fig.
18 are primarily due to the spiral motion of the liquid within the
casing. For such a spiral motion to develop, the casing has to be
adequately wider than the outlet passages of the blades. In the
pump tested in Benra et al. (2007) (a diffuser pump), the width of
the casing was of the same order of magnitude as the outlet
passages of the impeller (see Fig. 20). This explains the
difference in the results of the tests shown in Fig. 18 and the
results reported in Benra et al. (2007).
339
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3 (2008)
340
In an impeller with backward-curved blades, the returning liquid
will usually penetrate only a short distance into the impeller
passages before it will be expelled into the volute for a second
time. In most cases, this means that the liquid will have to make
at least one additional turn around the impeller before exiting the
volute. In an impeller
with straight, radial blades, the liquid may even return all the
way back into the impeller-eye. This has been confirmed by the
experiments reported and discussed in Yedidiah (2002) and Yu, Cua
and Leo (2001). This can be clearly seen in the impeller passages
facing the tongues of the (double) volute in Fig. 18.
Fig. 17 Effect of the distance between the base circle of the
volute and the impeller rim on the flow within an impeller
(schematic).
Fig. 18 Effect of volute geometry on the flow within an impeller
(Yu, Cua and Leo, 2001; Yedidiah, 2002).
n
C
m
Fig. 19 Some of the frequently used cross-sections of a
volute.
(a) (b) (c)
VOLUTE DIFFUSOR
Fig. 20 Effect of casing geometry on the spiral motion of the
liquid at the outlet of an impeller (schematic).
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3 (2008)
At present, we do not know of any method which is capable of
determining the optimum relative distance between the impeller-rim
and the base circle of a volute. Only CFD is offering logical tools
which are capable of leading us to the desired solutions. The above
study also illustrates the enormous amount of knowledge which is
needed for arriving at a desired solution. At the present explosion
of information, we cannot expect a single person to become an
expert in more than one single area of engineering. The only
obvious solution to that situation is teamwork. To arrive at
meaningful useful solutions, there is a need for close cooperation
and mutual understanding between experts in CFD and centrifugal
pump specialists. Finally, an engineer also needs to keep in mind
that the capability of a logical tool for solving a given problem
does not always mean that its use is the best choice for the given
task. This is illustrated by the following case history.
9. A CASE HISTORY IN WHICH THE USE OF CFD HAS PROVEN INFERIOR TO
OTHER APPROACHES
An improved design tool means that it provides a faster, more
reliable, and less expensive means for achieving the targeted
objective(s). In many cases, only CFD might be able to satisfy the
above requirements, particularly in situations where a significant
amount of trial and error work is required. However, in practice,
there exist many areas in which CFD offers very little advantage
over other known design procedures. Such a case history is
discussed below. Ballesteros-Tajadura et al. (2006) presented the
results of a study in which CFD has been used in an attempt to
reduce the noise generated by a centrifugal fan. This study has
correctly predicted that the most intensive noise will occur at the
passing frequency of the blades. The fact that the passing
frequency of the blades is a major source of noise is known to the
pump engineers for more than half a century (Yedidiah, 1996:
229231). What an engineer needs to know is whether the intensity of
that noise will be within acceptable limits. The graphs presented
in the published paper seemed to show a good agreement between the
predicted and the tested intensities of noise. However, the author
also presented tests results which showed that the predicted
intensities were significantly different from the tested ones. This
means that the approach used in Ballesteros-Tajadura et al.
(2006)
can hardly be regarded as a reliable means for predicting the
level of noise which a given design will develop. We can expect
that experts in CFD will find a more reliable way for predicting
the intensity of that noise in the future. In that case, it will be
possible to predict the maximum noise levels of several different
designs, and to select this one, which has been found capable to
operate within the acceptable limits of noise. This, however, will
not solve a different problem. To the best of the authors
knowledge, it took a mainframe, high-speed computer 15 hours to
arrive at the results presented in Ballesteros-Tajadura et al.
(2006). If we shall add to this the time spent by the engineering
and the auxiliary staff, and take into account that this has been
spent on only one single design, we arrive at the conclusion that
the use of CFD for predicting the levels of noise of a number of
different designs is rather a very costly and time-consuming
venture, considering that there is known a much simpler and by far
less expensive way of reducing the noise generated at the passing
frequency of the blades (Ref. 18: Figs. 11-17 and 11-18). The
author has applied this method in 1965, long before CFD became a
popular subject of study. This solution, while it can hardly be
classified as being based on fluids dynamics, has proven to be so
simple, effective and inexpensive that the author made it a
standard feature of all his subsequent pump designs. Since the
introduction of that design feature, the author has never again
encountered any problem of excessive noise caused by the passing
frequency of the blades.
10. CONCLUSIONS
Properly applied, CFD has the potential of providing the
practicing engineers with enormous assistance in their quest for
better designs. However, to know how, where and when the use of CFD
is the best choice for a given task requires an in-depth knowledge
and understanding of the problem(s) to be handled. At the present
explosion of information, this requires close cooperation between
the pump experts and the specialists in CFD. Such cooperation,
however, can be effective only if each member of a team knows and
understands what the others are doing. The presented discussion, as
well as the studies in Yedidiah (2003, 2004, 2007 & Ref. 18),
indicates that the needed mutual understanding between the pump
experts and the specialists in CFD can be
341
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3 (2008)
significantly facilitated by considering the physical meaning of
mathematical expressions.
NOTATION
B blade width mC meridian velocity component of the
liquid mavC average value of Cm, at any given radiusuC velocity
component, in the direction of
U, of the liquid which has been directly affected by the
blade
CE equivalent magnitude of Cu, related to the total flow of the
pumped liquid
CU magnitude of Cu for the case when Cm = Cmav
E energy per unit of time (=power) g acceleration due to gravity
H total head h head added to the liquid by a strip of a
moving blade K coefficient used in equations
L length of a blade element along the blade surface
oL length of a strip of a blade which projects into the eye of
an impeller
n total length of a normal line which extends between two
consecutive blades (Fig. 13)
sN specific speed: (m3/sec)0.5
(rev/min) / (m)0.75 Q flow rate of pumped liquid
qQ flow rate between two consecutive blades (=Q/Z)
bQ volume of liquid displaced by a strip of a moving blade
vQ volume displaced by a finite blade element R
R radius oS distance between the shrouds, measured
along the axis #10 respectively axis #6 (shown in Fig. 5)
S Distance from suction shroud, (relates to Figs 6, 7, and
8)
S length of an arc of a given radius R (relates to Fig. 13)
T time t pitch of blades U peripheral velocity of a blade V
velocity of the liquid
Z,Y,X location, determined by Cartesian coordinates
zZ, number of blades b blade angle (=w+) w angle of relative
velocity
angle of incidence density angular speed radians/sec efficiency
blockage due to finite thickness of the
blades head coefficient = gH / U22 flow coefficient = Cm2 / U2
Subscripts 0 refers to the impeller-eye 1 refers to the leading
edge of a moving
blade 2 refers to the outlet tip of a moving bladei refers to
the inlet tip of a stationary
blade ou refers to the outlet tip of a stationary
blade av average value based on the magnitude
of Cmav b refers to a blade n refers to the n-th element of a
blade n-1 refers to the preceding blade element R refers to the
radius, at which the given
dimension was taken x,y,z refers to the directions of
Cartesian
coordinates
REFERENCES
1. Acosta AJ, Bowerman RD (1957). An experimental study of
centrifugal pump impellers. Trans. ASME J. of Fluids Eng.
81:18211838.
2. Ballesteros-Tajadura Rafael et al. (2006). Prediction of the
aerodynamic noise generated by a centrifugal fan. ASME-Publication
FEDSM2006-98507.
3. Benra FK et al. (2007). Measurement of the periodic flow
field in a radial diffuser pump by the PIV method. ASME-Paper
FEDSM2007-37400 [CD-ROM].
4. Hureau F et al. (1993). Study of internal recirculation in
centrifugal impellers. ASME-Publication FED. Vol. 154, 151157.
5. Prang AJ, Oates DM (1971). Suction Chambers for In-line
Centrifugal Pumps with Inducers. Worthington, Canada.
6. Saalfield K (1966). Einige neuere Gedanken zur
Laufradberechnung von radialen und halbaxialen Kreiselpumpen (in
German). KSB-Technicsche Berichte, KSB-Company, Frankenthal
(Pfalz), Germany. Aug. 1966, 11.
342
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
-
Engineering Applications of Computational Fluid Mechanics Vol.
2, No. 3 (2008)
343
7. Stepanoff AJ (1957). Centrifugal and Axial Flow Pumps. John
Wiley & Sons, Inc., New York.
8. Yedidiah S (1996). Centrifugal Pump Users GuidebookProblems
and Solutions. Chapman & Hall, New York. (At present: Springer
Science and Business Media, Inc., Norwell, MA).
9. Yedidiah S (1997). How to design more efficient pumps. ASME
Paper FEDSM97-3383.
10. Yedidiah S (2001). Practical applications of a recently
developed method for calculating the head of a rotodynamic
impeller. Proc. Inst. Mech. Eng. Vol. 215, Part A, 119131.
11. Yedidiah S (2002). Effect of pump geometry on the flow
within a centrifugal impeller. Proc. Inst. Mech. Eng. Vol. 216,
Part C, 11451149.
12. Yedidiah S (2003). An overview of methods of calculating the
head of a rotodynamic impeller, and their practical significance.
Proc. Inst. Mech. Eng. Vol. 217, Part E, 221232.
13. Yedidiah S (2004). A bridge between science and practical
engineering. ASME Paper HT-FED2004-56011 [CD-ROM].
14. Yedidiah S (2005a). An updated study of recirculation at the
inlet of a rotodynamic impeller. ASME-Paper FEDSM2005-77048.
15. Yedidiah S (2005b). Certain effects of blade-geometry and of
inlet-recirculation on the distribution of the meridian velocities
at the impeller-outlet. ASME-Paper FEDSM2005-77041 [CD-ROM].
16. Yedidiah S (2006). Translating equations into their physical
meaning, as an effective tool of engineering. ASME-Paper
FEDSM2006-980012 [CD-ROM].
17. Yedidiah S (2007). Application of CFD to the design of
centrifugal pumps. ASME-Paper FEDSM2007-37349 [CD-ROM].
18. Yedidiah S. Centrifugal Pumps: Theory, Design, Use,
Diagnostics, and Problem-solving (unpublished manuscript).
19. Yu SCM, Cua LP, Leo HL (2001). The angle-resolved velocity
measurements in the impeller passages of a model biocentrifugal
pump. Proc. Inst. Mech. Eng. Part C. Journal of Mechanical
Engineering Science 215(5):547568.
Dow
nloa
ded
by [1
20.16
4.45.3
3] at
07:58
09 Ju
ne 20
15
List of FiguresFig. 1 Centrifugal pump with side-inlet suction
nozzle.Fig. 2 Suction chamber used in the tests reported in Fig.3
and Fig.4.Fig. 3 Distribution of the axial and of the tangential
velocity components at the inlet of the tested pump at 56 m3/hr,
when tested as a straight end-suction unit and with the suction
chamber shown in Fig.2.Fig. 4 Performance of the experimental pump,
when tested as a straight end-suction unit and as a side-inlet pump
with the suction nozzle shown in Fig.2.Fig. 5 Axes, along which the
normal velocity components shown in Figs.6, 7 and 8 were
measured.Fig. 6 Shroud-to-shroud distribution of the normal
velocity components upstream of the inlet edges of the blades along
axis 10 shown in Fig.5 at optimum flow and at 0.41 of the optimum
flow.Fig. 7 Shroud-to-shroud distribution of the normal velocity
components upstream of the outlet edges of the blades along axis 6
shown in Fig.5 at optimum flow and at 0.41 of the optimum flow.Fig.
8 Shroud-to-shroud distribution of the normal velocity components
along axis 10 shown in Fig.5 at different fractions of the optimum
flow rate.Fig. 9 Effect of the blade parts which project into the
eye of the impeller on recirculation.Fig. 10 A simplified model of
the effect of a strip of a blade which projects into the eye of an
impeller on the intensity of recirculation. Fig. 11 Distribution of
the tangential and of the axial velocity components at the outlet
of two different side-inlet-suction nozzles (nozzle C and nozzle D)
at two different flow rates.Fig. 12 Performance of the same pump
when tested with suction nozzle C and suction nozzle D.Fig. 13
Schematic flow between impeller blades of a centrifugal pump
(hypothetical).Fig. 14 Effect of the width of the passages between
the overlapping portions of the blades on the flow of the liquid
within an impeller (Yu, Cua and Leo, 2001).Fig. 15 Performance of a
5 sewage disposal pump, designed in 1958.Fig. 16 Typical shape of
the impeller blades of a sewage disposal pump.Fig. 17 Effect of the
distance between the base circle of the volute and the impeller rim
on the flow within an impeller (schematic).Fig. 18 Effect of volute
geometry on the flow within an impeller (Yu, Cua and Leo, 2001;
Yedidiah, 2002).Fig. 19 Some of the frequently used cross-sections
of a volute.Fig. 20 Effect of casing geometry on the spiral motion
of the liquid at the outlet of an impeller (schematic).
1. INTRODUCTION2. EFFECT OF A PAIR OF FORCED VORTICES AT THE
PUMP INLET ON ITS PERFORMANCE3. THE PHYSICAL MEANING OF EULERS
EQUATIONS4. EFFECTS OF BLADE GEOMETRY ON INLET RECIRCULATION5.
EFFECT OF THE STRENGTH OF FORCED INLET-VORTICES ON THE NPSH
REQUIREMENTS OF A PUMP6. A DESIGN PROBLEM IN WHICH CFD MIGHT PROVE
EXTREMELY USEFUL7. IMPROVING THE EFFICIENCY OF A CERTAIN CLASS OF
CENTRIFUGAL PUMPS8. EFFECT OF THE DISTANCE BETWEEN THE BASE-CIRCLE
OF THE VOLUTE AND THE IMPELLER-RIM9. A CASE HISTORY IN WHICH THE
USE OF CFD HAS PROVEN INFERIOR TO OTHER APPROACHES10.
CONCLUSIONSNOTATIONREFERENCES