Chapter 10

Performance of
Centrifugal Pumps
PAUL COOPER
GEORGE TCHOBANOGLOUS

CONTRIBUTORS
Richard O. Garbus
Robert J. Hart
Carl W. Reh
Lowell G. Sloan
Earle C. Smith

The purpose of this chapter is to introduce those axial, or a combination of both (mixed). Strictly de-
fundamentals of centrifugal pump theory that are fined (as in European practice), a centrifugal pump is
useful for background and sometimes necessary for a radial-flow pump only. But colloquial usage is fol-
selecting and specifying centrifugal pumps in water lowed here, and thus centrifugal pumps are divided
and wastewater pumping applications. A correspond- into three groups:
ing body of theory for positive-displacement pumps
. Radial-flow pumps
can be found in the literature [1, 2, 3].
. Mixed-flow pumps
This introduction includes: (1) the general classifi-
. Axial-flow or propeller pumps.
cations of centrifugal pumps, (2) pump application
terminology and usage, (3) pump operating charac- These classifications are derived from the manner
teristics, (4) cavitation and net positive suction head, in which the fluid moves through the pump (see
(5) pump characteristic curves and operating ranges, Figure 10-1). Thus, the fluid is displaced radially in
and (6) an introduction to pumping system analysis. a radial-flow pump, axially in an axial-flow pump,
Standards and specifications such as ANSI/HI and both radially and axially in a mixed-flow pump.
9.8-1998 are commonly (and sufficiently) identified The physical features of centrifugal pumps are de-
by the call number and, hence, are not referenced in scribed in greater detail in Chapter 11.
Section 10-10.

10-2. Pump Application Terminology,

10-1. Classification of Centrifugal Pumps Equations, and Performance Curves

In colloquial usage in the United States, a ‘‘centrifu- The basic terminology, equations, and curves for
gal pump’’ is any pump in which the fluid is energized defining pump performance and solving pump prob-
by a rotating impeller, whether the flow is radial, lems are given in this section. The development of the

10.1
10.2 Chapter 10 Performance of Centrifugal Pumps

Figure 10-1. Typical flow paths in centrifugal pumps. (a) Radial flow, vertical; (b) mixed flow; (c) radial flow,
horizontal; (d) axial flow.

pump performance curves, which are used to define Head

the operating characteristics of a given pump, is
reviewed briefly at the end of this section (this subject The term head (h or H) is the elevation of a free
is covered in more detail in Chapter 12). surface of water above (or below) a reference datum
(see Figures 10-2 and 10-3). For centrifugal pumps,
the reference datum varies with the type of pump, as
Capacity shown in Figure 10-1. Head is expressed in meters (m)
or feet (ft). Pressure can also be expressed as the
The flow rate (capacity, discharge, or Q) of a pump is equivalent head of water. See Appendix A.
the volume of liquid pumped per unit of time, usually Distances (heads) above the datum are considered
expressed in SI units in cubic meters per second positive and distances below the datum are consid-
(m3 =s) for large pumps or liters per second (L/s) and ered negative. Each term, defined graphically in Fig-
cubic meters per hour (m3 =h) for small pumps. In ures 10-2 and 10-3, is expressed as the height of a
U.S. customary units, the capacity of a pump is ex- water column in meters (feet) of water. H is used for
pressed in gallons per minute (gal/min), million gal- total head, whereas h is used for head from the datum
lons per day (Mgal/d), or cubic feet per second (ft3 =s). or for headloss. The subscripts s and d denote the
Equivalent units of measurement are given on the pump suction and discharge, respectively. Other sub-
inside back cover and in Appendix A. scripts are defined as follows:
 10-2. Pump Application Terminology, Equations, and Performance Curves 10.3

Figure 10-2. Terminology for a pump with a positive suction head. (*) The gauge is located to show theoretical
pressures at the inlet and outlet flanges; see ‘‘Field Pump Tests’’ in Section 16-6 for practical gauge locations.

Total static head (Hstat ): The total static head is the ing is the height that a water column would attain in a
difference in elevation in meters (feet) between the vertical pipe. It is also the distance to the hydraulic
water level in the wet well and the water level at gradeline (shown dashed in Figures 10-2 and 10-3).
discharge (hd  hs ). Manometric head (Hg ): The increase of pressure
Static suction head (hs ): The static suction head is head, expressed in meters (feet) generated by the
the difference in elevation between the wet well liquid pump (hgd  hgs ).
level and the datum elevation of the pump impeller. If Friction headloss (hfs , hfd ): The head of water that
the wet well liquid level is below the pump datum, as in must be supplied to overcome the frictional loss in the
Figure 10-3, it is a static suction lift, so hs is negative. pipe. The frictional headloss in the suction (hfs ) and
Static discharge head (hd ): The static discharge discharge (hfd ) piping systems can be computed with
head is the difference in elevation between the dis- the Hazen–Williams or Darcy–Weisbach equations
charge liquid level and the pump datum elevation. (Equations 3-9 and 3-10).
Manometric suction head (hgs ): The suction gauge Velocity head (v2 =2g): The velocity head is the
reading is expressed in meters (feet) measured at the kinetic energy in the liquid being pumped at any
suction nozzle of the pump and referenced to the point in the system. The energy gradeline (shown
pump datum elevation and atmospheric pressure. solid in Figures 10-2 and 10-3) is always above the
Manometric discharge head (hgd ): The discharge hydraulic or piezometric or manometric gradeline
gauge reading is expressed in meters (feet) measured (dashed line) by v2 =2g. The velocity head in the dis-
at the discharge nozzle of the pump and referenced to charge pipe, v2d =2g; is lost if the pipe discharges freely
the centerline of the pump impeller. The gauge read- in air or if it discharges abruptly below the surface of
10.4 Chapter 10 Performance of Centrifugal Pumps

Figure 10-3. Terminology for a pump with a negative suction head. (*) The gauge location is to show theoretical
pressures at the inlet and outlet flanges; see ‘‘Field Pump Tests’’ in Section 16–6 for practical gauge locations.

a reservoir. Some of the velocity head can be recov- individual losses for each fitting and valve, but, as
ered if turbulence is inhibited by a gradually expand- explained in Section 3-4, the total may in reality be
ing section, but this is ordinarily impractical with the somewhat less or much greater than this sum.
pipe velocities normally encountered (i.e., up to Total dynamic head (Ht or TDH): The total dynamic
2.5 m/s or 8 ft/s). head is the head against which the pump must work. It
Fitting and valve losses (hfvs , hfvd ): As a fluid flows is determined by adding the static suction and dis-
through fittings and valves, energy is lost due to eddy charge head (with respect to signs), the frictional head-
formation and turbulence. Because the head lost in losses, the velocity heads, and the fitting and valve
fittings and valves is small compared with the friction headlosses. The expression for determining the total
loss in long piping systems, the losses in fittings and dynamic head for the pumps shown in Figures 10-2 and
valves are sometimes called ‘‘minor losses’’ and often 10-3 is given by Equations 10-1, 10-2, and 10-3.
ignored. But the lengths of pipes within a pumping
station are short, and the total headloss through the n2d n2s
fittings and valves is likely to be greater than the pipe Ht ¼ hgd  hgs þ  (10-1)
2g 2g
friction loss. Regardless of the shortness of pipe
length, both the frictional headloss and the ‘‘minor where
losses’’ should always be computed.
The loss of head through each individual fitting or hgd ¼ hd þ hfd þ hfvd (10-2)
valve is estimated using Equation 3-16 (h ¼ Kv =2g) 2
and
with K values taken from Tables B-6 and B-7 (or
from the literature). The total fitting and valve losses v2s
hgs ¼ hs  hent  hfs  hfvs  (10-3)
in the piping system are determined by summing the 2g
 10-2. Pump Application Terminology, Equations, and Performance Curves 10.5

Substituting Equations 10-2 and 10-3 into Equa- where P is the water power in horsepower (hp), g is
tion 10-1 and noting that hd  hs ¼ Hstat , the specific weight of the fluid in pounds per cubic
foot (lb=ft3 , see Table A-9), Q is the flow rate in ft3 =s,
Ht ¼ Hstat þ hent þ hfs þ hfd þ hfvs H is the total dynamic head in ft, 550 is the conver-
sion factor from foot-pounds per second to horse-
v2d
þ hfvd þ (10-4) power, W is weight of water pumped in pounds per
2g minute (lb/min), 33,000 is the conversion factor from
foot-pounds per minute to horsepower, q is flow rate
Some designers consider all headlosses except pipe in gal/min., and 3960 is used to convert gallon-feet
friction to be ‘‘minor’’ losses, hm , and rewrite Equa- per minute to horsepower for water at 15 to 208C (59
tion 10-4 as to 688F).

Ht ¼ Hstat þ hfs þ hfd þ hm (10-5)

Input Power

Power Pump performance is measured in terms of the flow

rate that a pump can discharge against a given head
Output Power at a given efficiency. The pump capacity depends on
the design, and design information is furnished by the
The power output of a pump is the useful energy pump manufacturer in a series of curves for a given
delivered by the pump to the fluid. In SI units, the pump. Pump efficiency, Ep , is the ratio of the useful
power output is defined as power output [water kilowatts (wkW) or water horse-
power (whp)] to the power input to the pump shaft.
Hence, the brake power (bkW) that must be supplied
qH
P ¼ gQH ¼ (10-6a) by the drive is, in SI units,
102

where P is the water power in kilowatts, g is the bkW ¼ gQH=Ep ¼ wkW=Ep (10-7a)
specific weight of the fluid in kiloNewtons per cubic
meter (kN=m3 , see Table A-8), Q is the flow rate in The power input (bhp) in U.S. customary units is
m3 =s, H is the total dynamic head in m, q is the flow
rate in L/s, and 102 is a conversion factor for water at gQH qH whp
15 to 208C. bhp ¼ ¼ ¼ (10-7b)
550Ep 3960Ep Ep
In U.S. customary units, power output is defined as
Pump efficiencies usually range from 40 to 92% and
gQH WH qH increase with the size of the pump (see Figure 10-4).
P¼ ¼ ¼ (10-6b)
550 33,000 3960 Energy losses in a pump are volumetric, mechanical,

Figure 10-4. Maximum pump efficiency attainable at the best efficiency point and at optimal specific speed.
10.6 Chapter 10 Performance of Centrifugal Pumps

and hydraulic. Volumetric losses are those of leakage in the stuffing boxes and bearings, by internal disc
through the small clearances between wearing friction, and by fluid shear. Frictional and eddy
rings in the pump casing and the rotating element. losses within the flow passages account for the hy-
Mechanical losses are caused by mechanical friction draulic losses.

Example 10-1
Evaluation of Pump Performance

Problem: Water is pumped by a radial-flow centrifugal pump at 208C (688F) through a new
cement-mortar-lined ductile iron piping (DIP) system such as shown in Figure 10-3, except that
there are two 90-degree elbows and a gate valve in the suction piping, and a gate valve and
three 90-degree elbows in the discharge piping. Other data known, measured, or computed are
. Flow rate, Q ¼ 0:1 m3 =s (3:54 ft3 =s)
. Total static head, Hstat ¼ 14 m (45:93 ft)
. Static suction lift, hs ¼ 2 m (6:56 ft)
. Power input to pump shaft, P ¼ 22:8 kW (30:5 hp)
. Suction piping ¼ 3.3 m of 300-mm (11 ft of 12-in.) pipe
. Discharge piping ¼ 47.6 m of 250-mm (156 ft of 10-in.) pipe.
Evaluate: (1) the total dynamic head, (2) the power delivered to the water, and (3) the efficiency
of the pumping unit.
Solution: Because pipe is exposed and joints are flanged within the pumping station, Class 53 is
needed (see Tables B-1 and B-2 for dimensions). Outside, the pipe is buried, joints can be
mechanical or push on, and the pipe can be Class 50 with a net ID of 264 mm (10.4 in.), which
includes the lining (see the DIPRA handbook [3]). Because most of the discharge piping can be
Class 50, ignore the slightly smaller diameter of the 250-mm (10-in.) Class 53 exposed piping
within the pumping station.
Total dynamic head: Find velocity, velocity head, and friction losses in suction and
discharge piping using Equation 10-4.

SI Units U.S. Customary Units

For the suction piping:
ID ¼ 312 mm (Table B-1) ID ¼ 123 in. (Table B-2)
As ¼ 0:0763 m2 (Table B-1) As ¼ 0:822 ft2 (Table B-2)
vs ¼ 1:31 m=s vs ¼ 4:31 ft=s
v2 =2g ¼ 0:0875 m v2 =2g ¼ 0:288 ft
Find the entrance loss assuming a bell mouth entrance is used (see Table B-6).
h ¼ 0:05  0:0875 ¼ 0:004 m h ¼ 0:05  0:288 ¼ 0:014 ft
Find the friction headloss using Equations 3-9a and 3-9b.
hfs ¼ 10,700 (Q=C)1:85 D4:87 hfs ¼ 10,500 (Q=C)1:85 D4:87
hfs ¼ 10,700 (0:1=145)1:85  (0:3121)4:87 hfs ¼ 10,500 (1590=145)1:85  (12:3)4:87
hfs ¼ 4:41 m=1000 m hfs ¼ 4:34 ft=1000 ft
hfs ¼ 4:41 (3:3 m=1000 m) ¼ 0:015 m hfs ¼ 4:34 (11 ft=1000 ft) ¼ 0:048 ft
The ‘‘minor’’ losses are found using Equation 3-16 (h ¼ Kv2 =2g). Note that there is an elbow as
well as a gate valve in the suction piping. Obtain K values from Tables B-6 and B-7.
h (bend) ¼ 0:25  0:0875 ¼ 0:022 m h (bend) ¼ 0:25  0:288 ¼ 0:072 ft
h (gate) ¼ 0:2  0:0875 ¼ 0:018 m h (gate) ¼ 0:2  0:288 ¼ 0:058 ft
Shfvs ¼ 0:040 m Shfvs ¼ 0:130 ft
 10-2. Pump Application Terminology, Equations, and Performance Curves 10.7

For the discharge piping:

ID ¼ 0:264 m ID ¼ 10:4 in.
Ad ¼ 0:0548 m 2
Ad ¼ 0:589 ft2
vd ¼ 1:82 m=s vd ¼ 6:01 ft=s
v2d =2g ¼ 0:169 m v2d =2g ¼ 0:561 ft
Find the friction headloss from Equations 3-9a and 3-9b.
hfd ¼ 10,700 (0:1=145)1:85 (0:264)4:87 hfd ¼ 10,500 (1590=145)1:85 (10:4)4:87
¼ 9:95 m=1000 m ¼ 9:82 ft=1000 ft
hfd ¼ 9:95 (47:6 m=1000 m) ¼ 0:474 m hfd ¼ 9:82 (156 ft=1000 ft) ¼ 1:53 ft
Find the ‘‘minor’’ losses using Equation 3-16.
h (check) ¼ 2:20  0:169 ¼ 0:372 m h (check) ¼ 2:20  0:561 ¼ 1:23 ft
h (gate) ¼ 0:20  0:169 ¼ 0:034 h (gate) ¼ 0:20  0:561 ¼ 0:11
h (bends) ¼ 2 (0:25  0:169) ¼ 0:084 h (bends) ¼ 2 (0:25  0:561) ¼ 0:28
Shfvd ¼ 0:490 Shfvd ¼ 1:62
The TDH computed using Equation 10-4 is
H stat ¼ 14:00 m H stat ¼ 45:93 ft
hent ¼ 0:004 hent ¼ 0:014
hfs ¼ 0:015 hfs ¼ 0:048
hfd ¼ 0:474 hfd ¼ 1:53
Shfvs ¼ 0:040 Shfvs ¼ 0:130
Shfvd ¼ 0:490 Shfvd ¼ 1:62
v2d =2g ¼ 0:169 v2d =2g ¼ 0:561
TDH ¼ 15:192 m TDH ¼ 49:833 ft
Use 15.2 m Use 49.9 ft
The slight discrepancies between SI and U.S. values are caused by rounding off coefficients
and powers to three significant figures. All of the terms in Equation 10-4 are included in these
additions. No allowance was made for the proximity of fittings or for swirling (see Section 3-4).
The friction loss in the discharge piping is 0.474 m (1.53 ft), but only about 10% of it would be
within the pumping station. The total pipe friction loss within the pumping station, therefore,
would be about (0.1) (0.474) þ 0.015 ¼ 0.062 m (0.2 ft), whereas all the ‘‘minor’’ losses (due to
fittings and valves) total about 0.53 m (1.75 ft)—almost 10 times as much.
Power delivered to the water. Use Equations 10-6a and 10-6b.

P ¼ gQH P ¼ gQH=550

(62:32 lb=ft3 )(3:54 ft3 =s)(49:9 ft)

¼ (9:789 kN=m3 )(0:1 m3 =s)(15:2 m) ¼
½(550 ft  lb)=s=hp
¼ 14:9 wkW ¼ 20:0 whp

Efficiency of the pumping unit. Use Equations 10-7a and 10-7b.

wkW whp
Ep ¼ Ep ¼
bkW bhp
14:9 20:0
Ep ¼ ¼ 65% Ep ¼ ¼ 65%
22:8 30:5
10.8 Chapter 10 Performance of Centrifugal Pumps

Pump Performance Curves 10-3. Pump Operating Characteristics

The head that a pump can produce at various flow The operating characteristics of pumps depend on
rates and rotational speeds is established in pump their size, speed, and design. Pumps of similar size
tests conducted by the pump manufacturer. During and design are produced by many manufacturers, but
testing, the capacity of the pump is varied by throt- they vary somewhat because of slight design modifi-
tling a valve in the discharge pipe, and the corre- cations. The basic relationships that can be used to
sponding head is measured. The results of these characterize and analyze pump performance under
tests and other tests with different impeller diameters varying conditions include
are plotted to obtain a series of head-capacity (H-Q)
. Energy transfer in pumps
curves for the pump at some given speed (see Figure
. Flow, head, and power coefficients
10-5). Simultaneously, the power input to the pump is
. Affinity laws
measured. The efficiency at various operating points
. Specific speed
is computed, and these values are also plotted in the
same diagram. Together, these curves are known as
‘‘pump characteristic curves.’’ However, pumps are Energy Transfer in Radial Centrifugal Pumps
not typically capable of operating continuously or for
protracted periods at all positions along their charac- The transfer of energy in centrifugal pumps is based
teristic curves. Severe damage can result from con- on the momentum principle, which, as applied to
tinuous operation too close to shut-off or too far to centrifugal pumps, can be stated as follows: ‘‘A net
the right of the best efficiency point (BEP). In accord- torque applied to a pump impeller causes a change in
ance with a standard adopted by the Hydraulic Insti- the angular momentum of the fluid.’’ It is this change
tute in 1997 (ANSI/HI 9.6.3), pump manufacturers in angular momentum that develops head or pressure
will indicate the limits of acceptable operation on as the fluid passes through a pump impeller.
pump characteristic curves. Refer to Section 10-6 In Figure 10-6, the net torque, T, applied to the
for more information. pump impeller equals the difference between the

Figure 10-5. Typical pump characteristic curve. Double suction impeller.

 10-3. Pump Operating Characteristics 10.9

u2 = r2ω

V1 V1
α1
u1,t Impeller inlet
ω, T tip speed

Figure 10-6. Velocity diagrams for a radial-flow pump impeller. V, absolute velocity; n, relative velocity;
u, peripheral velocity; a1 , angle to V1 at the hub; a2 , angle to V2 at the periphery; b, angle to n at the periphery;
v, rotational velocity; T, torque.

moment of momentum at the inlet and outlet of the be seen that the theoretical head developed by a
impeller. For ideal conditions, the net torque is given pump depends only on the increase of the moment
by of velocity within the impeller and is independent of
specific weight and viscosity of the fluid (see the
g literature [2, 4–10] for a more complete discussion of
T ¼ Q (r2 V2 cos a2  r1 V1 cos a1 ) (10-8)
g pump theory). The actual total dynamic head, TDH,
or simply H of the pump is less than the theoretical
To obtain the power, the torque is multiplied by the head, Hth ; by the amount equal to the head losses of
angular velocity. the fluid as it flows within the pump. Therefore, H is
the product of the theoretical head Hth and the pump
P ¼ Tv ¼ gQHth (10-9) hydraulic efficiency Ep .

where P is the shaft power, T is the torque, v is the

Flow, Head, and Power Coefficients
angular velocity (rotational speed) in radians
per second, and Hth is the theoretical head.
In centrifugal pumps, similar flow patterns occur in a
The theoretical head, Hth , added to the fluid by
series of geometrically similar pumps. By applying
the pump can be obtained by equating T in Equations
the principles of dimensional analysis, the following
10-8 and 10-9
three independent dimensionless groups can be de-
rived to describe the operation of centrifugal pumps
g gQHth
Q (r2 V2 cos a2  r1 V1 cos a1 ) ¼ and other rotodynamic machines. Note that the equa-
g v tions are the same in either SI or U.S. customary
units. Only the value of the coefficient changes.
Substituting u=v for r
Q
CQ ¼ (10-12)
(u2 V2 cos a2  u1 V1 cos a1 ) nD3
Hth ¼ (10-10)
g
H
CH ¼ (10-13a)
n2 D2
For radial flow, cos u1 ¼ 0 and
or, for dimensional correctness,
u2 V2 cos a2
Hth ¼ (10-11) gH
g CH ¼ (10-13b)
n2 D2
A similar set of equations can be derived for mixed- P
CP ¼ 3 5 (10-14)
and axial-flow pumps. From Equation 10-10, it can rn D
10.10 Chapter 10 Performance of Centrifugal Pumps

where C is a coefficient and the subscripts Q, H, and  3

P1 n1
P correspond to capacity, head, and power. Q is flow ¼ (10-17)
P2 n2
rate in m3 =s (ft3 =s), n is speed in radians per second
(revolutions per minute or rev/min), D is impeller
where Q is flow rate, H is head, P is power, n is rota-
diameter in m (ft), H is head in m (ft), g is
tional speed, and subscripts 1 and 2 are only for corre-
9:81 m=s2 (32:2 ft=s2 ), P is the power input in kilo-
sponding points. Note that Equations 10-15 and 10-16
watts or kW (horsepower or hp), and r is the density
must be applied simultaneously to ensure that Point 1
in kilograms per cubic meter (slugs per cubic foot).
‘‘corresponds’’ to or is ‘‘homologous’’ with Point 2. Cor-
Although Equation 10-13b is dimensionally correct,
responding points fall upon parabolas through the origin.
Equation 10-13a is commonly used in the United
They do not fall upon system H-Q curves. These rela-
States for both SI and U.S. customary units.
tionships, known collectively as ‘‘the affinity laws,’’ are
Equations 10-12 through 10-14 apply only to ‘‘cor-
used to determine the effect of changes in speed on the
responding points,’’ the operating points at which
capacity, head, and power of a pump.
similar flow patterns occur. Thus, every point on a
The affinity laws for flow rate and head are accur-
head-capacity curve for a large pump corresponds to
ate as found from actual tests for all types of centrifu-
a point on the head-capacity curve of a geometrically
gal pumps, including axial-flow pumps. The affinity
similar but smaller pump that operates at the same
law for power is not as accurate because efficiency
speed. This correspondence can be corrected for differ-
increases with an increase in the size of the pump. If
ent speeds provided that the speeds differ by no more
the affinity law for power is used, the computed value
than about 40%.
of power can be corrected using an equation of the
type proposed by Moody [2, 7, 8], which accounts for
efficiency as a function of pump size, but it is often
Affinity Laws
necessary to modify that equation (based on experi-
ence) as recommended by the pump manufacturer.
For a pump operating at two different speeds, the
In applying these relationships, remember that
following relationships can be derived from Equa-
they are based on the assumption that the efficiency
tions 10-12 through 10-14. (Note that neither dia-
remains the same when transferring from a given
meter nor density changes.)
point on one pump curve to a homologous point on
another curve. Because the hydraulic and pressure
Q1 n1
¼ (10-15) characteristics at the inlet, at the outlet, and through
Q2 n2 the pump vary with the flow rate, the errors produced
 2 by Equation 10-17 may be excessive, although errors
H1 n1
¼ (10-16) produced by Equations 10-15 and 10-16 are very
H2 n2 small. The application of these relationships is illus-
trated in Example 10-2 and in Section 15-3.

Example 10-2
Application of Affinity Laws

Problem: A pump with a normal operating speed of 705 rev/min and a 356-mm (14-in.)
diameter impeller was tested at rotational speeds of 705, 625, 550, 450, and 350 rev/min. Using
the head and capacity data collected at 705 rev/min, generate pump curves for rotational
speeds of 625, 550, 450, and 350 rev/min and compare the curves generated with the measured
data points.

SI Units U.S. Customary Units

Flow rate Head Flow rate Head
(m3 =h) (m) (gal/min) (ft)

705 rev/min
1596 5.91 7025 19.4
1361 7.28 5994 23.9
1140 8.66 5020 28.4
 10-3. Pump Operating Characteristics 10.11

Flow rate Head Flow rate Head

(m3 =h) (m) (gal/min) (ft)

1002 9.39 4410 30.8

866 10.21 3550 33.5
363 11.28 1600 37.0
0 13.11 0 43.0
625 rev/min
1420 4.57 6250 15.0
1136 6.25 5000 20.5
908 7.32 4000 24.0
681 7.92 3000 26.0
454 8.63 2000 28.3
227 9.45 1000 31.0
0 10.36 0 34.0
550 rev/min
1249 3.51 5500 11.5
943 5.12 4150 16.8
772 5.82 3400 19.1
454 6.55 2000 21.5
363 6.80 1600 22.3
0 7.92 0 26.0
450 rev/min
1022 2.44 4500 8.0
795 3.35 3500 11.0
568 3.96 2500 13.0
227 4.72 1000 15.5
0 5.43 0 17.8
350 rev/min
568 3.13 2500 7.0
472 2.41 2080 7.9
363 2.59 1600 8.5
0 3.05 0 10.0

Solution: Plot the pump head-capacity (H-Q) data obtained at 705 rev/min (shown as solid
circles in Figure 10-7) and draw a smooth pump curve (shown as a solid line) through the
plotted points.
Develop H-Q curves for the other rotational speeds from the smooth pump curve by
applying the affinity laws for flow rate and head simultaneously. For example, the flow rate
and head values at the point on the 625-rev/min curve that corresponds to point a on the
705-rev/min pump curve, with coordinate values of 1250 m3 =h and 8.10 m (5504 gal/min and
26.6 ft), are determined as follows.
Using Equation 10-15
Q2 ¼ Q1 =(n1 =n2 ) ¼ 1250=(705=625) ¼ 1108 m3 =h
Using Equation 10-16

H2 ¼ H1 =(n1 =n2 )2 ¼ 8:10=(705=625)2 ¼ 6:37 m

So, point b with coordinates of 1108 m3 =h and 6.37 m (4875 gal/min and 20.9 ft) on the
625-rev/min curve corresponds to point a on the 705-rev/min curve. Enough other corresponding
points are computed in the same manner to allow the entire curve to be drawn for 625 rev/min.
The process is repeated for curves at 550, 450, and 350 rev/min, which pass through points c, d,
and e, respectively. The curves are shown as dashed lines to indicate that they are derived
mathematically from the curve at 705 rev/min. The measured values for head and capacity are
plotted in Figure 10-7 as open circles. The correspondence between measured values and
computed pump curves is excellent.
10.12 Chapter 10 Performance of Centrifugal Pumps

Figure 10-7. Pump curves developed from the affinity laws compared with test data.

Points f, g, h, i, and j in Figure 10-7 are unique. Because the flow is zero at these points,
Equation 10-15 can be ignored and Equation 10-16 can be used alone. However, the only time
that Equations 10-15 and 10-16 need not be solved simultaneously is when the flow is zero. The
errors in the four points derived from point f are 0.6, 0.8, 1.7, and 5.90%, respectively. In
variable-speed pumping, pumps rarely operate at less than 60% of full speed, so in practice the
errors in using the affinity laws for calculating head and flow would rarely exceed 2 or 3%.
Points a, b, c, d, and e lie on a parabola that passes through the origin. The equation for the
affinity parabola can be found by solving Equations 10-15 and 10-16 simultaneously to
eliminate n.

 2  2
Q1 n1 H1
¼ ¼
Q2 n2 H2

So H1 ¼ H2 (Q1 =Q2 )2 , which is a parabola through point O. Hence, all corresponding points
lie on parabolas that pass through point O, the origin. Station H-Q curves are also parabolas,
but because such parabolas do not pass through the origin (unless the static head is zero), it
follows that corresponding points cannot lie on a station H-Q curve. Expressed in the simplest
terms, the foregoing statement means that the flow from a pumping station (with a static lift) is
not proportional to the speed of the pump—a conclusion that might appear, superficially, to be
at odds with Equation 10-15.
 10-3. Pump Operating Characteristics 10.13

Approximate Relationships for Equation 10-19 applies only to two pumps of differ-
Radial-Flow Pumps ent size (i.e., the same scale-up in three dimensions).
For a single pump, the discharge area for a radial-
To cover a wide range of flows with a minimum flow impeller is a function of impeller width and
number of pump casings and impeller designs, manu- volute size, and both are nearly constant even if the
facturers customarily offer a range of impeller diam- impeller diameter is reduced. Because area is propor-
eters for each size of casing (see Figure 10-5). In tional to diameter squared, Equation 10-19 can be
general, radial-flow impellers are identical as cast, converted to an equation for a single pump by factor-
but the size is reduced by machining the impeller to ing out (D1 =D2 )2 , which yields
a smaller diameter. Equations 10-12 through 10-14,
written for pumps of different sizes, can be modified Q1 D1
into equations for impellers of reduced diameter in ¼ (10-20)
Q2 D2
the same pump. Dividing Equation 10-13 for an im-
peller of diameter D1 by the same equation for an Power is a function of head times discharge, so multi-
impeller of diameter D2 and canceling CH and n2 plying Equation 10-18 and Equation 10-20 gives
(because density, viscosity, and rotational speed are
the same) gives  5
P1 D1
¼ (10-21)
 2 P2 D2
H1 D1
¼ (10-18)
H2 D2
These relationships, when used to predict the effect of
changing the diameter of a radial-flow impeller, are
Although the performance of many pumps can be somewhat less accurate than the affinity laws because
modeled with this relationship with small errors, sig- the angle of the blade decreases slightly and the clear-
nificant deviations can arise, because CH may not ance between impeller and casing increases (which
remain constant (as assumed here) when only D is changes the geometry) as the diameter is reduced
reduced without reducing all other dimensions in the (see Figure 10-6). Consult the pump manufacturer
same proportion, as would be required to maintain before applying the modified affinity laws to more
strict geometrical similarity. than a 5% change in impeller diameter.
The same analysis applied to Equation 10-12 gives It should be noted that the modified affinity
laws (Equations 10-18 through 10-21) are less accur-
 3
Q1 D1 ate for trimming impellers of mixed- and axial-flow
¼ (10-19) pumps.
Q2 D2

Example 10-3
Effect of Changes in Impeller Diameter and Speed

Problem: Data from Figure 10-5 for a pump with an impeller diameter of 0.4463 m (17.57 in.)
operating at 1170 rev/min are tabulated as follows.

SI Units U.S. Customary Units

Point Flow rate Head (m) Point Flow rate Head (ft)
(m3 =h) (gal/min)

a 0 47.6 a 0 156
b 100 46.3 b 440 152
c 200 45.3 c 880 149
d 300 44.0 d 1320 144
e 400 41.0 e 1760 135
f 500 36.5 f 2200 120
g 600 29.8 g 2640 98
10.14 Chapter 10 Performance of Centrifugal Pumps

(1) Develop a new head-capacity curve for the same pump fitted with a new impeller 0.3810 m
(15 in.) in diameter. (2) Compare the results with the test curve for the new impeller shown in
Figure 10-5. (3) Find the increase in rotational speed that would be required for the new
impeller to match the performance of the original impeller at 1170 rev/min.
Solution:
H-Q for the smaller impeller. Compute this using Equations 10-20 [Q2 ¼ Q1 (D2 =D1 )] and
10-18 [H2 ¼ H1 (D2 =D1 )2 ]. Sample calculations for point d are
Q2 ¼ 300 (0:3810=0:4463) ¼ 256 m3 =h Q2 ¼ 1320 (15=17:57) ¼ 1130 gal=min
H2 ¼ 44:0 (0:3810=0:4463)2 ¼ 32:1 m H2 ¼ 144 (15=17:57)2 ¼ 105 ft
Compare the results to Figure 10-5. The computed values of points a to g for the new 15.00-in.
impeller diameter are summarized below.

SI Units U.S. Customary Units

Head (m) Head (ft)

Flow rate Flow rate

Point (m3/h) Computed Test Point (gal/min) Computed Test

a’ 0 34.4 34.4 a’ 0 113 113

b’ 85 33.8 34.4 b’ 376 111 113
c’ 171 32.9 33.2 c’ 751 108 109
d’ 256 32.0 31.7 d’ 1130 105 104
e’ 341 29.9 29.0 e’ 1500 98 95
f’ 427 26.5 24.7 f’ 1880 87 81
g’ 512 21.3 — g’ 2250 70 —

Compare the computed with the tested heads in the last two columns or compare calculated
points a’ to g’ in Figure 10-5 with the 15.00 curve. Minor differences (up to 0.6 m or 2 ft) are
due to errors in plotting and scaling. Larger differences are due to the fact that actual losses in
the pump are not considered in Equations 10-18 and 10-20. pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Rotational speed for the trimmed impeller. Use Equation 10-16 (n2 ¼ n1 H 2 =H 1 ) at zero
flow to estimate the new rotational speed required to obtain the original flow rate with the
smaller impeller.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n2 ¼ 1170 47:6=34:4 ¼ 1376 rev=min n2 ¼ 1170 156=113 ¼ 1375 rev=min

Specific Speed dividing the square root of Equation 10-12 by the

three-fourths power of Equation 10-13a:
Although the desired discharge or flow rate has the
1=2
major impact on the size and layout of a pumping CQ (Q=nD)3 nQ1=2
station through the sizes of the pipes and the intake nq ¼ ¼ ¼ (10-22a)
3=4
CH (Ht =n2 D2 )3=4 Ht
3=4
and discharge bays, the geometry of each pump
must also be known for establishing the detailed con- where ns is the specific speed (also called the ‘‘type
figuration of the station. The shapes of the impeller and number’’ in Europe), n is in revolutions per minute
adjacent stationary flow elements of a pump are deter- (not radians per second), Q is flow rate (strictly m3 =s
mined by the ‘‘specific speed,’’ ns (Note: The term ns is but often m3 =h), and Ht is the total dynamic head in
used generically without regard to units.), which is a m. In spite of being dimensionally incorrect, it is this
combination of the pump speed, flow rate, and the form of the equation that is used in the United States.
total developed head. The SI value of ns is denoted as (A dimensionally correct expression would be derived
ng, whereas Ns is the symbol used for U.S customary by using Equation 10-13b instead of 10-13a.)
units.
For a geometrically similar series of pumps oper-
nQ1=2
ating under similar conditions, the diameter term in Ns ¼ 3=4
(10-22b)
Equations 10-12 and 10-13a can be eliminated by Ht
 10-3. Pump Operating Characteristics 10.15

In U.S. customary units, Ns is the specific speed, n is Equation 10-22, because Q and H are specified ini-
in rev/min, Q is flow rate in gal/min, and Ht (or tially. Of course, there could be more than a single
simply H) is the total dynamic head in ft. The relation pump, and a high head requirement could dictate the
between specific speeds for various units of flow rate need for a multistage pump or pumps in series to
and head is given in Table 10-1, wherein the numbers optimize pump efficiency and cost.
in bold type are those customarily used. For example, The variations in maximum efficiency to be
if the specific speed nq is expressed in metric units expected with variations in size, flow rate, and spe-
(e.g., Q ¼ m3 =h), the corresponding specific speed in cific speed are shown in Figure 10-8 for single-stage,
U.S. customary units (e.g. Q ¼ gal/min), Ns, is single-entry centrifugal pumps. The progressive
obtained by multiplying the metric value by 0.861. changes in impeller shape as the specific speed in-
For any pump operating at any given speed, Q creases are shown along the bottom of Figure 10-8.
and H must be taken at the point of maximum effi- The efficiency increases with the size of the pump
ciency. When using Equation 10-22 for pumps with because of a reduction in the friction losses due to
double suction impellers, one-half of the flow rate is the lesser relative roughness of the internal surfaces
used unless otherwise noted. However, the full flow and a corresponding drop in the volumetric losses.
rate is widely used in the United States. For multi- The greatest efficiency is achieved in single-stage,
stage pumps, the head is the head per stage. single-entry pumps with a volute casing [5, 6]. In
The pump speed, usually selected by the pump general, for a given rotational speed,
manufacturer, is based on a consideration of (1) the
available speeds of the drive motor or engine; (2)
critical speeds that might cause structural resonance; . If the nq value is less than 30 ðNs < 1500), Q must
and (3) the limit imposed by the suction capability of be low and H must be high.
the pump. The suction limit on the speed is computed . If the nq value is 30 to 80 (Ns = 1500 to 4000), Q
from the ‘‘suction specific speed’’ nss as follows: and H must be of intermediate value.
. If the nq value is more than 80 (Ns > 4000), Q must
be high and H must be low.
(NPSHA)3=4
N ¼ nss  pffiffiffiffi (10-23)
Q
In summary, the significance of specific speed is
apparent from Figure 10-8. In particular, a high-ns
where N is the pump speed in rev/min, and nss is axial-flow pump has pipes as large as the impeller (or
characteristic of the pump and ranges from 150 to propeller), which feed the fluid in axially and usually
250 in SI units (8,000 to 13,000 in U.S. customary conduct it away in a radial direction through a sweep-
units), NPSHA is the available ‘‘net positive suction ing, gentle, 90-degree bend downstream from the
head’’ in m (SI) or ft (U.S. customary units) and is impeller and any flow-straightening vanes. The low-
defined in Section 10-4, Q is the flow rate in m3 =s (SI ns radial-flow pump, on the other hand, has smaller
units) or gal/min (U.S. customary units). Once N is pipes and a volute collector surrounding the impeller
known, the specific speed ns can be computed from with a tangential or radial conical diffuser exiting to

Table 10-1. Factors for Converting Values of Specific Speed Expressed in One Set of Units to the Corres-
ponding Values in Another Set of Units. Customary units are gal/min in the United States and m3 =s in Europe.
Some Europeans use m3 =h.

ns rev/min rev/min rev/min rev/min rev/min rev/min

Q L=s m3 =s m3 /min m3 =h gal=min ft3 =s
Ht m m m m ft ft

1 0.0316 0.2449 1.897 1.633 0.0771

31.623 1 7.746 60 51.628 2.437
4.082 0.1291 1 7.746 6.665 0.3146
0.5270 0.0167 0.1291 1 0.860 0.0406
0.6123 0.0194 0.1500 1.162 1 0.0472
12.975 0.4103 3.178 24.618 21.183 1
10.16 Chapter 10 Performance of Centrifugal Pumps

Figure 10-8. Pump efficiency as related to specific speed and discharge. After Flowserve Corp.

the downstream piping system. There are other con- will occur—and to what extent—depends on several
figurations, as is apparent in Example 10-4. factors, the primary ones being these:
. The Net Positive Suction Head (NPSH) sup-
10-4. Cavitation plied at the pump inlet, (1) and (2)
. Off-Design Operation [i.e., at flow rates (Q)
Cavitation is the formation and collapse of vapor through the pump that are different from
cavities within the pump. When cavitation occurs in its BEC (design flow rate or capacity, (1) and (2)]
pumps, it has the potential to cause (1) performance . The ‘‘Suction Energy’’ or ‘‘Energy Level’’ at
degradation, namely, loss of head, capacity, and effi- which the pump is operated, (2) only
ciency and (2) permanent damage due to erosion and How these factors influence the adverse effects (1)
mechanical failure of pump components and struc- and (2) is explained in the following subsections (see
tures. Whether these adverse effects [(1) and/or (2)] also Cooper [11]).

Example 10-4
Determination of Pump Speed and Configuration

Problem: A flow of 1 m3 =s (15,850 gal/min) must be pumped against a head of 25 m (82 ft).
The available net positive suction head is 6.1 m (20 ft). Assuming a single stage and the suction
specific speed nss to be 200 (10,330 in U.S. customary units), estimate the pump rotational
 10-4. Cavitation 10.17

speed (rev/min) and discuss the corresponding pump type and configuration. Also estimate the
pump efficiency.
Solution: Estimate the pump rev/min using Equation 10-23 as follows:

SI Units U.S. Customary Units

(6:1)3=4 (20)3=4
n ¼ 200  pffiffiffi ¼ 776 rev=min n ¼ 10,330  pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 776 rev=min
1 15,850

However, if this pump is driven by an electric motor at 50 Hz (or 60 Hz), the nearest
synchronous speed is 750 rev/min (or 720 rev/min); so the actual pump speed would probably
be reduced accordingly from this estimate of 776 rev/min. Nevertheless, continuing with the
776-rev/min estimate, the specific speed is found as follows from Equations 10-22a and 10-22b.

pffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
776  1 776  15,850
nq ¼ ¼ 69:4 Ns ¼ ¼ 3,585
(25)3=4 (82)3=4

As illustrated in Figure 10-8, this is a mixed-flow pump with a radial-discharge, overhung

impeller and would achieve as much as 91% efficiency. The pump would most likely have an
inlet pipe feeding axially into the impeller, a volute collector around the periphery of the
impeller, and a tangential conical diffuser exiting to the downstream piping.
Alternatively (not shown in Figure 10-8), this pump could have a double-suction impeller
hung between bearings with a specific speed nq ¼ 49:1 (Ns ¼ 2535) per side, which is based on
one-half the total flow or 0:5 m3 =s (7925 gal/min). The efficiency could be as high as 92%;
although one inlet passage and one volute handling both half-flows makes this pump
something like the higher-ns one above, which would not exceed 91% efficiency. The double-
suction impeller would be essentially two impellers of the third type from the left in Figure 10-8
joined back-to-back and fed by a tangential inlet passage that splits symmetrically to feed flow
to both sides of the impeller. The discharge pipe would come off the volute tangentially and in
the same direction, making this an ‘‘in-line’’ pump, which is commonly used in pipelines and
which is illustrated in Figures 10-2 and 10-3.

Occurrence of Cavitation and Its Effects edge is illustrated in Figure 10-9a. The difference
between p1 and pv is called NPSH when expressed
As fluid flows from the pump inlet port and into the as head and augmented by the pump inlet velocity
impeller, it undergoes a drop in pressure because it head V12 =2g. As the NPSH becomes lower, the vol-
accelerates and because the pump blades that drive the ume of bubbles increases and so does the likelihood
fluid have higher-than-average pressure on their driving that these cavities will distort the liquid flow lines as
faces and, therefore, lower-than-average pressure on illustrated in Figure 10-9a. Sufficient distortion, in
their trailing faces. If the entering liquid contains dis- turn, causes performance degradation. If NPSH gets
solved air, tiny bubbles of this air come out of solution low enough, the impeller becomes filled with vapor
as the pressure drops. If the pump inlet absolute pres- (or vapor-locked) as shown in Figure 10-9b, and the
sure, p1 , continues to drop and zones develop within the pump can no longer generate any pressure rise or
pump wherein the absolute pressure equals the vapor liquid head, TDH. Many of the bubbles occurring
pressure, pv ; of the liquid. The liquid now boils, which in partial cavitation collapse on the blade surfaces
causes an explosive increase in the number and size of and create the erosion seen in Figure 10-10a and b.
bubbles (cavities), and the liquid’s own vapor then be- Such erosion is less severe in full cavitation, wherein
comes the main constituent of these cavities. pumping ceases.
Degradation of the performance is illustrated viv-
Performance Degradation and NPSH idly in Figure 10-11, where three H-Q performance
curves are shown, each set corresponding to a given
The occurrence of cavitation on the trailing face or suction lift. If the suction lift increases, the absolute
suction side of a pump impeller blade near its leading pressure, p1 , at the pump suction port and the NPSH
10.18 Chapter 10 Performance of Centrifugal Pumps

Figure 10-9. Formation of vapor bubbles in a pump impeller. (a) Partial cavitation; (b) full cavitation. After Addison
[4].

decrease. It is at the lowest NPSH that the greatest flow rate where cavitation is fully formed, as in Fig-
degradation in performance is to be expected, and ure 10-9b. For pumps with specific speeds nq between
that is exactly what is portrayed in this figure. These 30 and 80 (Ns between 1500 and 4000), the pump
curves are typical of centrifugal pumps of specific performance curves fall more gradually until the cut-
speeds, nq , less than 30 (Ns < 1500). Here, there is a off flow rate is reached. For pumps with specific
sharp drop or cutoff in the H-Q curves. Vapor bub- speeds nq greater than 80 (Ns > 4000), there is no
bles begin to form at a lower flow rate than the cutoff distinct cutoff point.

Figure 10-10. Cavitation damage in a vertical pump. (a) Impeller; (b) suction bell.
 10-4. Cavitation 10.19

150 occur unless pc is large enough and the material of

construction weak enough. The value of pc increases
with the suction energy of the pump. The suction
energy, in turn, can be expressed as the inlet-tip-
Pump H-Q curve speed head u 21,t =2g of the pump blade tip speed, u1,t ,
at the impeller inlet, namely the linear speed of the
100 blade at its maximum diameter, as seen in Figure 10-
11 and defined in Figure 10-6. Water pumps with
Head, m

u1 ,t2 =2g less than 30 m (100 ft) usually experience

no erosion due to cavitation. The inlet-tip-speed
8-m Suction lift
head is about one-quarter of the head TDH of a
50 centrifugal pump, and because the TDH is readily
6-m Suction lift
known, it is easier to remember that a centrifugal
pump with less than 120 m (400 ft) of head should
zero Suction lift
usually experience no erosion due to cavitation. How-
ever, for a mixed- or axial-flow pump, such as that
0 shown in Figure 10-10, the TDH can be substantially
0 20 40 60 80 100 less than the inlet-tip-speed head, so that it is not
Flow rate, m3/h unreasonable to expect erosion damage at a TDH of
Figure 10-11. Pump performance curves at various suc- less than 30 m (100 ft) in an axial-flow pump.
tion lifts with cavitation. Note how rapidly the cavitation
causes the head to drop.
Energy Level, TDH, and Cavitation Erosion Rate

Cavitation Damage The TDH is also a measure of the energy level of a

pump stage; and, as with suction energy, if the pump
As the cavities illustrated in Figure 10-9a flow down- is operated above a critical energy level, cavitation
stream into the higher-pressure zone of the impeller, erosion will occur. The TDH for this critical energy
they can no longer exist (i.e., the vapor in them must level falls with increasing specific speed, ns , the type
condense). This condensation occurs quite rapidly, of impeller changing from centrifugal (or radial-out-
and the bubbles collapse as soon as they are exposed flow) to axial-flow as ns increases, as illustrated in
to an absolute pressure, p, that exceeds the vapor Figure 10-8. From the above discussion, it is apparent
pressure, pv . The instantaneous collapse pressure, pc , that high-ns pumps exhibit more cavitation erosion at
can be immense; in fact, it can be great and frequent lower TDH than do low-ns pumps [1]. (Note: As
enough to cause fatigue failure of the material of the defined earlier in this chapter, the SI-value of ns is
blade or other surface next to which repeated collaps- denoted as nq , whereas Ns is the symbol used for U.S.
ing of successive bubbles at the same site occurs. The customary units of ns .)
resulting pitting or erosion is illustrated in Figure 10- The consequences of operating a pump above the
10a, seen as expected from Figure 10-9a on the suc- critical energy level are severe. If a pump is near the
tion (visible) sides of the blades downstream of the critical energy level at a given rev/min, it will experi-
leading edges. The damage shown in Figure 10-10b is ence cavitation erosion if the rev/min is increased. In
caused by recirculation (and backward flow) in a a high-energy pump, the rate of penetration of cavi-
pump that operated too far to the left of the BEP. tation erosion into the blades or other affected sur-
In addition to pitting or erosion, the collapsing faces increases approximately as the sixth power of
vapor cavities can also cause noise and vibration. the blade inlet tip speed u1, t , and it greatly shortens
Noise is produced by the implosions of the collapse the life of the pump [1].
phenomenon, and the vibration is due to the imbal-
ance and surging caused by the uneven distribution of
collapsing vapor bubbles. Off-Design Operation

If the pump is operated at flow rates different from the

Erosive Damage and Suction Energy BEC, the locations of the cavities change. As Q is
increased above the BEC, the low-pressure zones even-
Even though there may be extensive cavitation with tually switch to the opposite sides of the pump
many bubbles collapsing, erosive damage does not blades from those shown in Figure 10-9a, and the
10.20 Chapter 10 Performance of Centrifugal Pumps

resulting performance degradation for a given NPSH the NPSH is progressively reduced to the point of
is greater. complete loss of this liquid head or pump pressure
When Q is much less than the BEC, the impeller is rise. The test described above is an expanded version
over-size for that low value of Q and it ejects highly of the pump manufacturer’s ‘‘NPSH-depression’’ test
unsteady, swirling fluid backward out of its inlet. This for finding the required value of NPSH for the pump
high-velocity, turbulent backflow is full of eddies with to run satisfactorily. The results of this expanded
low-pressure centers that therefore contain bubbles. version of a typical NPSH depression test are sum-
Driven by the backflow into adjacent upstream walls marized in Figure 10-12.
and stationary vanes, these bubbles collapse under the
accompanying higher stagnation pressures. In Figure
10-10b, the stationary upstream guide vanes have been Difference Between Cavitation Inception
perforated by this violently erosive phenomenon. As and Incipient Head Reduction
the pressure increases around the bubble as it col-
lapses, the collapse pressure, pc , becomes greater— In pump research laboratories, the actual cavitating
even greater than occurs within impellers operating fluid entering the pump blades is observed during this
at the BEP. Damage of the magnitude seen in Figure test through special windows with high-speed photog-
10-10b also happens to the driving face or pressure raphy and strobe lights. The observer gets a view very
sides of the impeller blades (the invisible sides of the similar to that of Figure 10-10a, except that the actual
impeller blades of Figure 10-10a), because these blades bubbly flow is also seen [1, 11]. The test begins with
drive the eddying backflow just described. an extremely high value of NPSH such that no bub-
bles can be seen. The inlet pressure to the pump is
then reduced (moving from right to left in Figure
Minimum Flow Limits 10-12). The point at which the first bubble is either
heard or observed is called the point of cavitation
From the above discussion, it is evident that at low- inception, and the NPSH value at this point is called
flow, off-design conditions, intense cavitation activity NPSHi . As indicated in Figure 10-12, no change in
accompanies the unsteady backflow. Erosion, noise, pump head (TDH or Ht ) has occurred down to this
and vibration are more severe because the values of
value of NPSH. In fact, a considerable further reduc-
the suction energy and TDH may not be conservative
tion in NPSH occurs with no head change—down to
enough for extended operation at low-flow, off-
the point where the TDH just begins to fall off. The
design conditions. These conditions have led to the
value of NPSH at this point is called NPSH0% , be-
specification of minimum-flow limits by pump manu- cause the head has fallen off by zero percent. At this
facturers. As might be expected, such limits vary with point, sheet-cavities of vapor are seen trailing off the
energy level. The minimum flow of high-energy leading edges of the impeller blades if the pump is
pumps is a considerably greater fraction of the BEC running at the BEP. At flow rates less than BEC,
than it is for pumps of lower energy [1]. These limits these sheet cavities become unsteady clouds of bub-
are also influenced by the NPSH. For example, as bles. Yet there has been zero reduction in head. Until
described below, no cavities exist within the pump at relatively recently, it was thought that the NPSH0%
sufficiently high NPSH, and hence, a lower minimum
point was the point of inception or first appearance of
flow can be permitted.
bubbles. Certainly, it is the point of incipient head
reduction. To avoid confusion between incipient
head reduction and inception of cavity formation,
Domains and Operational Criteria for NPSH however, the term NPSHi is confined here to the
latter, usually much higher value of NPSH. Because
The net positive suction head, NPSH, was defined NPSHi is so great, it is unusual to find a pump whose
earlier in this Section 10.4 as the absolute static pres- first stage is not cavitating.
sure at the pump inlet port, p1 , plus the velocity head
V12 =2g at that same location minus the vapor pres-
sure, pv , of the liquid. Several criteria for NPSH are Net Positive Suction Head Required
used that apply to the various domains of pump
operation with respect to cavitation and its effects. Obtaining repeatable results for NPSH0% has proved
These domains are best described in terms of a con- so difficult that the idea of a measurable reduction in
stant-speed, constant-flow rate (discharge) test in head that would be both repeatable and acceptable
which the pump head, TDH or Ht , is recorded as has been established at 3% of the non-cavitating
 10-4. Cavitation 10.21

3% drop

Point of cavitation inception

Head
NPSH at 3% drop

NPSH0% (at 0% drop)

NPSHi (at point of inception)

NPSH
Figure 10-12. Net positive suction head criteria as determined from pump tests.

head of the pump. The value of NPSH at this point measurable length C. At much lower flow rates, re-
has been given the name NPSH3% or NPSHR, be- circulating flow breaks this sheet into clouds of bub-
cause this is the value of NPSH that is required to bles, and a distinct value of C cannot be determined.
maintain pump performance. At this point, the sheet For flow rates below about 110% of the BEC, the
cavity (or clouds) of vapor on each blade extends all cavity exists on the suction or trailing face of the
the way to the leading edge of the following blade. blade. At higher flows it is on the pressure or driving
That condition causes only 3% reduction in TDH, face. For the typical usage (flow rate <110% BEC),
but as seen in Figure 10-12, any further reduction in the following can be said about the value of C:
NPSH leads to a complete breakdown in perform-
ance (i.e., total loss of pump liquid head, TDH). The . At NPSHi, C is zero.
NPSH at this point is sometimes called the ‘‘break- . At NPSH3% (almost at breakdown), C is essentially
down-NPSH.’’ the distance from the leading edge of one blade to
that of the next blade.
Potential for Cavitation Damage In other words, at breakdown, this sheet-like cav-
ity stretches into the passage between two adjacent
The potential for cavitation damage arising from blades. Determination of C at intermediate values of
erosion and vibration exists throughout the range of NPSH in the above range is best made through actual
NPSH from NPSHi to breakdown. Vibration be- visualization in pump development laboratories. As
comes acute if the flow rate is less than the ‘‘minimum that is inconvenient for users, estimate C by interpol-
flow’’ specified by the pump manufacturer. The rate ating between the two extremes in this NPSH range,
of erosion is roughly described by Equation 10-24 or (for an approximate equation) see page 2.91 in
[1, 11]: Karassik et al. [1].

P / B 6 (C  N)3 (10-24) Domains versus Flow Rate (Discharge)

where P is rate of penetration through a pump blade Applying the foregoing explanations to the entire
or other surface, B is impeller blade inlet tip speed range of flow rate of a pump at constant speed yields
(see Figure 10-6), C is sheet cavity length, and N is the curves in Figure 10-13. The potential for damage
NPSH. is seen as covering the range up to the inception-
So, the erosion peaks at an NPSH value some- NPSH line; NPSHi is typically five or more times
where between NPSHi and breakdown—usually the value of NPSHR [11]. As the energy level in-
between NPSH 0% and NPSH3% , depending on the creases, so does the NPSH at which the damage is
pump design. The severity of the erosion depends significant. For example, in power-generating sta-
on the energy level of the pump as represented by tions, the boiler feed pumps are often ‘‘high-energy
the term, B 6 in Equation 10-24. pumps’’ and need a value of NPSH to be supplied
The limits of application for Equation 10-24 are that is close to this NSPH0% -curve to have an accept-
within about 25% of the BEC. In this range the able life in resistance to cavitation erosion and mech-
cavity trailing off the leading edge of each blade anical damage. The supplied value of NPSH is called
appears to laboratory observers as a ‘‘sheet’’ of finite, the available NPSH or NPSHA.
10.22 Chapter 10 Performance of Centrifugal Pumps

TDH without cavitation

3% loss of TDH used for NPSHR
BEP

NPSHi (inception)
BEP
Head, TDH, fraction of TDH

1.0
NPSH0% - NPSH for 0% loss of TDH

~5ha
Erosion
damage*

~2ha
Mechanical
damage
ha

NPSHR for 3% loss of TDH

0
0 1.0
Flow rate, Q, fraction of BEC
Figure 10-13. Effect of NPSH on TDH and pump damage. NPSHR at BEP for 3% loss of TDH is termed ha . Courtesy
of DuPont Engineering [12] altered. *Note: NPSH to suppress erosion and mechanical damage increases with pump
energy level or suction energy. No damage occurs if NPSH > NPSHi .

Determining the Available Net Positive Suction

Air Injection Head, NPSHA

Injection of small quantities of air into the pump The NPSH ‘‘available’’ (NPSHA) in the actual instal-
inlet reduces the hydraulic shock imposed on the mech- lation is calculated using Equation 10-25.
anical equipment. Although air injection may improve
mechanical stability by acting as a shock absorber,
NPSHA ¼Hbar þ hs  Hvap  hfs
there are distinct disadvantages and seldom does it (10-25)
permit long-term operation, because: (1) oxygen con-  hm  hvol  FS
tent increases corrosion, (2) gas increases the NPSHR,
and (3) air causes noise and some vibration. where:

Hbar is the barometric pressure in m or ft of water

Taking Precautions column corrected for elevation above mean sea
level (see Table A-6 or A-7). Note that storms can
In conclusion, serious erosion can occur as a result of reduce barometric pressure by 1.7%.
blindly accepting data from pump catalogs, because hs is the static head of the intake water surface above
of the current standard (a 3% drop in head) used by the eye of the impeller. If the water surface is below
most pump manufacturers. In critical installations the eye, hs becomes a negative number.
where continuous duty is important or where the Hvap is vapor pressure of the fluid at the maximum
pump is to be operated at a reduced speed, the manu- expected temperature. See Table A-8 or A-9 for the
facturer should be required to furnish information on equivalent height of the water column.
the potential for cavitation erosion, minimum flow hfs is pipe friction in m or ft between the suction
rate, and NPSHR test results. Typically, NPSHR is intake and the pump.
plotted as a continuous curve for a pump (see Figure Shm is the sum of minor pipe friction losses such as
10-5). When impeller trim has a significant effect on entrance, bend, reducer, and valve losses. These
the NPSHR, several curves are plotted. ‘‘minor’’ losses are significant.
 10-4. Cavitation 10.23

hvol is the partial pressure of dissolved gases such as the NPSHR) was adequate, and some very know-
air in water (customarily ignored) or volatile or- ledgeable engineers continue to recommend such
ganic matter in wastewater (customarily estimated safety factors for water and wastewater pumps.
to be about 0.6 m or 2 ft). More conservative engineers warn that anything less
FS is a factor of safety used to account for uncertainty than 1.5 m (5 ft) or less than 1.35 times the NPSHR is
in hydraulic calculations and for the possibility of risking damage to the pump. Consult the pump manu-
swirling or uneven velocity distribution in the intake. facturer to decide which rules to follow, and be cer-
tain the pump manufacturer understands the
There is no term for velocity head in Equation application and the potential variation in operating
10-25, because velocity head is part of the absolute conditions to be imposed upon the pump. Be aware
dynamic head. that to eliminate cavitation and its effects entirely, the
As shown in Figure 10-13, the NPSHA should be NPSHA must, depending on the operating flow rate
greater than the NPSHR by a factor of safety that relative to the BEP, be two to five times the NPSHR.
depends mainly on the energy level of the pump. The effect of NPSH on erosion and mechanical dam-
Although the 3% loss in TDH that defines the age is illustrated in Figure 10-13. Note that NPSH
NPSHR is typically insignificant, the shock on the to suppress erosion and mechanical damage increases
equipment can reduce the life of bearings, seals, and with pump energy level or suction energy. It is assumed
impellers. Surprisingly large variations in NPSHR are that little damage can occur if NPSHA > NPSHi .
sometimes observed in tests of supposedly identical
pumps. Small pumps are more sensitive to casting
imperfections than large ones. Additionally, when
the pump operates at any point but the BEP, the Calculating the Required Net Positive Suction
NPSH may have to be increased substantially to Head, NPSHR
avoid cavitation, so the need for a generous factor of
safety is apparent. In the past, it was customary to Normally the NPSHR of a pump is found from the
consider that 0.6 m or 2 ft (but no less than 20% of manufacturer’s tests; however, there is a body of

Example 10-5
Calculation of Net Positive Suction Head Available (NPSHA)

Problem: Estimate the NPSHA for the pump system shown in Figure 10-3. Use the data given
in Example 10-1. Assume the water temperature is 408C (1048F) and the elevation is 500 m
(1600 ft).
Solution: Use Equation 10-25 with data from Appendix Tables A-6 to A-9 and calculations in
Example 10-1.

SI Units U.S. Customary Units

Hbar (Table A-6) ¼ þ9.74 Hbar (interpolate Table A-7) ¼þ32.0
hs (Example 10-1) ¼ 2.00 hs (Example 10-1) ¼  6.56
Hvap (Table A-8) ¼ 0.76 Hvap (interpolate Table A-8) ¼ 2.48
hent ¼ 0:004 hent ¼ 0:014
hfs ¼ 0:015 hfs ¼ 0:048
Shm ¼ 0:022 þ 0:018 ¼ 0:040 Shm ¼ 0:072 þ 0:058 ¼ 0:130
Subtotal ¼ þ6.921 Subtotal ¼ þ22.768
Safety factor a
¼ 1.5 Safety factor a
¼  5.0
NPSHA ¼ þ5.421 NPSHA ¼ þ17.768
Use þ5.4 Use þ18
a
Note that according to some engineers, the safety factor should be at least 1.5 m (5 ft) and also not less
than 35% of NPSHR. According to others, 0.6 m (2 ft) and also not less than 10% of NPSHR is adequate
for water and wastewater.
10.24 Chapter 10 Performance of Centrifugal Pumps

theory and empirical correlations that are used to

make estimates when the test results are not available
[1, 8]. The method of Rutschi [14], which is presented
here, provides a correlation for Thoma’s cavitation
parameter, s, the ratio of the NPSHR to the pump
head TDH or Ht [8, 9, 10, 13]:

NPSHR
s¼ (10-26)
Ht

There is disagreement in the literature as to whether

NPSHR (¼ NPSH3% ) or NPSH0% should be used in
the definition of Thoma’s cavitation parameter.
NPSHR is used here because, as seen in the next ex-
ample, applying Rutschi’s method yields an NPSH
result that is only 4% greater than the manufacturer’s
NPSHR value. Moreover, as indicated in Figure 10-
12, NPSH0% tends to be more nearly twice NPSHR
and is difficult to establish repeatedly.
With s equal to the ratio NPSHR=Ht , it follows
from the definition of specific speed, ns (Example 10-
4), and suction specific speed, nss (Equation 10-23),
that s ¼ (ns =nss )4=3 . As Wislicenus [10] demonstrated,
nss tends to be constant over the entire range of ns .
Thus Rutschi [14] developed a correlation for
Thoma’s cavitation parameter that depends on ns .
Pump efficiency is also involved, as low pump effi- Figure 10-14. Cavitation constant, s, versus specific
ciency means greater internal losses and pressure speed of pumps. After Rutschi [14].
drops and therefore greater NPSHR (lower nss ). Plot-
ted in Figure 10-14, Rutschi’s relationship is as fol-
lows:
Figure 10-14 deviate by as much as 25% from the
4=3 line of best fit. Second, the K coefficients in Table
K(nq or Ns )
s¼ (10-27) 10-2 vary more widely than do the efficiencies. (A
106 change of 30% in efficiency reduces K by half.) Fi-
nally, the efficiency of pumps (upon which K is so
where the values of K are given in Table 10-2. dependent) can easily vary by 2%. Although the cal-
The accuracy implied by the results of Example culations cannot be trusted to yield errors of less than
10-6 may be misleading. First, the data points in 30 or 40%, they nevertheless provide both insight and
approximate results.

Table 10-2. Values of K for Equation 10-27a Predicting NPSHR at Off-BEP Flow Rates

Pump efficiency (%) SI unitsb U.S. customary unitsc If the pump operates at low head at a flow rate
considerably greater than the BEC, Equation 10-28
70 1726 8.99
is approximately correct:
80 1210 6.30
90 796 4.14
NPSHR at operating point
a
For double suction pumps, use the same formula with Q equal to
NPSHR at BEP
half of the actual value.  
Use with nq values in m3 =s, m, and rev/min.
Q at operating point n
¼
b
(10-28)
c
Use with Ns values in gal/min, ft, and rev/min.
Q at BEP
 10-4. Cavitation 10.25

Example 10-6
Determination of NPSHR and NPSHA

Problem: At a total dynamic head of 36.5 m (120 ft) and a rotational speed of 1170 rev/min,
the pump described in Figure 10-5 delivers a flow of 500 m3 =h (2200 gal/min). Assume that the
pump is to operate at sea level at a temperature of 158C (598F) and that the suction losses are
2.0 m (6.6 ft).
(1) Estimate the required NPSHR, (2) compare the computed NPSHR to the pump
manufacturer’s NPSHR value given in the figure, and (3) estimate the allowable suction head.
Solution:
(1) NPSHR. Equate Equations 10-26 and 10-27 to

Ht Kðnq or Ns Þ4=3
NPSHR ¼
106

and calculate the specific speed using Equations 10-22a and 10-22b. Because the pump curves
shown in Figure 10-5 are for a split-case pump with a double-suction impeller, one-half of
the flow rate is used in computing ns . Use K values of 1140 and 6.0 at 82% efficiency (see Table
10-2).
SI Units U.S. Customary Units

 
250 1=2
nq ¼ 1170 (36:5)3=4 Ns ¼ 1170(1100)1=2 (120)3=4
3600
¼ 20:8 ¼ 1070
36:5  1140 (20:8)4=3 120  6:0 (1070)4=3
NPSHR ¼ ¼ 2:38 m NPSHR ¼ ¼ 7:88 ft
106 106

(2) Compare NPSHR to the pump manufacturer’s NPSHR. The computed value is 4% greater
than the value of 2.28 m (7.4 ft) given in Figure 10-5.
(3) Estimate the allowable suction head. Rewriting Equation 10-25 in terms of the suction
head yields

hs ¼ NPSHA  Hbar þ Hvap þ hvol þ hfs þ Shm þ FS

where, for this example, hvol is assumed to be zero, and hfs þ Shm are given as 2 m (6.6 ft).
Substituting the more conservative value of NPSHR (the computed value in this example) for
NPSHA, the value of hs is

hs ¼ 2:38  10:33 þ 0:17 þ 2:0 þ 1:5 hs ¼ 7:88  33:9 þ 0:58 þ 6:6 þ 5:0

¼ 4:28 m ¼ 13:8 ft

where the exponent n varies from 1.25 to 3.0, depend- Prevention and Control of Cavitation
ing on the design of the impeller. In most water and
wastewater pumps, n lies between 1.8 and 2.8. The The adverse effects of cavitation can be avoided most
NPSHR at the BEP increases with the specific speed easily by simply raising the NPSHA sufficiently to
of the pumps. For high-head pumps, it may be neces- ensure that no vaporization takes place within the
sary either to limit the speed to obtain the adequate pump. For most pumps this increase would be far
NPSH at the operating point or to lower the elevation more than necessary; and as NPSHA would then
of the pump with respect to the free water surface on have to exceed the inception-NPSH or NPSHi , this
the suction side. figure is so high as to be as much head as many
10.26 Chapter 10 Performance of Centrifugal Pumps

pumps generate. Furthermore, as discussed previ- . Add a booster pump or an inducer upstream from
ously, for sufficiently low pump head the energy the main pump impeller. Illustrated in Figure
level is so low that the cavitation does little or no 10-15, an inducer is an auxiliary, low-head, axial-
damage. The real requirement is for just enough flow propeller mounted just upstream from the
NPSH—as suggested by the explanation accompany- impeller that supplies some of the NPSH required
ing Figure 10-13—to ensure adequate pump life by the impeller. An inducer differs from a typical
against cavitation damage for the energy level of the propeller in that it has long, tightly wrapped helical
pump involved. A low-energy pump [e.g., a centrifu- blades (usually two or three) that require much less
gal pump with a head of less than 120 m (400 ft)] can NPSHR than does an impeller. The inducer has a
theoretically get along indefinitely with NPSHA very high suction-specific-speed (nss ) capability—
being no greater than NPSHR. To guard against nss (see Equation 10-22a) being from 300 to 600 in
uncertainties in NPSHA and/or NPSHR and a pos- SI units of rev/min, m3 =s; and m (or 15,000 to
sible loss of performance, a factor of safety in the 30,000 in U.S. customary units of rev/min, gal/
form of some added NPSHA is good practice. min, and ft) [1, 9, 13]. Inducers normally run with
Preventing both performance degradation and long cavities on the suction sides of their blades,
damage, therefore, begins with (1) raising NPSHA, and it is the unique behavior of these cavities that is
(2) reducing NPSHR, and (3) changing the pump connected with the high nss capability of inducers.
design. Some ways of accomplishing these prevention The difficulty with inducers is that the range of
approaches are: flow rate Q is quite limited unless the suction en-
ergy is quite low. In particular, for Q lower than
the low end of this range, the cavities evolve into
Raising NPSHA surging clouds of vapor, backflowing at unequal
rates and so imposing fluctuating loads (both ra-
. Decrease the suction lift (see Figure 10-3) or in- dial and axial) on the inducer, coupled with exces-
crease the suction head, hs (see Figure 10-2). sive erosion. The resulting wear would require
. Decrease the suction losses (e.g., with larger pipes, replacement of the inducer, which may be preferred
more gradual turns, etc.). to replacing the more expensive impeller. However,
. Where possible, lower the vapor pressure.
the violence of the surging might first cause mech-
anical failure of the whole machine, especially if it
Reducing the NPSHR is a high-energy pump. Passive methods of elimin-
ating this off-design ‘‘cavitation surge’’ have been
. Reduce the impeller rotating speed (rev/min)—pos- devised but are not widely available [1].
sible if the pump is too large and the flow rate is . Take advantage of the favorable vaporization
being throttled at the current speed to drop both characteristics of some liquids. Compared with
head and flow rate to those required. what occurs when cold water vaporizes, the volume
. Avoid operation at or near the manufacturer’s spe-
cified ‘‘minimum flow.’’
. Inject air at the inlet of the pump. However, as
explained above, that may be of doubtful value.
. Decrease the maximum flow rate (which reduces
NPSHi ) to prevent cavitation.
. Use a variable-speed drive that reduces rev/min
when possible (e.g., when throttling flow rate, less
pump head is needed so the speed can be reduced).
Also, because the BEC (BEP flow rate) decreases
directly with rev/min, a lower speed brings BEC
closer to the actual, lower Q.
. Change the pump or the impeller to one designed for
a lower NPSHR, although such a change is often
accomplished by increasing the impeller inlet diam-
eter or blade inlet tip diameter. Such an increase also
raises the suction energy and could cause more dam- Figure 10-15. Typical inducer (axial-flow impeller) at-
age at much lower flow rates should the pump be tached to a conventional pump impeller. After Turton
required to operate at such reduced flows. [9].
 10-5. Pump Characteristic Curves 10.27

of the cavities formed within the pump is far less 10-5. Pump Characteristic Curves
for high-temperature water. Reductions in NPSHR
below that specified by the pump manufacturer for In most pump curves, the total dynamic head H in m
room-temperature water are possible for high- (ft), the efficiency E in percent, and the power input P
temperature water [1, 8]. in kW (hp) are plotted as ordinates against the flow
rate Q in m3 =s or m3 =h (gal/min or Mgal/d) as the
abscissa.
Changing the Pump Design

. Specify a material for the impeller (and adjacent Nondimensional Pump Characteristic Curves
vanes—Figure 10-11—and other parts exposed to
the collapsing cavities) that is more resistant to Nondimensional pump curves are obtained by ex-
cavitation erosion, such as stainless steel, titanium, pressing head, flow rate, power input, and efficiency
etc. [1]. See Table 10-3. as percentages of the corresponding values at the best
. Utilize an axial inlet nozzle that is free of elbows or efficiency point, and such curves are useful for (1)
other shapes that change the flow direction appre- comparing the hydraulic properties of pumps belong-
ciably, as this would eliminate distortion of the ing to the same type, and (2) assessing the perform-
flow entering the impeller eye, which subjects the ance of pumps of different specific speeds for various
blades to varying angles of attack or ‘‘incidence’’ of applications [6]. The general shape of these curves
the incoming fluid, the larger incidence on the varies with the specific speed, the form and number
blades creating larger cavities. of blades on the impeller, and the form of the casing
. Redesign the impeller blades to minimize the local used. As shown in Figure 10-16, the slope of the head-
reductions of pressure that produce cavities and capacity curve becomes steeper as the specific speed
thus reduce the inception NPSH of the pump increases. Notice the shape of the power curves in
(NPSHi in Figures 10-12 and 10-13). NPSHR is Figure 10-16a. As head decreases below the BEP,
not reduced, but if NPSHi is reduced the damaging the power required increases. Thus, overloading the
action of bubbles at NPSH > NPSHR is reduced motor is most likely to occur in pumps with low
or eliminated [1, 11]. This design approach is a specific speeds and with flat head-capacity (H-Q)
relatively recent development found in some high- curves. Because of the shape of the characteristic
energy pumps. curves in Figure 10-16, mixed-flow and axial-flow

Table 10-3. Relative Resistance to Cavitation and Costs for Various Metals. Courtesy of Fairbanks Morse
Pump Co.

Impeller costa, $

Material Relative cavitation resistance Original Replacement

Cast iron 1 1450 3000

Cast iron with 3% Ni 1þ 2030 4200
Tin bronze (gun metal) 1.2 5800 12,000
Aluminum 1.7 4350 9000
Carbon steel 3.4 2900 6000
Manganese bronze 4.5 5800 12000
Ni-resist 4.6 8700 13,000
Monel 6.8 10,150 21,000
CB7 Stainless steel 8.8 5800 12,000
CF8M ss 14 4800 9900
Titanium 18 29,000 60,000
Aluminum bronze 28 5800 12,000
CA15 ss 410BHN 35 4800 9900
a
Approximate cost in 2004 for a 75-kW (100-hp) wastewater pump impeller about 400 mm (16 in.) in diameter.
10.28 Chapter 10 Performance of Centrifugal Pumps

Figure 10-16. Typical dimensionless characteristic curves for centrifugal pumps. (a) Radial flow; (b) mixed flow; (c)
mixed flow; (d) axial flow.

pumps are non-overloading in their operating range; adsorbed by the fluid rises steeply as the flow rate is
that is, the power required by the pump does not decreased to zero. Thus, with mixed- and axial-flow
increase as the flow rate increases above the BEC. pumps, some means of unloading the pump (such as a
However, depending on the impeller blade design, a bypass pipe with a pressure-activated valve) is needed
dip (and consequent instability) can occur as shown as the head rises so as to avoid overloading the pump
in H-Q curves for pumps of high specific speeds. driver. Sometimes, it is more cost effective to oversize
Prolonged operation of high–specific speed pumps the motor than to install an unloading device. A
in this head-flow rate range should be avoided. comparative cost analysis should be made if there is
As the specific speed increases, the effects on the doubt.
power input are marked. With radial-flow pumps,
where the specific speed, nq , is about 35 (Ns about
1800), the energy input decreases with a decrease in Stable and Unstable Pump Curves
flow. For a mixed-flow pump with a specific speed of
about 85 (Ns about 4300), the power input is nearly The head-flow rate curves for radial-flow centrifugal
constant. In pumps with specific speeds, nq , greater pumps can be either stable (Figure 10-17a) or un-
than about 125 (Ns greater than 6400), the power stable (Figure 10-17b), whereas the head-flow rate
 10-5. Pump Characteristic Curves 10.29

curves for mixed-flow and axial-flow pumps are al- for a given value of head in unstable pump curves, as
ways stable (except for a small area of operation). can be seen from Figure 10-17b.
Unstable pump curves are usually limited to specific The use of pumps with unstable pump curves can
speeds, nq , of less than 20 (Ns < 1000). With stable lead to pumping instabilities in systems with signifi-
pump curves, there is only one flow rate for each cant capacitance or where pumps are operated in
value of head, whereas two flow rates are possible parallel. In such situations, a static and dynamic

Figure 10-17. Typical pump H-Q curves. (a) Stable and (b) unstable.
10.30 Chapter 10 Performance of Centrifugal Pumps

analysis of the entire pumping system should be con- pumping equipment to achieve optimum life from the
ducted to ascertain whether the system is stable under equipment. The document establishes two operating
all conditions. If this analysis cannot be done or gives regions, the Preferred Operating Region (POR), de-
doubtful results, it is best to avoid the use of pumps fined as the region where ‘‘ . . . [T]he service life of the
with unstable H-Q curves entirely. In single-stage pump will not be significantly affected by hydraulic
pumps, stable pump head-capacity curves can be loads, vibration or flow separation.’’ According to
obtained by reducing the number of impeller blades, the document, for most centrifugal pumps, this re-
changing the exit blade angle, and changing the con- gion is nominally bounded between 70 and 120% of
figuration of the blades [6]. BEP flow rate, BEC. For vertical (column-type)
pumps, the POR is the same as defined for centrifugal
pumps for specific speeds, nq ,less than 87 (Ns ¼ 4500)
10-6. Pump Operating Regions and between 80 and 115% of BEP for higher specific
speed pumps. The standard explains that operation
A pump operates best at its BEP. Not only is the outside these limits may be possible, but at the risk of
efficiency maximum, but radial loads on the impeller increased vibration, damaging hydraulic forces, cavi-
and the problems of cavitation are minimized. In tation effects, and reduced service life. This latter
1997, the Hydraulic Institute issued ANSI/HI 9.6.3- region, which is to be defined by the individual
1997, American National Standard for Centrifugal pump manufacturer, is termed in the standard as
and Vertical Pumps Allowable Operating Region. the Acceptable Operating Region (AOR). Many
This standard provides guidance on how to specify manufacturers are now indicating the AOR in their

Example 10-7
Estimating the Size of the Pump Required

Problem: Water is to be pumped at a flow rate of 0:2 m3 =s (3166 gal/min) to a height of 30 m

(98.4 ft) at a rotational speed of 1150 rev/min. Determine (1) the type of pump, and (2) the
impeller diameter.
Solution:
(1) Type of pump. Calculate the specific speed using Equations 10-22a and 10-22b and relate
the type of pump to the specific speed.

SI Units U.S. Customary Units

nq ¼ 1150(0:2)1=2 (30)3=4 ¼ 40:1 Ns ¼ 1150(3:166)1=2 (98:4)3=4 ¼ 2071

From Figure 10-8, a pump with a mixed-flow (e.g., screw or helicoidal) type of impeller is
needed.
(2) Impeller diameter. Referring to Figure 10-16, the normalized shut-off head for a pump
with a specific speed, nq of about 40 (Ns about 2000) is estimated to be about 1.4 times the
value at the BEP. At shut-off head, the head developed by a pump is proportional to u2 =2g,
where u is the peripheral speed of the impeller (see Figure 10-6). Assuming that the head at the
BEP is 30 m (98.4 ft), the shut-off head is

H ¼ u22 =2g ¼ 1:4  30 ¼ 42 m H ¼ u22 =2g ¼ 1:4  98:4 ¼ 137:8 ft

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u ¼ 42  2  9:81 ¼ 28:7 m=s u ¼ 137:8  2  32:2 ¼ 94:2 ft=s
u2 ¼ pDn=60 u2 ¼ pDn=60
D ¼ 60u=pn ¼ 60  28:7=(3:14  150) D ¼ 60u=pn ¼ 60  94:2=(3:14  150)
¼ 0:477 m impeller diameter ¼ 1:56 ft ¼ 18:8 in: impeller diameter
 10-6. Pump Operating Regions 10.31

catalogs, but only a few to date have also displayed provide POR and AOR limitations within their lit-
the POR. The standard should be used by the speci- erature and catalogs, the users will have a greater
fying party to define the requirements for pump se- appreciation for the capabilities and limitations of
lections to individual applications. The following individual equipment selections and will be able to
rules are suggested: judge more readily the relative merits of competing
designs.
. Routine, day-in, day-out (i.e., daily water demand,
or dry weather flow) operating conditions should
be located in the pump’s POR
Radial Thrust
. Peak operating conditions (i.e., peak fire flow,
peak wet weather flow) can be located in the
Pump volutes are so designed that at optimum flow
pump’s AOR because these infrequently experi-
rate (at the BEP) the average pressure around the OD
enced episodes will not result in significant equip-
of the impeller approaches uniformity. At other flow
ment deterioration.
rates, the pressures around the OD of the impeller
The advent of this ANSI/HI standard represents a vary considerably. The unbalanced radial forces are
departure in the process of promoting the capabilities represented as the radial thrust, P2  P1 , in Figure
of and specifying pumping equipment. If pump 10-18a, where P is the pressure times the projected
manufacturers follow through with the standard and area of the impeller. Radial thrust always occurs in

Figure 10-18. Development of radial thrust in a radial-flow centrifugal pump shaft. (a) Plan of pump; (b) impeller
and shaft; (c) radial thrust versus flow rate.
10.32 Chapter 10 Performance of Centrifugal Pumps

centrifugal pumps, but it is of greatest concern in end- Fixed Efficiency Loss

suction overhung shaft designs as shown in Figure
10-18b. For a given impeller diameter, the operating range
The magnitude of radial thrust depends on such of a pump can be established by (1) setting a limit on
operating conditions as flow rate, head, and rota- the minimum acceptable efficiency, and (2) setting
tional speed. Radial thrust can be astonishingly upper and lower limits on the allowable speed
large. For example, tests of a 350-mm (14-in.) pump changes. Alternatively, if the diameter of the pump
rated at 0:5 m3 =s (7800 gal/min) at 50-m (163-ft) head impeller is to be changed, the operating range of the
showed that the radial thrust at shut-off was 36 kN pump can be established by (1) setting a limit on
(8200 lb) at 1160 rev/min and 24 kN (5400 lb) at a the minimum acceptable efficiency, and (2) setting
variable speed when operated along the station’s sys- upper and lower limits on the allowable impeller
tem curve (see Figure 10-18c). diameters.
Be aware of the potential for damage by radial The operating range of a pump based on these
loads that may be too great for the load-carrying criteria is illustrated in Figure 10-19.
capacity of standard shafts and bearings. Pumps
have occasionally been destroyed in a few days or
even a few hours through excessive wear in bearings, Pump Operating Range
seals, shaft sleeves, and wearing rings. Bearings,
shafts, and casings must be designed to resist the An alternative approach used to establish the pump
anticipated radial loads. Radial loads can be approxi- operating range is to set limits on the pump flow rate
mated by the method given in ANSI/HI 1.3-2000. as a percentage of the BEC. An operating range of 60
Standard pumps can often be customized with larger to 125% of the BEC is often recommended for pumps
shafts and bearings. One practical means of dealing with specific speeds, nq , less than 40(Ns < 2000) [15]
with radial thrust is to define several critical operat- (see ANSI/HI 9.6.3). Pump operating envelopes are
ing points and require the manufacturer to furnish a defined using these values, as illustrated in Figure
pump with a specified life at such points (see Appen- 10-20, for changes in speed as well as for changes in
dix C). the impeller diameter.

Cavitation 10-7. Elementary Pump System Analysis

When the pump flow rate increases beyond the BEC, In any pump–force main system, the head developed
the absolute pressure at the pump inlet required to by the pump must equal the total dynamic headloss in
prevent the possible adverse effects of cavitation in- the system and, of course, the e flow rate in pump and
creases so that cavitation is a potential problem if pipe are equal. This relationship determines the
there is insufficient NPSH margin. The NPSHR ‘‘pump operating point,’’ a single point on both the
curves are the basis for selecting the maximum per- H-Q curve of the pump and the H-Q curve of
missible flow rate, but the engineer must be aware the piping system. There are several ways of finding
that the amount of margin, in terms of percentage of the pump operating point, including computer analy-
NPSHR, increases dramatically as flow rate exceeds sis and several graphical methods, but the simplest
approximately 120% of BEC or falls below 85% of (for simple problems) and the most universally used
BEC. When the pump flow rate decreases toward method is given in this section.
zero at the shut-off head, the radial load increases
and the recirculation of the pumped fluid within the
impeller becomes a problem. This recirculation can System Head-Flow Rate (H-Q) Curves
cause vibration and could combine with cavitation to
ruin the impeller (see Section 10-4). In wastewater The transmission pipeline H-Q curves in Figure 10-21
pumps, recirculation can result in plugging caused are plotted over a range of flows from zero (where the
by recirculated flow tumbling rags and stringy mater- only head is due to static lift) to the maximum
ial in the pumped flow. expected (which also includes all friction, fitting,
Pumps should not be operated at or near the shut- and valve losses). Because transmission pipeline or
off head for extended periods of time because such force main head-capacity curves are approximate
operation also leads to heating of the liquid and can functions of v2 =2g, the curve is approximately a
cause severe wearing-ring rub and other problems. parabola with its apex on the zero Q line. Actually,
 10-7. Elementary Pump System Analysis 10.33

Figure 10-19. Pump operating envelopes based on a fixed efficiency loss. (a) Change of rotational speed; (b) change
of impeller diameter.

Figure 10-20. Pump operating envelopes based on the percentage of flow rate at the BEP. (a) Rotational speed;
(b) impeller diameter.

headloss is a function of Q1:85 —not Q2 —in the posits, or deterioration may reduce the H–W
Hazen–Williams equation. In the Darcy–Weisbach coefficient C to as little as 120 for pipes lined with
equation, the value of the friction factor, f, changes cement-mortar (see Table B-5 and Section 3-2).
with the Reynolds number. However, the use of a Under rare conditions, the coefficient C may fall to
parabola is close enough for practical purposes. 100 or less.
New pipe is expected to be very smooth for some A simple envelope of transmission pipeline H-Q
indeterminate time. But bacterial slime, other de- curves is shown in Figure 10-21.
 10.34 Chapter 10 Performance of Centrifugal Pumps

Single-Pump, Single-Speed Operation

Station H-Q curves
The point on any specific system head-flow rate curve
at which a single-speed pump must operate is deter-
0
mined by superimposing the pump H-Q curve on the
10 transmission pipeline H-Q curve as shown in Figure
c = 20 140
1 c= 10-22. The point of intersection is the pump operating
c=
120 point. As the pipe ages and the pipeline H-Q curve
Head

c= rises, the operating point moves up and to the left

100 140
c= c= along the pump H-Q curve.
As pumps start and stop (or as they change speed)
because of the change in water level, the path of
operating points is extended still further along the
pump curve. Pumps and impellers must be chosen
Total Friction Total for a range of operating conditions. The efficiency
static static
curve is usually rather broad at the top. Hence,
head for head
for choose an impeller that operates from slightly left to
HWL
LWL slightly right of the BEP. If the range in operating
Flow rate points is too great to fit the efficiency curve nicely,
choose an impeller that fits in the early years when
Figure 10-21. Typical envelope of transmission pipeline the pipe is smooth. Then a new impeller of larger
head-capacity (H-Q) curves.
diameter can be installed when conditions warrant
it. Pay attention to the motor sizing.

25
80

70
20
System head-
60 capacity
curve
Pump head
50 15 capacity curve
Head, m
Head, ft

40 Pump operating point 80

Efficiency, %

10 70
30
60

20 50
5 Efficiency
hs 40
10

0 0
0 50 100 150 200 250
Flow rate, m3/h

0 500 1000
Flow rate, gal/min
Figure 10-22. Determining the operating point for a single-speed pump with a fixed value of hs .
 10-7. Elementary Pump System Analysis 10.35

Single-Pump, Variable-Speed Operation determine the rotational speed at any other desired
operating point on the system curve. Again, the op-
The point on any head-flow rate curve at which a erating point moves along the pump H-Q curve,
variable-speed pump must operate is determined as although not so far, because the change between
described previously for the single-speed pump. The HWL and LWL is less than with single-speed oper-
affinity laws are used, as shown in Example 10-8, to ation.

Example 10-8
Application of a Variable-Speed Pump

Problem: The pump curve shown in Figure 10-23 is for a pump operated at 1150 rev/min that
pumps 130 m3 =h (572 gal/min) at a head of 13.2 m (43.3 ft), which is shown as point a. If the
pump must discharge 150 m3 =h (660 gal/min) at a head of 15.5 m (50.8 ft), which is shown as
point b, what is (1) the required new operating speed, and (2) the efficiency of the new
operating point?
Solution: (1) Operating speed. Extend the system H-Q curve from point a to an intersection
with the desired flow rate at point b. The new pump curve must pass through point b. Note
that Equation 10-15 (Q1 =Q2 ¼ n1 =n2 ) cannot be used to find the pump speed because points a

25

80
n=
12
70
re
70 v/m
20 in
n=
11
50
60 re
v/m
in
b
15 Original
50 pump head-
Head, m
Head, ft

capacity curve a
c
40 System
head-capacity
10 curve
30
l a
bo
ra
pa

80
w

20
la

Efficiency

5 d
ity
fin

70
Af

10 hs
Efficiency 60

0 0
0 50 100 150 200
Flow rate, m3/h

0 200 400 600 800

Flow rate, gal/min
Figure 10-23. Determining the pump operating points for a single variable-speed pump and a system curve with a
fixed value of static lift (hs ).
10.36 Chapter 10 Performance of Centrifugal Pumps

and b are not corresponding points (see Example 10-2 and the text following Equation 10-17).
One way to solve the problem is to create an ‘‘affinity law parabola’’ by solving Equations
10-15 and 10-16 simultaneously to eliminate n:

Q21 n21 H1
¼ ¼ , H1 ¼ H2 (Q1 =Q2 )2
Q2 n2 H2

which defines a parabola passing through the origin and point b in Figure 10-23. Every point
on the parabola is a corresponding point. Because point b corresponds to point c, Equation
10-15 can be used to find speed at point b after Q is scaled at point c.

SI Units modified to m3 / h U.S. Customary Units

Qc ¼ 136 m3 =h (by scaling) Qc ¼ 598 gal= min (by scaling)

From Equation 10-15 (nb ¼ nc Qb =Qc ),

nb ¼ 1150 (150=136) ¼ 1270 rev= min nb ¼ 1150 (660=598) ¼ 1270 rev= min

By scaling H at points b and c, the speed can also be found from Equation 10-16

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(nb ¼ nc Hb =Hc )
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
nb ¼ 1150 15:4=12:6 ¼ 1270 rev= min nb ¼ 1150 50:5=41:3 ¼ 1270 rev= min

(2) Efficiency. The efficiency of the new operating point is found by projecting a line from point
c downward to point d on the efficiency curve, which gives an expected efficiency of 75%.

Multiple Pump Operation headlosses from a to b to c are subtracted from the H-

Q curve for pump P1. Similarly, the losses from a to d
In pumping stations where several pumps are in- to c are subtracted from the H-Q curve for pump P2.
stalled, two or more pumps are usually operated in The modified pump curves are shown as dashed lines
parallel. Occasionally, pumps are operated in series.
The elements of the methods used to develop the
combined H-Q curve for multiple pumps are outlined
in the following subsections. A more extensive
explanation for multiple pumps is given in Section b P1
a c
10-8.
d
a
P2 P2
Parallel Operation
P1
Head

When two or more pumps are discharging into the j

k e
same header or manifold, the total flow is found by p MP1 MP2
h g f
adding the individual flows at a given head. A minor
n
error is introduced, however, because the system
curve does not include headlosses in the suction and
discharge piping. One method for eliminating the MP1 MP2
error is to subtract from the pump curve those station
I m
headlosses in the suction and discharge piping up to
the intersection of the last pump discharge and the Flow rate
manifold (header). In the inset of Figure 10-24, the Figure 10-24. Operation of two pumps in parallel.
 10-8. Practical Pumping System H-Q Curve Analysis 10.37

labeled MP1 and MP2 in the figure. Adding the 10-8. Practical Pumping System H-Q Curve
abscissas (or flow rates) of MP1 and MP2 produces Analysis
the pump curve MP1 k MP2, where the symbol ‘‘k’’
means ‘‘in parallel with.’’ Thus, he ¼ hg þ hf. The Review the fundamentals of pumping system analysis
intersection of MP1 k MP2 with the system curve for before reading this section in which additional details
the manifold and force main downstream from point add practicality to this subject. The ultimate objective
c is the operating point (point e) for the pumps when in analyzing a pumping system is to describe the
both are running. required range of operating conditions in sufficient
To find the flow rate contributed by each pump, detail to be able (1) to understand the application
draw the horizontal line, eh, which intersects the completely, and (2) to describe the application to
modified pump curves at f and g. The discharge equipment suppliers in sufficient detail for them to
from P1 is hg; from P2 it is hf. To find the total apply their products intelligently.
head at which each individual pump operates, project As noted in Section 3-2, hydraulic calculations are
a vertical line from f to j and from g to k. The actual never precise, regardless of the care taken in assessing
operating point of P1 is point k and the head on the headlosses in valves, fittings, and piping. The head on
pump is kl. The operating point of P2 is point j and pumps is influenced by a number of variables includ-
the head is jm. With one pump off, the other operates ing:
at point n (for P1) or point p for P2. The pump
specifications must be written for pump P1 operating . Variations in static lift
either at point k or point n and for pump P2 at point j . Variations in fluid viscosity
or point p. . Variations in pumping rate
. Deviations from assumed pipe roughness coeffi-
Series Operation cients.

Pumps are operated in series, as shown in Figure The last two variables have been the cause
10-25, when the head requirement is greater than of many failed or poorly performing pumping sta-
can be obtained with a single pump. The combined tions. Engineers tend to focus obsessively on max-
head capacity curve, P1 þ P2, is found by adding imum flow and head for pump selection and
ordinates—that is, the heads developed at the same acceptance tests, with insufficient emphasis on daily
flow rate . Thus, ad ¼ bd þ cd. The operating point is operation at minimum flow rates. Maximum design
point c for pump P1 and point b for pump P2, and the flow rates rarely occur, whereas, during dry weather
operating point for the system is point a. or average demand days, low flow rates occur daily,
and these usually impose the most severe operating
conditions on pumping equipment—particularly in
variable speed applications. As headloss calculations
are inaccurate, prudent engineers develop system
curves based on both the worst and best expected
combination of headloss factors. Actual pumping
P1 + P2
conditions will lie somewhere between these two ex-
P1 P2
a tremes, so pumps must be capable of operation
System curve within their AOR at any combination of head and
flow between the two extremes. Furthermore, pumps
should operate within their POR during the most
Head

b common flow rates and heads—usually at or close

Pump P2
to runout.
When considering operating regimes, visualize
c what is happening in the system and do not ignore
Pump P1 characteristics that may alter system requirements.
These include:
. Combination valves that may open to alter static
d
lift at lower pumping rates
Flow rate . Operation of other pumping stations that may be
Figure 10-25. Operation of two pumps in series. connected to the same discharge pipeline.
10.38 Chapter 10 Performance of Centrifugal Pumps

Constant-Speed (C/S) Pumps in Parallel both HWL and LWL in the wet well, forebay, or
supply main, and the extremes of friction losses in
A typical schematic diagram of a pump and piping the force main. (See Section 3-2, Tables B-5 to B-7,
system for a pumping station with three duty C/S and Figures B-2 and B-3.) Curve A represents losses
pumps (plus a standby pump) that discharge into at the greatest static lift (LWL in the wet well) and the
a manifold and thence into the force main is shown worst condition for pipeline resistance losses—in this
in Figure 10-26. The system is divided into three example for Hazen–Williams C ¼ 120. Curve A’ rep-
basic components: force main, suction piping, and resents the lowest static lift condition (HWL in the
discharge piping. Head-capacity (H-Q) curves are wet well) and similar pipeline losses. Curve A’’ repre-
drawn for each component. The H-Q curves for the sents the lowest static lift condition with the most
pumps are then superimposed. In Figure 10-27, favorable (best) pipeline resistance losses—assumed
dynamic headlosses in suction and discharge piping here to be C ¼ 145. The worst and best are the
are combined into Curve BþC. The pump, boundary conditions that describe the range of sys-
together with its suction and discharge piping (be- tem head characteristics. Minor losses in the force
tween manifold and wet well), is herein termed a main caused by valves and fittings can, perhaps, be
‘‘pumping unit.’’ ignored if the pipe is very long.
Curve BþC is the sum of friction and turbulence
losses from the suction intake to the manifold.
Station Curves (Losses on the inlet side of the pump are calculated
and accumulated separately, because they are needed
In Figure 10-27, curves A, A’, and A00 are drawn first. to calculate NPSHA.) Here, turbulence losses in pipe
They are combinations of static head plus the dy- fittings and valves are more significant than pipe
namic headlosses in and from the manifold to the friction losses. Typical loss coefficients are given in
point of discharge. The curves must encompass the Tables B-6 and B-7. Curve BþC is drawn for the
entire envelope of operating conditions, including maximum expected losses.

Pumping unit

Figure 10-26. Schematic diagram of pump and piping system. One of four (3 duty) pumps is shown.
 10-8. Practical Pumping System H-Q Curve Analysis 10.39

Mfrs H-Q curve impeller as cast, M

Mfrs H-Q curve for trimmed impeller, M
M-(B+C)=M
2M
3M
st
Wor
3 =120
1 A C
Head

2
=1 20
y23 A C
st
=14 5 Be
A C 6
LWL 5 (Runout)
y45
HWL
4

es
loss
B+C

y23 y45

0.33Qfirm 0.67Qfirm Qfirm Qmax

Flow rate
Figure 10-27. System H-Q curve analysis for three duty C/S pumps per Figure 10-26.

Analysis through Point 2 and, by extension, to Point 4 on the

best curve (Curve A’’). The abscissa of Point 4 repre-
Under the worst conditions, three pumps and their sents the maximum flow rate that can pass through
associated piping, BþC (the pumping units) discharge the pump, and the pump operating point is Point 5
the firm pumping capacity, Qfirm , at the maximum (called runout), y45 above Point 4. When the system is
head of Point 1, so each pumping unit discharges new, the pump will normally operate at or near run-
Qfirm =3 at an equal head, Point 2. The headloss out whenever the pump is activated by the HWL
from sump to manifold (Curve BþC) is y23 , so the signal. For the life of the station, all operating points
pump must operate at Point 3, y23 above Point 2. As fall between Points 3 and 5.
Point 3 represents the greatest head on the pump Make further efforts to select a pump for which
(unless some valve is closed), it is called the ‘‘war- Points 5 and 3 are within the POR. If that range is too
ranted point,’’ the only point for which both head large, make sure that Point 5 is within the POR and
and flow rate are warranted by the manufacturer. Point 3 within the AOR.
Note that the pumps will rarely—if ever—operate at What happens during the highest flow rate to the
Point 3, because Qfirm , the maximum flow that can station? From the above, the H-Q curve for a pump-
occur, is always chosen to exceed expected peak flow ing unit is Curve M’’. When two pumps operate, the
rates, and furthermore, peak flows rarely persist for curve is 2M’’, and three pumps produce Curve 3M’’.
more than an hour or so. The maximum station flow rate that can occur is
Choose a pump with an H-Q curve (here, Curve labeled Qmax with the head at Point 6. If the same
M) that includes Point 3—that is, above and to the construction for Points 1, 2, and 3 is developed for
right of Point 3. With the proper impeller trim, Curve Point 6, the operating point for the pump falls be-
M becomes Curve M’, passing through Point 3. (Re- tween Points 3 and 5.
view Example 10-3 to ensure the affinity laws are Now refer to the manufacturer’s NPSHR curve
applied properly for trimming impellers.) If the and determine whether there is sufficient NPSHA
pump and its associated piping (CþB) is considered for a comfortable margin—not less than 130% of
to be a pumping unit, its H-Q curve is M’’ passing NPSHR if the operating point is within  15% of
10.40 Chapter 10 Performance of Centrifugal Pumps

BEC. Outside of that boundary, the margin should be Hydraulic Profile

much greater—as much as 180% of the NPSHR. The
alternative is either to lower the planned setting for The next step (which should not be omitted) is to plot
the pumps or seek another pump performance curve. the hydraulic profile for the system curves against the
Always bear in mind that pump performance is profile for the pipeline under the intended modes of
worse, in terms of vibration, potential cavitation, and operation. Although extreme accuracy is not critical,
mechanical damage to the pump, at operating condi- it is important to assess the system as it is intended to
tions different from the best BEP. As a general rule, function. Try to visualize how the system will behave
off-the-shelf pumps perform best if operating condi- under all operating conditions. Questions to be asked
tions lie between about 70 and 115% of BEC. when performing this task are:

. Are all operating conditions properly defined by

Specifications for Purchase Orders the system curves?
. What happens in the pipeline under each operating
The conditions for both Points 3 and 5 must be condition?
published in the purchase order or project specifica- . Are there any peculiarities in the operation of the
tions, because that is the only legally defensible system that have not been considered in the con-
method for communicating performance require- struction of the system curves?
ments to the equipment manufacturer. When specify- . Has everything that might influence pump operat-
ing Point 3, list head, flow, and NPSHA; however, ing conditions (and, hence, pump performance)
when specifying Point 5, list only the head and been considered?
NPSHA because every pump manufacturer has a
pump curve of a different shape and the intersects An example of where one can go astray is seen in
for these curves with Curve A’’ can vary. Figure 10-28, in which a knee is shown in the pipeline

Figure 10-28. Hydraulic profiles for Figure 10-27.

 10-8. Practical Pumping System H-Q Curve Analysis 10.41

profile. In this example, the operating point (Point 3 fewer pumps are needed because of the V/S feature,
in Figure 10-27) for three pumps is properly defined. so piping and space requirements are reduced; and (3)
But when only one pump is running, the operating in properly designed systems, the difference between
point (Point 5 in Figure 10-27) is wrong because the LWL and HWL in V/S pumping is small, slightly less
hydraulic profile in Figure 10-28 is below the high than the diameter of the influent pipe discharging
point in the pipeline. The air-vacuum valve will open, into the wet well, so the average static lift is less
and flow will stop until the pump increases the pres- than that for C/S pumping.
sure enough to send water over the high point. So by
plotting hydraulic profiles, it can be seen that the
transmission pipeline H-Q curve must be raised, as Full-Speed Operation
shown in Figure 10-29, to accommodate a new static
head. The true flow rate is less than the flow rate For full-speed operation with all duty pumps (two in
obtained from Figure 10-27, so the system will not this example) operating, the construction for Points
perform as intended. Blunders of this sort are un- 1, 2, and 3 is the same as that for C/S pumps.
The HWL for a single operating pumping unit is at
likely to inspire client confidence.
the mid-depth of the influent pipe, Point 4 on Curve
M’’, because for any water level higher, the pump
controller is programmed to start a second pump.
Variable-Speed (V/S) Pumps in Parallel
(For reasons, see Figure 12-22 and the accompanying
text.) Point 5 (Runout) is y45 above Point 4.
The important difference between V/S and C/S pump-
ing is that V/S pumps operate over a wide speed
range, and care is required to make certain the Low-Speed Operation
pumps always operate near their BEP at all speeds
(see Section 15-4). Other differences are: (1) storage is The pump controller is programmed to operate a
neither required nor desirable for V/S pumping; (2) single pump at variable speed on a straight line

M

M

Operating point
-27
re 10
m Figu
LWL fro
Head

at LWL
New static head

10-27
m Figure
New static head at HW
L HWL fro

Minimum flow rate per Figure 10-28.

Flow rate
Figure 10-29. System curve analysis for one pump operating at minimum flow as per Figure 10-28. Compare to
Figure 10-27.
10.42 Chapter 10 Performance of Centrifugal Pumps

basis for water levels between the invert and the half- basis), so Point 9, at y89 above Point 8, appears to fall
depth of the influent pipe. Above mid-depth, the on Line A (but in reality, it does not).
controller turns two pumps on, each operating at
50% capacity. (The straight-line basis is only an ap-
proximation, as can be seen by examining Figure B-4, All Pump Speeds
but the errors are insignificant.) As the water level
falls below the mid-depth, it is wise to let the control- Obviously, the pumps must be able to operate at
ler continue to operate two pumps until the total flow Points 3, 5, 7, and 9. They must also be able to
rate is, say, 35% of Qfirm so as to prevent frequently operate anywhere within the cross-hatched region.
switching pumps off and on as the water level fluctu- Other intersections can be tried, but the others all
ates around mid-depth. fall within the cross-hatched area.
Point 6 for 0.35 Qfirm should be located 0.35 d
(where d is the diameter of pipe) above the invert
(0.65 d below the soffit) for the straight line approxi- Communicating with Manufacturers
mation. From Figure B-4, it should be located 0.42 d
above the invert (0.58 d below the soffit). In Figure It is the custom of pump manufacturers to guarantee
10-30, it is located 0.6 d below the soffit. (Note that no more than one operating point, so the guaranteed
as water level rises, the line representing water level operating condition (Point 3 in Figure 10-30) should
falls in Figure 10-30 because the head falls.) Point 7 is represent the required head and capacity at the worst
y67 above Point 6. assumptions for losses. Certain other requirements
For a single pump at minimum flow rate, 0:18Qfirm , must also be met, and the following is a good example
Point 8 is 0.18 d above the invert (on a straight line for accomplishment:

M, Mfr’s H-Q curve Control system requirements:

1. A single pump operates between
M, trimmed impeller invert and mid-depth of influent pipe.
2. Two pumps operate between soffit
M, = M − (B+C) and mid-depth of influent pipe.

2 M

WL
3
=120 L
Influent pipe
Head

s tAC WL
y23 Wor 120 H
2 A C=
Invert

L
y89 5 (Runout)
145 HW
A C=
9
7 Best
y45
d

y67
4
8 6
d/2
0.8d

0.6d

s
losse
B+C
soffit

y23 y45
y89 y67
0.18Qfirm 0.35Qfirm 0.5Qfirm Qfirm Qmax
Flow rate
Figure 10-30. System H-Q curve analysis for two duty V/S pumps per Figure 10-26.
 10-10. References 10.43

. Condition A. The required flow rate at maximum known to be as flexible and to follow the principles
head is specified. (Point 3 in Figures 10-27 and stated herein as closely as PUMPGRAF21 [16].
10-30.)
. Condition B. The minimum head (Point 5 in Fig-
ures 10-27 and 10-30) is specified. The flow rate is 10-9. Complex Pumping System H-Q Curves
not specified because each pump manufacturer’s
curve unique, and to specify a second rate of flow The trend in engineering today is to depend on the
would probably limit the choice to only one manu- newer and more advanced computer programs such
facturer. This operating head should lie within the as Cybernet1 [17] or KY Pipe [18] to determine
pump’s POR at the maximum operating speed. The operating points in complex piping systems. Hence,
maximum flow rate (runout) occurs at this min- the graphical method for solving complex problems
imum head. described by Frey [19] in the first edition is omitted in
. Condition C. The required efficiency is specified at this edition.
the maximum operating speed, but the point at
which it is attained is not specified. (It is unreason-
able to expect and virtually impossible for a manu-
10-10. References
facturer to meet specified head, flow, and efficiency
at any one point.)
1. Karassik, I. J., J. P. Messina, P. Cooper, and C. C.
In most instances, runout (or near runout) repre- Heald (Eds.), Pump Handbook, 3rd ed., McGraw-Hill,
sents the most service time, so it should not be over- New York (2001).
looked when developing the requirements for pump 2. Warring, R. H., Pumps: Selection, Systems, and Appli-
cations, Trade and Technical Press, Morden, Surrey,
selection. If the pumps are not selected for runout, it
UK (1979).
is likely they will be subject to severe cavitation, 3. DIPRA, Handbook of Ductile Iron Pipe, 6th ed., Ductile
damaging vibration, and costly repetitive repairs. Iron Pipe Association, Birmingham, AL (1984).
The probable peak flow at the probable lowest head 4. Addison, H., Centrifugal and Other Rotodynamic
(Point 5 in Figures 10-27 and 10-30) will be a far more Pumps, 3rd ed., Chapman & Hall, London (1966).
frequent operating condition than Point 3). Hence, 5. Anderson, H. H., Centrifugal Pumps, 3rd ed., Trade and
select pumps so that the head and flow rate at Point 5 Technical Press, Morden, Surrey, UK (1980).
falls within the POR as defined by ANSI/HI 9.6.3. 6. Lazarkiewicz, S., and A. T. Troskolanski, Impeller
The message is clear: System hydraulic information Pumps, Pergamon Press, Warsaw, Poland (1965).
should be developed for a range of assumptions and all 7. Norrie, D. H., An Introduction to Incompressible Flow
Machines, Edward Arnold Publishers, London (1963).
potential operating conditions. Make certain that the
8. Stepanoff, A. J., Centrifugal and Axial Flow Pumps, 2nd
all operating conditions are known and considered by ed., Wiley, New York (1957).
the pump manufacturer for pump selection. 9. Turton, R. K., Principles of Turbomachinery, L. E. &
F. N. Spon, London (1984).
10. Wislicenus, G., Fluid Mechanics of Turbomachinery,
Deep Wet Wells Vol. 1, 2nd ed., Dover, New York (1965).
11. Cooper, P., ‘‘Cavitation: How does it happen?’’ Pumps
Storing the peaks of storm water runoff in long tun- & Systems, 11,16–21 (June 2002).
nels far (~ 30 m or 100 ft) below ground surface is 12. DuPont Engineering, P.O. Box 80840, Wilmington, DE
becoming popular. The tunnels vary from about 2.5 19880–0840.
to 4.9 m (8 to 16 ft) in diameter, and hence the static 13. Grohmann, M., ‘‘Extend pump application with in-
head can vary by the same amount. The total dy- ducers,’’ Hydrocarbon Processing, 59, 121–124 (Decem-
namic head varies still more, and such large vari- ber 1979).
14. Rutschi, K., ‘‘Die fleiderer-Saugzahl als gutegrad der
ations pose severe problems for designers, who must
Saugfahigkeit von Kreiselpumpen,’’ Schweizerische
be very careful and thorough in selecting pumps and Bauzeitung, No. 12, Zurich (1960).
designing systems. 15. Metcalf and Eddy, Inc., Wastewater Engineering: Col-
lection and Pumping of Wastewater, G. Tchobanoglous
(Ed.), McGraw-Hill, New York (1981).
Computer Modeling 16. Wheeler, W., PUMPGRAF21. For a free copy of this
computer program with instructions, send a formatted
There are a number of digital computer programs 1.4-MB, 312 -in. diskette and a stamped, self-addressed
that can be used to calculate pump performance mailer to 683 Limekiln Rd., Doylestown, PA 18901–
curves and pump operating points. But none is 2335.
10.44 Chapter 10 Performance of Centrifugal Pumps

17. Cybernet1, Haestad Methods, Inc., 37 Brookside Rd., 19. Sanks, R. L., G. Tchobanoglous, B. E. Bosserman, D.
Waterbury, CT 06708. Newton, and G. M. Jones, Frey’s Method in Section
18. KY Pipe, Dept. of Civil Engineering, University of 10–8 in Pumping Station Design (1st ed.), Butterworth-
Kentucky, Lexington, KY 40506. Heinemann, Newton, MA (1989).