Y

'--<''

Centrifugal -pultps
and
systenhydralrlcs

Centrifugal pumps and their associated liquid

systems are pervasive in the chemical process
industries. This report provides detailed
information on pump performance, suction
capabilities, viscosity efrects, operation at
off-design conditions, and energ-y conseryation.
Igor

J. Karassik,

Worthington

L,lost processes in the chemical process industries

(ce) involve the transportation of liquids, or their

transt'er from one level of pressure or static energy to

ano_her.

The p-mp is the mechanical means for achieving this

transport or t:,ansfer. and thus becomes an essential part
of all processes. In turn, the growth and development of
such processes are linked to the improvement of pumpng equipment and to a better understandng of how

purnps work and how they should be applied.

The centrifugal pump accounts for not less than B0%
of the total pump production in the world because it is
more suitable for handling large capacities of liquids
than the positive-displacement pump. For this reason,

ve will examine the centrifugal pump exclusively in

this article, and specifically try for a better understand_

ing of centrifugal-pump and system hydraulics.

Head and system-head curves

Pumping is the addition of kinetic and potential energy to a lqud for the purpose of movng t from one
point to another. This energ-y wiil cause the liquid to do
,ork, suc}r as ilowing through a pipeline or risins to a

higher level.
A centrifugal purp transforms mechanical eneIg-y
from a rotating impeller into the kinetic anci potential
energy required. Although the centrifugal force devel-

*59.02

bar

Di., McGraz,-Edison Ca
oped depends on both 1the peripherai speed of the impeller and the ciensity of the liquid, the amount of energy imparted per pound of liquid is independent of the
density of t}e liquid. T\erefore, for a given pump operating at a certain speed; and handling a definite volume
of liguid, the energy applied and transferred to the liquid (in ft-lbllb of liquid) is the same for any iiquid,
regardless of the density. (The only qualification to this
statement is thAt the viscositv of the liquid does affect
this energ,y, as we shail'see later.) The pump head or
energy in ft-lb/lb must, iherefore, be expressed in feet.
Within the pumping system itself, we must renember
that (1) head can be measured in various units, such as
ft of liquid, pressure in psi, in. of mercury, etc.; (2) pressures and head readin can be in gage or absolute units
(the difference beteen gage and absolute units is affected by the existing.atmospheric pressure and there_
fore by the aititude); ad (3) the pressure at any point
in a system handling liquids must never be permitted to
fall beiow e vapor presure of the liquid.
A column of cold water 2.31 ft high will produce a
pressu.re of 1 psi at its base. Thus, for water at ordinary
ambient temperatures, any presure calculated in
pounds per square inch can be converted into an eqrivaient head in feet of water by muitiptying by 2.31. For

liquids other than cold water, the coiumn of liquid

equivalent to 1 psi presure can be calculated by dividing 2.31 by the specific gravity of the liquid. The effect

Repinted from GHEMCAL ENGNEERiNG, oct. 4, 1982. copyright @ 1982 by Mccaw_Hill lnc.
1221 Avenue of the Americas, New York, N.Y. 10020

Hot water,

45o"F \232.2"cl

Gasoiine

(sp. gr. =0.822)

(sp. gr. = 0.75)

Mercu ry

(sp. gr. = 13.6)

170 ft

51.8 m

of specific glavity on the height of a column of various

liquids for equal pressures is iilustrated in Fig. 1. Formulas tbr the converson of pressure and head data are

pump centerline, it is usually spoken of as a "static suction lift." If either the suction or discharge liquid level is
under a pressure other than atmospherc, ths pressure
can be considered either as part of the static head or
eparately as an addition to the static head.

to work in terms of gage pressure, a complicated probm can occasionaily be clarified by working entirely in

Friction head

given in Table I.
Fig. 2 iilustrates the relationship befiveen gage- and
absolute-pressure readings. While it is usually feasible

-terms oI absolute pressure.

System head
In strct terms, a pump can only operate wthn a
To deiiver a given volume of liquid through
that sytem, a pump must mpart energ,y to the lqud,
made up of the following components:
l static head^
Difference in pressures on liquid surfaces.
Friction head.
I Entrance and exit losses.

The friction head is the head (expressed in feet of the

liquid being pumped) that is necessary to overcome the
friction losses caused by flow of liquid through piping,

system.

Gage pressure + Atmospheric pressure

Absolute pressure

u. s. units

.:

1]];l

1 atm = 34-ft column

static head

The static head refers to a difference in eievation.

Thus, the "total static head" of a system is the difference in elevation between the iiquid levels at the discharge and the suction points of the pump (Fig. 3). The
"static discharge head" is the difference in elevation
between the discharge tiquid level and the pump centerline. The "static suction head" is the difference in
elevation between the suction liquid level and the
PumP centeriine. If the static suction head has a nega-

:1+

]]4i:]i]i

of cold water

34 t

::--=-=Z.Jln,/psl
l4./ psl
. head in ft
osl=
2.31
Head n f =

..

^so,or_

ps X 2,31
sp, qr,

;*#*=
pressure in bur =

1a,2m/bar

h"l1i!

1o.2 ',

X sp. gr.

Headinm=barX]o,2
sp. 9r.

tive value because the suction liquid levei is below the

CHE\,IICAL

::,, !,,l

1 atmosphere = 14.7 pi

o-J

CENTRIFUCAL PUMPS

Atmospheric pressure

Any pressure above atmospheric

Absolute pressure
gage prsure +

barometric pressure

(variable with altituUe ,.,

and c imatic co1{l',l"{

..
Anv nressure below

gage pressure)
I
I

atmosoheric l
l

Barometric

a.

Suction level below pump centerlina

Atmospheric pressure

presUre
I

Abolute

] pressure

Absolute zero pressure

Static
discharge
head

valves, fittings and any other elements such as heat exchangers. These losses vary approximately as the square
of the flow through the system. They also vary with the
size, type and surface condition of the piping and fit-

tings and with the character of the liquid pumped.

In calculating friction losses, we must consider that
they will increase as the piping deteriorates with age. It
is usual to base the losses on data established for average piping that is 10 or 15 years old. These data are
readily avalable from sources such as the Hydraulic
Institute Standards and the Pump Handbook [1,2].

b. Suction level above pump centerline

Pdi""hrrg"

Entrance and exit losses

If the supply of a pump originates in a reservoir, tank
or ntake ch3mber, losses occur at the point of connection of the suction piping to the source of supply. The
magnitude of these losses depends on the design of the
pipe entrance. A well-designed bell-mouth provides the
lowest possible loss. Similarly, on the discharge side of
the system where the discharge lne termnates at some
body of liquid, the velocity head of,the liquid is entirely
lost, and must be considered as part of the total friction

Total

static
head

-tI
I

Static
suction

losses o[ the system.

head

System-friction and system-head curves

As mentioned earlier, friction, entrance and

exit

losses vary approximately as the square of the flow

through a system. For solving pumping problems, it is
convenient to show the relatonship between capacty
and friction-head losses graphically. These losses are
B6

,,l,"
discharge
head

oHEMICAL ENGINEERING oCToBER 4,

c. Suction and discharge levels under pressure

,,

l9B2

:""-, ,
a, ":,
^.
] ;;

,_"!

Head

\.'
G

'9!,

curVe

/
\_

SVsten f rictiol] curve

rlcton
losses

Static pressure or head

Capacity, O
PUmp's
...p9u.yu
']ea.l-."^_

do

u3

a2

a hdx)

Capacity, O

System-head curve.

1r'

o
o
T

Capacity, O

Capacity, O

therefore calculated at some predetermned flow, either

expected or design, and then calculated for all other
flows, using the square of the flow relationship. T'he resulting curve is called tlre system-friction curve, as
shown in Fig. 4a.
When we combine the static heads, pressure differences and friction-lread losses of any system and plot
them against the capacity, the resulting curve (Fig. ab)

is called the system-head curve. Superimposing a pump

head-capacity curve at constant speed on tls systemhead curve (Fig. ab) will permit us to determire tlre
capacty at the pont where the two curves ntersect.

This is the capacity that will be delivered into the system by that pump at that particular speed.
For systems having varying static heads or pressure
differences, it is possible to construct curves correspond-

ing to the minimum and maximum conditions,

as

shown in Fig. 5. The corresponding intersections witl

the pump's head-capacity curve will then determire the
minimum and maximum flows that the pump will deliver into the system.

variations in desired flow

I
do

o
T

.u.*IlEd

lrur'rr.
p static head
l

2|
E1

)
.E

X
E

Capacity, O

'l

Minimum

It is unusual for a system to require operation at a

single fixed capacity. Generally, the process served by
the centrifugal pump is variable in its demand. A given
pump operating in a given system will only deliver that
capacity corresponding to the intersecton between the
head-capacity and the system-head curves. In order to
vary the capacity, t becomes necessary to change the
shape of either one or both curves.
The pump head-capacity curve can be changed by
operating the pump at variable speed (Fig, 6a). (For a
guide to variable drives, see "Making the proper choice
of adjustable-speed drives," Chem. Eng., Aug. 9, l9B2, p.
46.) Or the system-head curve can be altered by creatirg friction loss tlrrough a throttling valve (Fig. 6b).
Obviously, the difference between the total head de-

(lllEN,t]CA|. llNGlND[,RINc oc]ToBliti 4.

9B2

o.7
ol

CENTRFUCAL PUMPS

100

60
c-50
{40
do
o
s30

*2O

1B0

+ 160
140

o 120
o
E ,l00
-6

|-

80

2oo

|,

p,lp J",r!

"ffi

_o

Capacity, m3/h
300
|

4o0
,,

500

++\*l

90
80}c
70 .9
a

60EtlJ
50

40o
30n

60

20T
O
103o

40

10

i9

_Speed: 1,760 rpm

20

2 4 6 8 10121416182022

c
,i
o

0o-

t00

otr
JJ

;9

50

40=
y

30

('
j

20g
10

P90
E

.6

o
=85
,E
!
o

,3
c,

cc

Capacity, hundreds of gpm

80

75

veloped by the pump and tlre head required by the

system-lread curve represents wasted energy lost n the
throttling process. On the other hand, the majority of
centrif-gal pumps today are drven by constant-speed

Affinity laws

practces.

speed.

squirrel-cage induction motors, and throttling the

punp clischarge is the only practical means of obtaining the desired variable capacity. As will be seen later in
tlris article, the coming revolutonary move to variable-frequency motor drives is about to change these

Affinity laws and rating curves

The hydratlic performance o[ a centrifugal pump

involves three basic parameters: (1) the capacity (expressed in units of volume per unit of time such as
cpm), (2) the total head (expressed in feet of the liquid
pumped), and (3) the speed at which the pump runs
(generally in rpm).

Normally, pump performance is presented in the

form of curves such as shown in Fig. 7, with the headcapacity curve plotted at a fixed speed. The curve also
shows the brake lrorsepower required at various flows,

and the corresponding pump efficiency. The capacity at

which the pump performs its function most efficiently is
called the "b...p." or best efficiency point.
The useful work done by a pump is the weight of
liquid pumped in a period of time, multiplied by the
head developed by the pump, and is expressed in terms
of }rorsepower, called water horsepower,I4/HP.It would
be more correct to refer to WHP as liquid horsepower,
which can be determined from:

q'!!;-')
wHp 3,960

()

where: WI-IP = water horsepower; Q - pump capacty, gpm; and H = total head, ft.
Tle power required to drve the pump is the water
horsepower clivided by the pump efficiency,7l.Hence,
dividig Eq. (l) by 1 gives:

BHP=-OIr(.b.
3,960

80

85

Calculated impeller diameter,

pr.\

(2)

4
c}lEtrtc.

90
%

95

100

of original diameter

The relationships that allow us to predct the performance of a pump for a speed other than that for
known pump characteristics are referred to as the "affinity laws." When the speed is changed:
1. The capacity, Q, for any given point on tle pump
characteristics varies directly a the speed, z.
2. T}e lead, H, varies as the square of the speed.
3. The brake horsepower, P, varies as the cube of the
ln other words, if Subscript l is assigned to the conditions under which the characteristics are known, while
Subscript 2 denotes the condtons at some other speed,
then:

*:?,

?:(*)'

(3)

=(3)",?:(?)'

$)

3,=(?)",

These relationships can be used safely for moderate

changes in speed. Eq. (3) may not. be as accurate for
large speed changes.
Similar affinity laws exist for changes in impeller diameter, D, within reasonable limits of impeller cutdown. In other words:

Dr. H2
Q, Dr' H,
Qz _

some deviation from these laws occurs even with rel-

atvely modest cutdowns, Fig. B shows the recommended cutdown related to the theoretical cutdown,

Specific speed

Tle principle of dynamical similarity when applied

to centrfugal pumps expresses the fact that two pumps

geometrically similar to each other will have similar

performance characteristics.
The term "specific speed" is the concept that links the
three main parameters of the performance characteristics-capacity, head, and rotative speed-into a single
term. The mathematical analysis used to establish the
relationship between the specific speed of a pump and
its operating characteristics need not concern us..In its

basic form, the specific speed is an index number, expressed as:

N"=!#

(5)

in w|rich: N" : specific speed; n : rotative speed, rpm;

Q = capacity, gpmi and H - head, ft (head per stage
lor a multistage pump).
Eq. (5) renains unclranged whether the inrpeller is
single-suction or double-suction. Therefore, it is customary when referring to a definite va]ue of specific
speed to mention the type of impeller.
While rve could calculate the specific speed for any

Capacity

Capacity

Capacity

80

given operating condition of head and capacity, the

definition of specific speed assumes that the head and
capacity used in the equation refer to those at the best
efiiciercy of the pump. The specific-speed number is

?, o

o
,6

A
tlJ 60

Specifc speed: N= n,/7x"o

U.S. units: n = rpm;O = gpn; H
Metric units: n =rpmiO = m3/s;B

independent of the rotative speed at which the pump is

operating,

We cannot overenphasize the fact that "specific

speed" is an index number-a concept quite similar to a
"family name" that is useful in identifying various
characteristics of a group. Just as the Browns, the Wil-

N s |met ri c|
/V{u,s.l/5'1.6
=

Specific speed, /y',,r9

sons or the Smiths may be said to have certain hair or

eye coloraton, certan general common features, so do

pumps of the sanre specific speed have a number of

cl,aracteristics that distinguish them from pumps having other specific

speeds.

For instance1 the plrysical characteristcs and the gen-

eral outlire of impeller profiles are intimately connected to their respective specific speeds. Thus, the
value of the specific speed will immediately describe the
approximate impeller shape, as shown in Fig. 9. Similarly, the specific speed of a given pump will be definitely reflected in the shape of the pump characteristic
curves, as shown at the top of Fig. 9. While some variation in the shape of these curves can be made by
clarges in the desig of tle inpeller and casing waterways, the variation that can be obtaned without adversely affecting pump efficiency is relatively small.
Another parameter affected by the specfic speed s
the maximum efliciency obtainable fron pump impellers of different specific speeds and different sizes, also
indicated in Fig. 9.

approximating that of a single-suction impeller having

(1 /2)|/', or 70,77o that of the double-

a specific speed of

suction impeller.

Rating curves

Rating curves are commonly reproduced in pump

bulletins and sales literature for standard lines of
pumps. A rating curve for a centrifugal pump shows in
a condensed form the possible range of applications of
that pump at some rated speed for a range of impeller
diameters (Fig. 13). A different chart is generally made
available for each motor speed for a particular pump.

Typ. characteristics for a pump

I[ tlre operating conditions for a pump at the design
speed (that is the capacity, head, efficiency and power
input at which the efficiency curve reaches its maximum) are taken as the |00% standard of comparison,
the head-capacity, power-capacity and e fficiencycapacity curves car all be plotted in terms of the percentage of their respectve values at the capacty for
maximum efficiency. Such a set of curves represents the

In addition, the rating curyes generally show a curve for

the required net positive suction head (NP.S11). To facilitate the selection of a pump, rating curves of an entire
line cf similar pump are prepared.

suction conditions

"type characteristic" or "l007o" curve of the pump.

Most centrfugal-pump troubles occur on the sucton

it is imperative to understand how to
relate the suction capability of a centrifugal pump to
the suction characteristics of the system in which it will

Tl,e l00% curves of pumps having specific speeds of

2,000, 4,000 and 10,000 are shown in Fig. l0. These
curves can be used to predict the approximate shape of
a pump's c}aracteristics once the specific speed of that
pump is known. 'lo avoid interpolation, the curves of
Fig. l l and 12 show the change in head and power in
relation to specic speed for single-suction impellers. A
double-suction impeller will have a type characteristic
cll II]N1l(lAl, DNGlNEliR

side. Therefore,
operate.

When pumping liquids, the pressure at any point

within the pump must never be permitted to fall below
the vapor pressure of the liquid at the pumping temperature. There must always be sufficient energy available
I

Nc oCToDEIt

4. | !B2

B9

ft

CEN,fRIFUGAL PUMPs

j
c

130

130

A]"

zo

.3 ,ll0

oo

Ego
a
.E
80
X

WF
1

Ezo
ts60
-s50

80 ,:

.e'_

70Ey
60: O

50a

40 ,'
30b

f20
o
o

;s

a./Y,

10

o
t

-z,ooo (U.S.

,l

80 100 12o 140

20 40 60

20ao
105

units|

=,-1o (metric)]

try,

Ratio of capacity to capacity at maximum efficiency,

o.

160

160

160

150

,l

140

140

130

,l

o 10
E 100
f
E
90

50

j
o
c
o
,6
120 E
o

so
20

110

?o

,l00
^/er

90

80

B0

o
!
o
o
E
o
E
@
o

60
40

40

o
o
,F

30

30

o
cf

7o

50

7o

20
,l0

60
50

b.Vs

20

40

[tr" =

60

20

-4,000 (U.t i. units)

-80

10

(metri c)]

80 100 120

l40

Ratio of capacity to capacity at maximum efficiency,

300

300
280

2o0

_a,7*
,

]ao

jQ

100

o
*o
.9

r6

E
f

.E

iB0 ox
E

i60 o

160

oo

o
}
o

240 A
o
220

2oo

140

o
o
}
o
o
o

'O

22o

Eo

x6

260 Oc

.P 24o

y
C
c?

E
*

o/o

320

?60

E
.E

32o
zeo

,;

tc
90! c
100

s
o40
3o 0
,F

-:

120 2
a
110,6

z.

zo

'oh

e80

140 o
3
120 o
o
o

100
B0

;60

60

Eao

(r

c.

20

=
^/s
[&.

20

40

60

&10,000 (lJ.S. uni ts}

-200 {me tric} l

B0 100 120

140

Ratio of capacity to capacity at maximum eff iciency,

90

40
20
0

o
B
o

o
o

.9

o
E

at the pump suction to get the liquid into the inpeller

and to overcome the losses between the pump-suction

nozzle and the impeller inlet. At this point, the impeller

vanes can act to add energy to the liquid;
Normally, pump performance is presented n the
form of curves (Fig. 14), with the head-capacity curve
for the pump plotted at a fixed speed. The curves also
show the brake horsepower required at various flows,
and the corresponding efficiency.
An additional characteristic of the pump is the
(NPS11) required. This is the energy in feet of liquid
head required at the punp sucton over and above the
vapor pressure of the liquid, to permit the pump to

deliver a given capacity at a given speed.

Changes in available (NP,Sf1) do not affect the pump

performance as long as available (NP"1/) remains

above the value of required (NPS). However, when
available (NPSH) falls below the required value, the
pump begins to cavitate and "works in the break," as
shown in Fig. 14, The characteristics in solid lines are
for values of (NP,Si/) available in excess of (NP.11) re-

quired. If available (NP.l1) falls below tlrat requiredas for instance if at 1,800 gpm the (NP,S1/) available
falls below 17 ft-cavitation starts, and the pump
produces less head. Some increase in capacity occur
with a further reduction in head, until about 1,970 gpm
is reached; then further reduction in head causes no
increase in capacity (see curve with broken line).

suction head and suction lift

As defined by the Standards of the Hydraulic Institute, the suction head, i", is the statc head on the
pump-suction line above the pump centerline, minus all
friction-head losses for the capacity being considered
(including entrance loss in the suction piping), plus any
pressure (a vacuum being a negative pressure) existing

in the suction supply.

Rather than express the suction head as a negative
value, the term "suction lift" is normally used when the

pump takes its sucton from an open tank under atmospheric pressure. Because the suction lift is a negative
suction head measured below atmospheric pressure, the
total suction lift (symbol also i*) is the sum of the static
suction lift measured to the pump centerline and the
friction-head losses as defined above. (It is sometimes
advantageous to express both suction and discharge
heads in absolute pressure, but usually it is more convenient to measure them above or below atmospherc
pressure.)

A gage on tlre suction line to a pump, with its readngs corrected to the pump centerline, measures the
total sucton head above atmospheric pressure minus
the velocty head at the point of attachment. Because
suction lift is a negative suction head, a vacuum gage
will indicate the sum of the total suction lift and velocty head at the pont of attachment.
The three most common suction-supply conditions
are illustrated in Fig. 15.
Case

involves a suction supply under a pressure

other than atmospherc, and located above the pump

centerline. It includes all the components of the suction
head, i.. If i, is to be expressed as a gage reading, and
P* is a partial vacuum, the vacuum expressed in feet of

CIlINIICAl, DNGINEERlNG OCTOBER 4, l9B2

j
O
c

,O

.a

o 200

_P

f
F

E 150

E
o
o
s
o
T
a
o
s
o
.9

ca

140

zoo

150

130

120
110
100

90E
80

l_

o_

l-

?0t

100
20

tv1
OO
OO
Oo
oi

e'

Specific speed

10

=
!
o

OC)
OO
O_ O_
<i, 0

0L

liquid would constitute a negative pressure head and

carry a minus sign. If the pressure, P", is expressed in
absolute pressure values, i" will also be in absolute pres-

sure values.
Case II involves a suction supply under atmospheric
pressure, located above the pump centerline. Because
the suction head (expressed as a gage value) has a P,

value of zero, the P, value can be dropped from the

formula given in Fig. 15.
Case III involves a suction supply under atmospheric
pressure, located below tlre pump centerline. It is opticnal whether the suction head be expressed as a negative suction head or in positive values as a suction lift.

50

Because the source of supply is below the pump centerline (which is the datum line), S is a negative value. The
suction-lift formula is the same as that for suction head,
except that both sides have been multiplied by ( - l). A
gage attached to the pump sucton flange, when corrected to the pump centerlne, wll regster a partal
vacuum or negatve pressure.
To deternine the suction head, it is therefore necessary to add the velocity head to this negative pressure
algebraically; or, if it is desired to work in terms of a
vacuum, the velocty lead must be subtracted from the

vacuum to obtain the suction lift.

For example, if the gage attached to the suction side

100

1B0

o
,6
o
E
f

E
X
o

E
O

oo

o
3
o

\|l
\ft
!l
f

60

g#>
6810

10,000

12 14

16,1

CHENlICAL DNGINEDRING oc,IoBER 4,

l9B2

,9

30o.
20j

-|P

B2022

Capacity, hundreds of gpm

Specific speed

.O

lJ

ll

l
5.000

[*i

60

2.000

effic iency

EB0
o

3
o

o
o
o

Point of

120

140

80
7oic

91

5
!

10 :

nUGAL

Case

-Suction from source under

PUI\,!Ps

of a pump having a 6-in. line and pumping at a capacity of 1,000 gpm of cold water showed a vacuum of 6 in.
Hg (equivalent to 6.8 ft of water), the velocity head at
the gage attachment would be 2.0 ft of water, and t}re
suction lread would be -6.8 f 2.0, or -4.8 ft of water,
or the suction lift rvould be 6.8 - 2,0 or 4.B ft o[ water.

pressure other than

atmospleric, and Iocated above pump centerline.

Net positive suction head

Point

A'

terl i ne

B-"hr=S-hrr-h,
Point

Case

l l

+P

-Suction from source under atmospheric

nd locateci above pump centerline.

pressure,

The use of "permissible suction lift" or of "required

sucton tread" has definte shortcomngs. Ether term
can only be applied to water, because it refers to the
energy of barometric Presure expressed ir feet of water.
Changes in barometric pressure, whether caused by differences in elevation above sea level or by climatic conditions, affect the value of tlrese terms. Changes in
pumping temperature also affect these values, since
they affect tlre vapor pressure of the liquid,
For this reason, all references to suction conditions
today are made in terms of (NPS//)-net positive suction head above the liquid vapor pressure.
Botlr suction head and vapor pressure should be expressed in feet of liquid being handled, and must both
be expressed in either gage or absolute pressure units. A
pump handling 62"F water (vapor pressure of 0.6 ft) at
sea level, witlr a total suction lift of 0 ft, has an (NPS11)
of 33.9 - 0.6, or 33.3 ft, whereas one operating with a

15-ft total suction lift lras

33.9

Point

Case l l l -Suction from source under atmospheric pressure,

and located below pump centerline.

h,

hr,

or

1B.3 ft.

an

(NP.11) of

l,,,-h,
+

lent to the vapor pressure correspondng to its temperature), this energy has to exist entireiy as a positive head.
If the liquid is below its boiling point, the suction head
required is reduced by the difference betwee tlre pressure existing in the liquid and the vapor pressure correspondng to the tenperature.
It is necessary to differentiate between available net
positive suction head, (NPS1I), and required net posi-

tive suction head, (i/P,Sl/)o. The former, ,vhich is a

characteristic of the system in which a centrifugal pump
works, represents the difference between the existing
absolute suction head and the vapor pressui^e at the prevailing temperature. The (NP.11)", which is a function
of the pump design, represents the minimum required
margn between the sucton head and vapor pressure.
The manner in which (NPSII)A at a given capacity
should be calculated for (l) a typical installation with a
suctio lift, (2) a pump taking its suction fron a tank,
and (3) a pump handling a liquid at the boiling point is
demonstrated in Fig. 16.
Botl (NZSIJ), and (NPS1l)n vary with capacity, as
shown in Fig. l7. With a given static pressure or elevation difference at the suction side of a centrifugal pump,
(NPSH)^ is reduced at larger flows by the friction losses
in tlre suction piping. On the other hand, (NP,S//)R,
beig a function of the velocities in the pump-suction

Entrace loss at Point A

Total friction loss from Point A to Point B
h," - Velocity head at Point B
/r, (gage reading at Point B corrected to punp
centerline) = l,- hu,
P., ft of liquid

9,2

15,

A
operating on suction lift will landle a certai maximum capacity of cold water without cavitatior. The (NPS) or amount of energy available at the
suction nozzle of such a pump is the atmospheric pressure minus the sum of the suction lift and the vapor
pressure of the water. To handle the same capacity with
any other liquid, the same amount of energy must be
availab]e at the suction nozz|e. Thus, for a liquid at its
boiling point (in otler words, under a pressure equiva-

h,=S-h,,-h

h, {suction head) = (-S) (--h,)(suction lift) = S I l,.

- 0,6 pun-p

CL]

l]vIl(]Al, ]aN(;l NIaERlNG O(j,],()l}li,lt 4. l 9B2

-lNpSH| A =

p,

z _ nr

pressure over liquid surface. psia

vapor pressure of liquid, psia

Static head, ft

/r

Friction losses, ft

sea level:

|NPSH|A-

At

pu

At

?}1Jt:&\

Water, 80"F
P, = 0.5 psia

2'31

Point B

(L4J-0,5) _
15_3= 14,Bft

5,000 ft above sea level:

{NPSHIA
a. suction

15

3 = 9.0 ft

lift

(=5psig

{NPSHI =s

(& _-P,p}

(h

--

r" + h i|

Pno= Vapor pressure of liquid at pumping temperature

hr. = Friction losses in suction losses from point A to point B
h, = Entrance loss at Point A
All units are to be expressed in feet or meters

=J
=

o
E
o

D
a
o
T

-Water, B0" F
P, = 0,5 psia

ricti on

hr=4it-

Capacity
2.31 |Ps_ Pvl

lNPSHlA _

sp.qr.

,| ? A
'-lF

Ps = pressure over liquid surface, psia

Pn = Vapor pressure of liquid, psia
Z = Static head, ft
= Friction losses, ft

lNPSH\A _

2.31 (14.7_+

S-O.S) +
10_4

passages and at the inlet of tle impeller, increases basi-

50.3 ft

b. Suction from pressurized tank

vanes, area between these vanes, shaft and impeller hub

diameter, impeller specific speed, and the shape of tlre
sucton passages-enter n some form or another nto
the determination of (NPSH)n. Different designers may
use dfferent methods to produce an mpeller that will
perform satisfactorily with a specific value of (NP./I)E.
As a result, t is not recommended that users attempt to
estimate (ilP,Sf/)fi from the knowledge of just one or
two of these factors. Instead, they should base their selections on data provided by pump manufacturers.

P, = 37,5 psig

ffia

Wl

Friction
hr

n-Butane at 1O0"F ,',

Pn = 52.2 psia
sp. gr. = 0.56

losses,tt

2i

Specific-speed and suction limitations

|NpsH\A=ffi!*z-n,
P,
Pn

h,
|NpSHl A

= Pressure

Vapor pressure of liquid, psia

Static head, ft

Friction losses, ft

|37,5

Specific-speed-limit charts have been prepared and

published by the Hydraulic Institute [1] for several
types of pumps:

over liquid surface, psia

J.!J -

cally as the square of the capacty.

A great many factors-such as eye dameter, sucton
area of the mpeller, shape and number of mpeller

52,2l +
10 _ 2= 8.0 ft

c. Suction from liquid at boiling point

l
I

Double-suction.
single-suction, with shaft through eye of impeller.
Single-suction, overhung-impeller.
Single-suction, mixed- and axial-flow.
Hot-water, single-sucton and doub]e-suction.
condensate pumps, wth shaft passng through eye
of impeller.
One such chart, containing specific-speed limits for

CIlEi\4lCAI- DNGIND]RING ocToI}ER 4,

t9B2

93

cENTRlI,UcAL PUMPs

tion-limitation charts suggests that the specific speed

indicated corresponds to the point of maximum efficiency. Yet this is the intended meaning. If a pump is
applied for conditions near to its capacity at the best
efficiency, there would be little error introduced by
using the rated conditions to determine chart limita-

4,000
3,500

HbrI- ,|
'

3,000

tion. On the other hand, if the rated and best-efEciency

conditions were to differ sgnficantly, the chart recommendations would be found to apply only to the best
efEciency point.

LPo.*o;
J ? -F

O
O

Suction specific speed

2,000

Application of the specific-speed-limit charts, as originally developed, had a very important shortcoming,
i.e., satisfactory suction conditions were tied directly to
the total head developed by the_pump. The perform-

Eo
O

o_

U,

1,000
900
B00

400

200 150 100 B0 50 40 30

Total head (first stage), H, ft

300

sigle-sction overhung-impeller pumps, such as the

ANSI pumps, is slrowr n Fg. 1B.
It s mportant to remember that these charts are
strictly empirical. In using tlem, it must be realized
that pumps built for the limit allowed are not necessarily the best design for the intended service and that a
lower-specific-speed type miglt be more economical.
It nust also be realized tlrat tle design of individual
pumps controls the application of the specific-speed

linit for maximum head ard suction conditions. For

exanple, the n,aximum recommended specific speed

fbr a double-suction, single-stage pump is 1,990 for a
200-ft total head and a 15-ft suction litt. It does not
follow that all double-suction, single-stage pumps of
1,990 specific-speed type are suitable for operation at
speeds that will cause them to develop a 200-ft total
head (at maxinum e{ficiency); nor that tlre pump, f
suitable lor operation at a 200-ft total head, is suitable
for operation with a l5-[t suction lift; nor that a pump
of this type operating against a 200-ft total head would

on test be found capable of operating on only a 15-ft

maxin,um suction li[t.

These charts are intended to indicate only tlre maximum rotative speed for which experience has shown a
centrifrgal pump can be designed with assurance of
reasonable and proper operation for the combination of
operating conditions. The Hydraulic Institute Stand-

ards suction-limitation charts should be considered

guidelines.

Nothing in the Hydraulic Institute Standards suc-

94

ance of an impeller from the point of view of cavitation

cannot be affected too significantly by conditions existing at the impeller's discharge periphey. Yet, these conditions are the prme factors in determnrg the total
head that the impeller will develop.
In other words, if an impeller exhibits certain suction
characteristics, cutting down its dameter within reasonable limits, and thus reducing its lread, should have
o influence o its suction capabilities. Since the total
head, H, is changed, a strict interpretation of the specific-speed-limit charts would indicate that unless the
suction lift is to be proportionately altered, the maxirnum permissible specific speed must be changed.
This inconsistency was finally resolved by the devel-

opment of the suction specific-speed concept. It is essentially an idex number, descriptive of the suctiol charis defined as:
acteristics of a given

::ud

(h")"n

(6)

where: S = suction specific speed; 'n = rotatve speed,

rpni; Q : flow, gpm. (For single-suction impellers, Q is
the total flow. For double-suction impellers, Q is take
as one-half of the total flow.); and i" = required
(NPS11), ft.

The specific-speed-limit charts (such as in Fig. lB)

have been revised several times since they were first
adopted as a guideline for centrifugal-pump suction
conditios. Unfortunately, they are still based on the

erroneous concept that the total head developed by the

pump plays a part in determining the maximum permssble rotatve speed for a given set of sucton conditions, despite the recognition of the concept of suction
specific speed in the Hydraulic Institute Standards. The
charts in the Standards today are based on values of S
ranging from 7,480 to 10,690. These values vary witlrin
each chart, as well as from chart to chart. It is hoped
that the charts will be revised and simplified so as to
make them easier to use. They will then resemble the
chart for the hot-water pump (see Fig. 19), in which the
required (NPS/1) can be read directly for any given flow

at various

speeds,

Cavitation and pump performance

Cavitation occurs when the absolute presure within
an inpeller falls be]ow the vapor presure of the liquid,
and bubbles of vapor are formed. These bubbles col-

CIl]].NlICAl. IaNGlNDnItlN(; otj'fOI}El( 4,

t9B2

lapse further out along the impeller blades when they

reach a region of higher pressure. The minimum required (NP.taI) for a given capacity and a given pump
speed s defined as that dfference between the absolute

suction head and the vapor pressure of the liquid

pumped at the pumpng temperature that s necessary
to prevent cavtaton.
Pump cavitation becomes evident when there is one
or more of the following signs: noise, vibration, drop in
the head-capacity and efficiency curves and-with
time-damage to the impeller by pitting and erosion.
Since all of these signs are obviously inexact, it became

necessary to agree to apply certan ground rules so as to

establish some uniformity for detecting cavitation.

The minimum (ffPSf/) is determined by a test in
which both total head and efficiency are measured at a
given speed and capacity under varying (NP,S) conditions. The results of such a test appear in a form similar
to that in Fig. 20. At the higher values of (NPS11), head
and efficiency remain substantially constant. - As

500
400

Sino la- uctior

ump_

ll

300

.;
a

l.

200

lI
I

lll t*t=]
,":s'l
L
tr*--.od
fi
;ped,
J^

s
co

,;:1oo

-l
*:

;80
o

:io

I,,l

.l

oo

i+o

zo

F;

20

10

lllll

L.tr

ource: H,ydraui

ll

l.

6 81O

lnst iu

Capacity, hundreds of gpm

lan( ja ,d

20 304050

(NPS11) is reduced, a point is finally reached where the

curyes break, showing the impairment of pump per-

formance caused by cavitation. The exact value of

(NPS11) where cavitation starts is difficult to pinpoint.

IJsually, a drop oA 3%i the head developed is taken as

evidence that cavitation s occurring, For that particular speed and the capacity being tested, the (NPSf1)
that produces a 3Vo drop in head is stated to be the

minimum required (niPS11).

(NPS14 tests of centrifugal pumps are normally carried out with cold water. Both the Hydraulic Institute
Standards curves and pump manufacturers' rating
curves indicate (NPS14 requirements for cold water.
Thus, it might be assumed that (NPSi| required by a
centrifugal pump for satisfactory operation is independent of the liquid vapor pressure at the pumping
temperature. Actually ths s not true.
Laboratory and field tests run on prrmps handling a
wide variety of liquids, and over a range of temperatures, have always shown that (NPS1) required for a
given capacity and with a given pump apparently varies appreciably. For example, the required (NP,Si4
when handling some hydrocarbons is frequently much
less than that required when the pump handles cold

water tests. The theory underlying this effect s fairly

siple bt -eed not be discussec.l in detail l,ere. It is
based on tire fact that nild and partial cavitation can
take place in a pump vithout causing extrenely urr[a-

vorable effects.
T}re degree of interference with the proper operation
of tlre pump caused by minor cavitation will bear a
definite relationslrip to the temperature of the liquid
hadled by tle pump. Wirer we say that a punp is
cavitating, we ITean tlat somewhere rvithin tl-e confines
of the punp, the pressure will }ave fallen below the
vapor pressure of the liquid at the prevailing tenperature. Thus, a small portior of t}e iiquid }and]ed by tle
pump wiil vaporize, and tlris vapor will occupy cosiderably more space witlrin the impeller than did tle

equivalent -ass of liquid.

If the pump is hadling water at normal tempera-

tures, tle vo]ume of a bubble of steam is tremendously

larger tlran tlre volume of the original quantity of the

water. Even when pumping water, there is definite evidence that required (NPSf4 decreases when the water

temperature ncreases.
Altogether it became evident that the reduction n
required (NPSI' nust be a function of the vapor pressure and of the characteristics of the liquid handled by
the pump. Thus, it was felt that rules could be developed to predict the effect of liquid characteristics on
required (NP,Si|.
Such rules have been developed by the members of
the Hydraulic Institute and incorporated in its Standards. We shall examine these rules, but before doing
this, let us consider the effect of temperature on required (NPS4 for watern as this may help us better
understand the effect of other liquids.

performance with water

Net positive suction head, H"

It has been noted that pumps handling hot water

seem to require a lower (NPS14 than shown by cold-

cllEIuICAL ENcINDERING oCToBER

4, l9B2

95

CliNTlltIlUGAl, PUMPS

being based on accepted theories. It was believed that a

reduced (niPsr ftr service

1,000

witl hydrocarbon fluids

could be justified for two reasons:

1. Oil companies' specifications generally called lbr a

500
400
300

maximum capacity and head at minimum (NP,S|. In

practce, it was unlkely that these two requrements
would be imposed simultaneously. In [act, some field
conditions are self-regulating-for instance, low capac-

10

200

150

5+

4i

,l00

3.=o

.o

Eso
a40
930

,E

,5i

1.0

0.5

zo

R
=

10

ity occurs at low (NP,fO, as a result of a reduced flow in

tle system. Under these conditions, even if pump capacity falls off, (NPSf1) is increased and equilibrium is
eventually attaned.
2. Cavitation with hydrocarbons was not as severe as
with water, i.e., the head-capacity curve does not break
off suddenly because (a) only the liglrter fractions will

boil first, and (b) the specific volume of hydrocarbon

vapors is very small in comparison witl that of water

vapor.

Obviously, this does not tell the whole story. Many

other factors affect the behavior of a pump landling
hydrocarbons at low (NP,f|. Tlus, attempt to arrive

at a more reasoned understanding continued, and led to

a conversion chart for hydrocarbons by the Hydraulic
Institute (since updated).The latest correction chart is
incorporated in the 1975 edition of the Hydraulic Institute Standards [1] (see Fig. 21). To use this chart, enter
with the pumping temperature and proceed vertically
upward to tle vapor pressure. From this point, follow
along or parallel to the sloping lines to the right side of
the chart, where reduction in (NPSf| may be read. If
tlris value is greater than one-half of (NP11) required

o
Temperatu re, F

water. For instance, at 50"F, one pound of water occupies 0.016 ft3, while steam at tlre same temperature oc-

cupies 2,44L ft3. Tlre ratio of the two volumes is

152,500. This ratio diminishes as water temperature
increases. At 2l2"F, ore pound of water occupies

0.0l67 ft3, and one pound of steam, 26.81 ft3, so that the
ratio of volumes is only 1,605-almost one lrundred
times less than at 50"F. Thus, the higher the temperature of the water, the greater the reduction in (NPS@,
tlrat can be permitted for tle same degree of effect on

pump performance.

Performance with hydrocarbons

Pump applications for hydrocarbon processes frequently impose restrictive linritations on available
(NPSH). On the otlrer hand, it was found in the past
that variations between tlre required (NPS@ when

handling hydrocarbons and when handling cold water

were generally favorable. These circumstances led both

pump designers and designers of refineries to direct

their efforts to understanding tlre plenomena and establishing rules that could be applied to predict the e[fect of any special liquid characteristics on the required
(NPS14 of any centrifugal pump.
At first, it was thought that these variations did not
exist and that if true vapor pressures or "bubble-point"
pressures were to be used i- the calcu]ations of test
(NP,S14, the discrepancies would disappear and there
would be complete correlaton with water-test cavitation data. Corrections for (i/PSt| with hydrocarbons
were nevertheless used, as a matter of polcy rather than
96

with cold water, deduct one-half of the cold-water

(NPSi| to obtain corrected required (NPS|. if the
value read o the chart is less than one-half of coldwater (NPS|, deduct ths value from the cold-water
(ilP.r| to obtain corrected required (NPS14.

Because of the absence of available data demonstrating (NPS/4 reductions greater than 10 ft, the chart has
been limited to that extent. Extrapolation beyond that
limit is not recommended.
Warnings are included in the Hydraulic Institute
Standards regarding the effect ofentrained air or gases.
This can cause serious deterioration of the head-capacity curve, of the efficiency, and of the suction capabilities, even when relatively small percentages of air or gas
are present.
An exhaustive analysis of the phenomena that take
place in a pump handling hydrocarbons is beyond the
scope of our discussion, Such an aralysis would at best
be open to argument, because several somewhat conflctng interpretatons stll exist wth respect to what

actually takes place.

It is probably best to use the correction factor for the

reduction in (NP.i| as an additional safety factor

rather than as a license to reduce available (NPSia).
This is a personal opinion, but one that is shared with a
number of rotating-machinery specialists of some of the
major petroleum and petro-chemical companies.

Inadequate suction conditions

When a system offers insufficient available (NPS|
for an optinum pump selection, tlrere are several ways
to deal with the problem. We can either find means to

C]}IDI\{lcAL ENGINDERINC oCToBl1ll 4. l9B2

increase available

(NPSa, or means

to reduce required

(ffP,Sf|, or do both.
To increase available (NPSI/), we can:
l. Raise the liquid level.
2. Lower the pump.
3. Reduce the friction losses in the suction piping.
4. Use a booster pump.
5. Subcool the liquid.

To reduce required (NPS|, we can

6. Slower speeds.

use:

A
B. A

double-suction impeller.
larger impeller-eye area.
9. An oversize pump.
10. Inducers ahead of conventional impellers.
11. Several smaller pumps in parallel.
Each of these methods presents advantages and disadvantages. We shall examine and evaluate these methods individually:
7.

l.

B0

70
'

25a

|.i

ll
]De i."n;;;

6
].,: o
i:;iiF

a:I*|

150

i,]il

:ili]

ll

a! l.*
hea

60

E .L

;e

u
c
O
,6

uJ

'"ro'u"''|*

50

ificie,ncv,
;;l,:,j,

30
+

|n

diJ

10

400 4B0 560

leel-At first glance, this appears to

be the simplest solution unless it is impractical because
(a) the liquid level is fixed, as in the case of a river, a
Raise the liquid

o_

640

Capacity, gpm

pond or a lake; (b) the amount by which the level must

be raised is completely impractical; or (c) the cost of

raising a tank or a fractionating tower is excessive. Frequently, it wIl be found that only a few extra feet may
permit the selection of a less-expensive or more-efficient
pump, and the savngs n first cost, energ-y or mantenance will far outweigh the additional costs incurred.
2, Lower tle PumP-Just as in the case of raising the
liquid level, the cost of lowering the pump may not be
as prohibitive as one might imagine because it may permit the selection of a higher-speed, less-costly and
more-efficient pump. An alternative approach may be
to use a vertical pump with the impeller located below
ground level.
The penalty for this solution is that the pump bearings may have to be lubricated by the liquid being
pumped. While successful bearing designs and materials have been developed for this purpose, it should be
understood that the pump life cannot compare with the
life obtainable from external bearings that are either
grease or oil lubricated. Thus, one should expect morefrequent scheduled overhauls with this method.
3. Reduce PiPing friction /osses-]This is recommended
under any circumstance; the cost of doing so will be
easily repaid both by improved suction conditions and
by savings in energy.
4. Use a booster PumP--This solution is particularly effective in the case of pumps intended for high-pressure
service, where the resulting permissible higher speeds
will yield great savings in first costs of the main pump,

higher efficiencies, and frequently a lesser number of

stages-which in itself leads to greater reliability. The

booster pump can be selected as a low-speed, low-head

pump of single-stage design.
5, Subcool the liquid-:Ihis approach increases.'available (NPS@ by reducing the vapor pressure of the liquid

being pumped. It is most readily accomplished by injecting liquid taken from somewlere in the stream
where it is available at a colder temperature. In many

cases, particular|y at hgher pumpng temperatures, the

amount of injected colder liquid is very small. As an

example, if we are pumping water at 325"F, the injecCI

tion of on|y 4Vo of 1 75 " F water will subcool our stream

to the point that available (NPS| will have been increased by 20 ft.
6. Use slower sPeeds-Once a reasonable value of suction specific speed has been selected, it becomes obvious
that the lower the pump speed, the lower the required
(NPSfl. The problem is that a lower-speed pump will
be more expensive and less efficient than a higher-speed
one selected for the same service. Thus, lowering the
pump speed will seldom prove most economical.
7. Use a double-suction imPeller-Particularly for larger
capacities, whenever a double-suction impeller is available for the desired conditions of service, this presents
the most desirable solution. It is based on the following:
If we select the same . value for both single- and
double-suction impellers such that:

"

nr(Q)'/'

(Hr,r)3/n

n"(Qr)"'
(Hur)3/n

(7)

where Subscript 1 refers to a single-suction impeller and

Subscript 2 refers to a double-suction impeller.
Since Q, - qr/z, we can assume first that:
flz : fl7

in which

case,

Hrrr:

0.63 H"rr, or;

Hrrz =

in which

case, n2

(8)

H"fl

(9)

|.414nr.

By keeping the pump speed the same in both cases, as

in Eq. (B), we can reduce (NP,SiQj? by 27% if we use a
double-suction impeller. Alternatively, with a given
(^/P,SfO, as shown in Eq. (9), we can operate a doublesuction pump at 4|.4%higher speed.
8. Use a larger imPeller-eye area-This solution reduces
required (ffPS14 by reducing the entrance velocities
into the impeller, These lower velocities may have little
effect on pump performance at or near the pump's best
efficiency point. But when such pumps run at partial
capacity, this practice can lead to noisy operation, hy-

lEi\rICAI- l]NCtNDl,.lttNC oC,I,oIJlilt 4. l9B2

97

f
!

C\

rRl|UG4!j!!!L---draulic surges and premature wear. This problem will

be discusseJ in greater detail later in the article. At this
Pont, suffice to say that this is a dangerous procedure
and should be avoided if possible.

9. {Jse an oaersized Punl-Because (NPS14 required by

a pump decreases as the capacity is decreased, a larger
pr-p nu., would otherwise be applied to the service is
occaionally selected. This practice is risky and can lead
to undesirable results. At best, the penalty is the choice
of a more expensive pump that operates at a lower efficiency than might otherwise have been obtained (see
Fg. 22). At worst, the operaton at a lower percentage
of tlre best efficiency flow can lead to exactly the same
as the use of excessively enlarged impeller-eye
::.}J."-.
10.

{Jse

an indcer-An inducer.is a low-lead axial-

type impeller with few blades, which is located ir front

of a conventional impeller (Fig. 23). By design, it re-

quires considerably less (NPS| tlran a conventional

impeller, so it can be used to reduce (NPSQ requirements of a punp, or to let it operate at higher speeds
with given available (NPS14.
The inducer is an adequate answe, for many situations but must be applied with care, as the permissible
operating range of pumps with inducers is generally
narrower than with conventional impellers.

||.

1B0

160

ea

D,140

\ \

'_ 12o
,o 100
80

&
,i

V
!;16

*,

?O

7o

60

50{
405,6

-'.tr ssu

c or1o

30EtJJ

o,oJo.rJ

,ri

400 Sst
l

00

sst,--,

l\.

o
0

100

2,00l) ssl

iU

200

300

Capacity, gpm

9B

400

500

Viscosity and entrained gases

Earlier, we stated that pump performance is indeof the characteristics of the liquid being
pumped, but qualified this by adding that the liquid
viscosity does affect performance. This is because two of
the major losses in a centrifugal pump are caused by
fluid friction and disk friction. These losses vary with
the viscosity of the liquid being pumped, so that both
the head-capacty output and the meclrancal output
differ from the values they have when the pump hanpendent

dles water.
The performance of a pump tested first on water (vis-

9nwer

in larallel-Obviously,

shall see later.

"o 1S,
i:

4,00( SSL

,,oJo

20

L-Sr.,

80

b10

a%

ls/ ,(L_.

-g
-o

^|

?o^"c

(ooo,

60

40
=uo
o
a30
.9

j
,

Use seueral smaller PumPs

smaller-capacity pumps require lower (NPSi1) values.

While this appears to be a costly solution, it is not necessarily so. In many cases, three half-capacity pumps, of
which one is a spare, are no more expensive than one
full-capacity pump plus its spare. As a matter of fact, in
many cases, two half-capacity pumps may be installed
without a spare, since part-load can still be carried if
one pump is temporarily out of service. In addition, if
the demand varies widely, operating a single pump during lighlload conditions will conserve energy, as we

600

cosity - 32 Saybolt seconds universal, or SSU), and

then on a variety of iiquids having viscosities ranging
from l00 to 4,000 SSU, is shown in Fig. 24. It is evident
that by the time the viscosity reaches 2,000 SSU, pump
performance wll have deteriorated to such an extent
tlat a positive-displacement pump will be more economical for the application.
It is not necessary to present here a complete dscussion on the effect of viscosity on the flow of liquids.
Flowever, all correction factors for viscosity effects on
pump performance have been developed exper.imen-

CtlEN,tIcAL I]NGINEERING OCTOBER 4, l9B2

ffi,

10o

E80
o
a
o

!
f

60

=
o

100

a
o

c O
j:
O O
o -

.E

(J

6
F

go

E60
o
,6

340

Liquid capacity, gpm

o
O

co
?
o
o

O
,6

aria*"pz"t,"{""ffi

4 6 810 1520

40

o
o
3

6080100

Capacity, hundreds of gpm

Capacity

Hydraulic lnstitute standards

i
l

tally. The Hydraulic Institute Standards include two

charts that are unversally accepted for correctng capacity, head and efficiency values obtained from pump
tests conducted wth water. One chart apples to pumps

having capacities of 10 to l00 gpm, and the other

(Fig. 25) for capacities from l00 to 10,000 gpm. For
further details on the method of applyng these correc-

ton factors to tests wth water, and on the selecton of a

pump for given performance conditions at a given viscosity, consult the Standards [l].

Entrained air or gas

If entrained air or gas is permitted to enter a centrifu-

gal pump along with the liquid, the performance of

the pump will be unfavorably affected. The most frequent way that ar enters the pump sucton s by vortex
formation at the free surface of the liquid. On occasion,
air leaks into the pump through the pump stuffing box
if proper precaution has not been taken to seal it properly. The amount of air or gas that can be handled with
impunity by a centrifugal pump is probably in the
range of %%bv volume (measured under suction condi-

tions). If the amount is increased to 6Vo, the effect is

almost disastrous, as can be. seen from a typical curve
shown in Fig. 26. The dotted line indicates the minimum capacity at which the pump can be operated. The
reason for this minimum is that if the pump capacity is
reduced further than indicated, air or gas can no longer
be partially swept out through the discharge, and the
pump becomes air-bourd.

Operation at off-design conditions

Theoretically, as long as available (NPS| is greater
than required (NP,S/|, a centrifugal pump is capable of
operating over a wide range of capacities. As explained
earlier, exact operating capacty is determined from the
intersection of the pump's head-capacity curve with the
system-head curve. This operating capacity can only be

changed by altering either one or both of these

curves-varying the pump speed will alter the headcapacity curve, while throttling the pump discharge
will alter the system-head curve.
At any given speed, the performance of a centrifugal

CHEMICAL ENGINDERING ocToBER 4,

9B2

l
}

!
I

l
,

CEN,I,RI,UGAL PLIN4PS

O
c

.9
a

!
o
T

o
o
,i6

t(/)
o_

Capacity

pump is at its optimttm at only one capacty point,

i.e., the capacity at which the efhciency curve reaches its

nraximtr. At all other flows, the geometric configuration of the impeller and of the casing no longer provides
an ideal flow pattern. Therefore, our definition of "offdesign" conditions must be any conditions wherein a
pump is req-ired to deliver flows either in excess or
below the capacity at the best efficiency.

Operation at high flows

There are two circ,mstances that might lead to the

operation of a pump at flows in excess of its best efficiency or even of its design point.
The first of these occurs when a pump has been oversized through the use of excessive margins in specifying
both head and capacity. Under these circumstances, the
pump perfornance and its relationship to the systemhead cuve might look as in Fig. 27 , The head-capacity
curve ntersects the system-lread curve at a capacty
much in excess of the required flow, using excess power.
Of course, the pump can be throttled back to the required capacity and tl-e power reduced somewhat. But
if, as frequently is the case, the pump runs uncontrolled,
it will always run at the excess flow indicated in Fig. 27.
Unless sufficient (ffP"14 has been made available, the
pump may suffer cavitation damage and, of course,
power consumption will be excessive.
The second occurs when two or more pumps are used
in parallel and one of them is taken out of service because demand has been decreased. Fig. 28 describes the
operation of two such pumps. Whenever a single pump
is running, its head-capacity curve intersects the system-head curve at flows in excess of design capacity.
This is called the "run-out" point. Here, again, available (NPSifl and the size of the driver must be selected in
such a manner as to satisfy conditions prevailing at this
run-out pont.

Operation at low flows

The rnost frequent cause for running a pump at reduced flows is the reduction in demand by the proces
100

served by the pump. Flowever, it can also happen that

two pumps operating in parallel may be unsuitable for
this service at reduced flows, and one of the pumps on
the line may have its check valve closed by the higher
presure developed by the stronger pump.
Operation of centrifugal pumps at reduced capacities
leads to a number of unfavorable results that may take
place separately or simultaneously, and must be antici-

pated or circumvented. Some of these are:

l operating at less than best efficiency-when reduced flows are required by the characteristics of the
process, they can be handled by using a variable-speed
drive, or by using several pumps for the total required
capacity and shutting down the pumps sequentially as
total demand is reduced. This last procedure will save
energy, as will be described later on in this report.
Higher bearing-load-If the pump is of the single-volute design, it will be subjected to a higher radial
thrust that will increase the load on its radial bearings.
A pump that s expected to operate at such flows must
be able to accept this higher bearing-load.
l Temperature rise-as capacity is reduced, the temperature rise of the pumped liquid increases. To avoid
exceeding permissible limits, a minimum-flow bypass
must be provided. Tlis bypass, whiclr can be made aulomatic, wili also protect against the accidental closng
of the check valve whi]e the pump is running.
l Internal recirculation-At certain flows below that
at best efficiency, all centrifugal pumps are subject to
internal recirculation, in both the suction and the discharge area of the impeller. Tlris can cause hydraulic
surging and damage to the impeller metal sinilar to
that caused by classic cavitation, but taking place in a
different area of the impeller

Internal recirculation
The first three unfavorable effects of operation at low
flows are well understood and need no further explanation. On the other hand, the subject of nternal recircu-

lation has been little understood outside of a small

number of pump designers. Because internal recircula-

cHDl\.{lcAL ENGINEERING ocToBER 4,

t982

,,l,,

.,

lmpact probe-__

c
o
F
o

90

o
!

80

.O

100

E=

Suction

o
,a

I
I

6
f

,6

60

,F

50LIl

o
c
o

Normal flow

R ec

rcu

lation

flow,

7o

6,000

':

8,000 10,000 12,000 14.000 16,000 18,000

Suction specific speed,., at best efficiency flow
(single suction or one side of double uction)

tion at the stction is nrost frequently the cause of field

problens, we s}ral] now discuss it.

Tle exact flow at which suction recirculation takes
place depentls on impeller design. The larger tle impeller eye dianeter and the larger the area at tlre impeller
suctio relative to its overall geometry (therefore, the
Iower the required (NP,SH) at a given capacity and
speed), tl,e [igler the capacity at which recirculation
takes place as a percentage of the capacty at best efficiency (see Fig. 29).
Internal recirculation causes the formation of very
intense vortices (Fig. 30) witlr lrigl velocities at their
core and, consequently, a sgnficant lowerng of the
statc pressure at that locaton. In t,rrn, ths leads to
intense cavitation acconpanied by severe pressure pulsations and nrise, and this can be damaging to the operation of the pump and ultimately to the integrity of
the impeller material.
Location of the material damage is an excellent way
to identify whether the cause is classic cavitation or internal suction recirculatio, If the damage is on the visible side of tlre inlet inpeller vanes, tlre cause is classic
cavitatio. lf the danage is to the hidden side (pressure
side) of the vanes and must be seen with the help of a
small mirror, tle cause is suction recirculation.
FitlC :bLt:r-: tr,:_*td b; tb.L phtnonenan obvio,:shoccurred from time to tire. But in the early 1960s, they
became intensified. Two factors led to this greater incidence of trouble:
1. More pressure was exerted by the users to have

pump desigers reduce required (NPS14 values. This

could only be achieved by increasing the impeller suction-eye areas, bringing the onset of internal recirculation closer to the best-efficency capacity.
2. Higher heads per stage, and larger pump capacities, were increasing the energy levels of individual impellers, intensifyig the unfavorable effects of internal

recirculation.
Information on the phenomenon was first released in
1972 n a lmted-crculaton paper and later, more
widely, in an article [3]. For obvious reasons, the matle-

matical solution was considered proprietary and was

not published rntil 198l [4,5].
Interim guidelies, however, were indicated suggesting that uction specific-speed values (S) not be allowed
to exceed a range of 8,500 to 9,500 so as to avoid field

problems

if

operation at significantly reduced flows

were contemplated.

Fraser |4) gave exact formulas for calculating the

flow at whicl internal recirculaton would start at the
sucton, once certan geometrc data were known about
an impeller. He provided some close-approximation
curves (of which Fig. 31 is one) in the event that these
data are not readily available [5].
There is a way to confirm the value of the suction
recirculation flow. An impact-head probe is installed
into the direction of flow, as indicated in Fig. 30, with
the probe directed into the eye. Under normal flow, the
reading will be essentially the suction pressure less the
velocity head at the eye. As soon as internal recirculation occurs, a flow reversal will show as a sudden rise in

pressure.

The suction specific speed, S, of the pump must always be calculated for the conditions corresponding to
the capacity at best efficiency. Guaranteed conditions of

service may or may not correspond to this best-efEcienc fl.ot,-the seldora do- S rn-*t al_*o he calcutated
on the basis of pump performance with the maximum
impeller diameter for which the pump is designed.
This constraint becomes obvious when we consider
that internal recrculaton at the sucton occurs because
of conditions that arise in and around the impeller inlet.
These conditions are not necessarily affected by cutting
down (i.e., reducing) the impeller diameter. Cutting
down will move the best efficiency point to a lower flow
value, but will not reduce the flow at whiclr suction

recirculation wil] occur.

Liquid characteristics do not affect the flow at which

internal recirculation takes place. But they have a profound effect on the severity of the symptoms and on the
extent of the damage, The reason is exactly the same as
in the case of required (NP.ll) to prevent the symptoms

ctlEMlCAL DNGINEl]RlNo ocToBER

4, 9B2

101

20,000

CI]NI'RIFUG\l- PlIMPS

ad damage caused by classical cavitation, as described

earlier in tlris article. Thus, internal suction recircula-

tion will always be less damaging when the pump handles lrot water, and particularly when it handles hydrocarbos, than ,hen it handles cold water.
Now that means are available to calculate the onset
of internal recirculatior, users and designers of centrifugal pumps should be in the position to establish sensible
linits for mirimum operating flows. To do so, it will be
necessary to establish some guidelines between mininrum flows and recircrlation flows. But there's t}re rub!
Setting up such guidelines is a fairly complex task for a
rumber of reasons.

Operating a pump at flows below the recirculation

flow leads to a variety c,f events-leading to unfavorable effects on pump pcrformance and on the ultimate
lile of tlre impeller. We can lunp all these events under
the term "distress." In turn, the degree o[ distress will
depend on a variety of factors, such as:
Size of the pump-i.e. capacity, total head and

horsepower.

value of the suction specific speed.

Fluid cl-aracteristics.
Materials of construction.
Lengtl o[ time the punrp operates below certain
critical flows.
The degree of tolerance the pump user permits to
the symptoms of distress exhibited by the pump.
The last factor makes the choice of minimum-flow
guidelines a sub.jective one. Some users will unquestionably accept the fact that a impeller may have to be
replaced every year; others will complain if an impeller
on exactly the same service lasts only three or four years.
Similarly, noise and pulsation levels perfectly acceptable to one user are cause lor conplaints by otler users.
One observation may give some guidance to users on
what constitutes an acceptable minimun flow with respect to the effect of the. value. Referring to Fig. 31

and assuming a hub-to-eye diameter, hl/D1, ratio of

, value o[ l4,000 will have
ts suction recirculaton occur at about l00% of its best

0.45, an impeller having an

efficiency flow. If dealing with a fairly substantial pump

handling cold vater, one should be most reluctant to
consider running it below this 100%. On the other
}and, a pump with the same hub-to-eye-diameter ratio
and an S value of 8,000 will have a recirculation flow of
ooly 567o of best efficency flow. I would have no hesitation running this pump when necessary a,t as little as
257o of best efficiency flow.

Vague as they may appear to be, here are some

guidelines:

1. Unless there is a compelling reason to do so, do not

specify (NPSH) values tlat result in ,S values much

above 9,000.
2, In the case of relatively small pumps-say under
100 hp-the effect of stcton recirculation is not apt to
be as significant as for larger pumps.
3. Pumps haidling hydrocarbons can be operated at
lower flows than similar pumps handling cold water.
4. Tlre risks o[ operating at flows much below the

recirculation flow can best be determined after the

pump is in operation. Provision should, therefore, be
made to increase the minimum-flow bypass if there is a
102

suspicion that too optimistic a decision as to this mininum flow was made at tlre time the pump was selected.
5. When the pump s not expected to operate at flows

below ts design conditon [such as coolng-tower

pumps operating in parallel, at constant speed and
unthrottled (see !'ig. 2B)], higher S values can be used
without concern for unfavorable effects from internal
sucton recrculaton.

Energy conservation and pumps

The high cost of energy and the scarcity of fuels have
become hard facts of life. More than ever, it is imperative to examine pump selection and operation with a
view to minimizing power consumption. We might address ourselves to the pump designers and ask them
what further improvements they may make in pump
efficiencies. Alas, the easy improvements have already
been incorporated in today's centrifugal pumps. The
curves in Fig. 9 show the maximum efficiencies attainable from commercial pumps o[ different sizes and at
different specific speeds. They have not changed materially sirce l955.

Effect of specific speed

The higher the specific speed selected for a given set
of operating conditions, the higher the pump ef{iciency
and, therefore, the lower the power consumption. Bar-

ring other considerations, the tendency slrould be to

favor higlrer specific-speed selections from the point of

view of energy conservation

Let us consider some typical examples (see Table II
and III) wlerein several alternative selections are examined for two sets of conditions. In both cases, it has been
assumed that the desgn point corresponds to the best
elficiency point of the pump,
In tl-e first case (Table II), the difference between selections (2) and (3) does not appear to be significant
enough to warrant considering a three-stage pump. On
the other hand, a two-stage pump saves 20.5 hp, which
at an energy cost of $300 to aool(hp)(yr) will produce

annual savings of $6,150 to $B,200.

Aganst these savings, we must consider a number of

counterweighing factors. Among these: (1) tle initial

cost of a 2-stage pump wll be higher than that of a

single-stage pump, and (2) a two-stage design will pre_
clude the use of a simpier overhung end-suction punrp
with a single stuffing box.
Similar factors must be weighed against the savngs
of 7.4 hp, or $2,220 to 2,960/yr, which favors a twostage pump in the second case (Table iII).
It may well be that the higher-specific-speed pumps
will slow sulficient savings to.justify their selecton. But
are tlese savings really there? The answer to this ques-

tion will depend entirely on the expected operatingcapacity range of these prmps. The shape of the power

consumpton curve vares consderably wth the specfic

speed of the pumps, as illustrated in Fig. 12. Before we

can decide whiclr of the possible selections is best in

terms of energy consumpton, we must examine power

consumption not only at the desgn point but also over

the entire range of capacities that t}re pump will encounter in service.

CHEN{ICAI. ENGINEERING ocToRDR .t, 9B2

Let us expand our analysis of the case examined in

Table III. In addition to the power at the 100% design
capacity, we shall compare the power consumption for
single- and two-stage pumps at 757o and 5O% flow, (see
laUle lV;. Much to our surprise, we see that instead of
saving energy at all flows, the two-stage pump uses 7,6
hp more at 75% flow, and lB hp more at 50%.
To establish the real energ"y balance between the two
selections, we need to predict the subdivision of operating hours at various loads. If we assume that this subdivision will correspond to that shown in Table V, we find
that the most efficient pump s not the best selecton
and that the single-stage pump will save 36,267 hp-h
yearLy over the two-stage pump. Obviously, the final
answer will always depend on the load factor for a
given installation.
Once the optimum combination of specific speed and
(ffP.f1) requirements has been established for any
given set of operating conditions, we still must choose
between various pump offerings that may differ somewhat in guaranteed efficiencies.
Our most obvious reaction might be to look with
greater favor

on pumps having higher

efficiencies,

favoring those that might exceed others by as little as

/"% or |Vo. A1| things being equal, there is some logic to
this approach. But, all things are not always equal and
small differences in guaranteed efficiencies may have
been obtained at the expense of reliability, by using
smaller running clearances or lighter shafts.
Savings in power consumption obtained from these
small differences in efficiency are insignificant when we
compare them with other approaches for reducing or

eliminating wasteful power consumption, While we

cannot examine every approach available, there are
three particular sources of significant power savings
that I wish to ana|yze in some detail. These are (a)
savings of power wasted by oversized pumps, (b)
savings that may be obtained by operating the minimum number of pumps in multiple-pump installations,
and (c) savings that can be obtained by restoring internal clearances at the right time.

Effects of oversizing
One of the greatest sources of power waste s the practice of oversizing a pump by selecting design conditions
having excessive margins in both capacity and total
head. It is strange, on occasion, to see a great deal of

attention being paid to a one-point difference in efficiency between two pumps, whle at the same tme potential savings of 5, 10 or even |SVo of the power are
ignored through an over-conservative attitude n selecting the required conditions of service.
Still, it is true that some margin should always be
included, mainly to provide against the wear of internal
clearances that with time will reduce the effective pump
capacity. How much margin to provide is a fairly complex question, because the wear that will take place varies with the type of pump, the liquid handled, the severity of the service, and a number of other variables. We
shall examine this question later when we analyze the
savings in power consumption that can be realized from
restoring internal clearances to their original values.
A centrifugal pump operating in a given system will

deliver a capacity corresponding to the intersection of

its head-capacity curve wi the system-head
prgvided t}at avaiiable (]VPr} i equal to or exceds
required (NPS|. To change this operating point requires changing either the head-capacity curve or the
system-head curve or both. The first can be accomplished by varying the speed of the pump or changing
the impeller diarneter, while the second requires altering the friction losses by throttling a valve in the pump

discharge.
In the majority of pump installations, the driver is a

CHEMICAL ENGINEER]NG ocToBER

4, l9B2

103

CENTi(II|UGAI PUMPS

of suclr a pump, with

posed on Curve A.

,40

14"/u'in. im}.d|^.

,20

H >ad::

****1-1

-)00

J4]]=J"q"]

lB0

its head-capacity curve and Curve B, the pump will

absorb 165 blp.
Tlus, the pump has been selected with too mucl
margin. We can safely select a pump with a snraller
impeller diameter-say 14 in. The head curve for the
smaller impeller will intersect Curve A at 2,820 gpm,

l60
l40

90

]20
.]00

80

-Stati

lBo

gvng us about 4Vo rnargn n capacty, whclr s suffi-

7o

IrrT

l60
140

l-f

|20

l00

508

.g

40E
tll

80

60 -"

Vo<
''r(Oi

30

1,000

2,000

3,000

Our savings may actually be even greater tlan

4,000

used

to change the punrp capacity. Thus, if we have provided too much excess margin in the selection of the

punp head-capacty curve, the pump wll have to operate witl considerable throttling to limit its delivery to
the desired value. On the other hand, if we permit the
pump to operate unt}rrottled, which is more likely, flow
into the system will increase until that capacity is
hed where the system-head and head-capacity

L-ufl/eS intersect.

'Ihe old "rules of tl-umb" on selecting pipe sizes and

valve sizes should be questioned. We slould evaluate
care[ully the economy of investng n larger ppes and
valves lraving lower frictior losses, against the logterm power savings resulting from these lower friction
losses.

Example of oversizing
Let us exanine a pumping sytem for which tlre max-

imum desired capacity is 2,700 gpm, static head is l l5

ft, and total friction losses, assuming 15-yr-old pipe, are

60 ft. Total lread required at 2,700 gpm is 175 ft.
We can now construct a system-head curve, shown as
Cuve A in Fig. 32. If we add a margin of about |0% to

the desired capacity and, as is frequently done, add

sone margn to the total head above tlre system-head

curve at this rated flow, we end up by selecting a pump
for 3,000 epn and 200 ft total head. The performance

i04

we

have shown. In many cases, the pump may be operated

unthrottled-the capacity beng permtted to run out

Capacity, gpm

constant-speed motor. l'his is the last means that

cient. We will still have to throttle the pump slightly,

and our system-head curve will become Curve C. The
power consumption at 2,700 gpm will now be oly 145
bhp instead of the 165 bhp required with our overiy
conservative selection. This is a respectable |27o saving
in power consumption. Furthermore, we no longer need
a 200-hp motor; a 150-hp motor will do quite well. The
saving in capital expenditures is another bonus from

not oversizing.

20

i0
4o

|a3/n-in. impeller, is superim-

This pump develops excess head at the maximum

desired capacity of 2,700 gpm. If we wish to operate at
that capacity, this excess head will have to be throttled,
as shown by Curve B.
At 3,000 gpm, the pump will take 175 bhp and we
will have to drive it with a 200-hp motor. At the desired
capacty of 2,700 gpm, operatng at tlre ntersecton of

-l^"]_-|--|-|J{
lzL.""_J-l-ll

,.l,"]+",|J

to the intersection of the head-capacity curve and

Curve A. If this is so, a pump with a l43fn-in.-dia. impeller would operate at approximately 3,150 gpm and
take 177 bip. If a 14-in. impeller were to be used, the
pump would operate at 2,820 gpm and take 14B bhp.
We could be saving over |6Vo in power consumption.
Our real margin of savings is actually greater than
indicated. The friction losses we used to construct the
system-head curve (Curve A) were based on losses
through 15-yr-old piping. The losses through new piping are only 0.613 times the losses we have assumed.
The system-head curve for new piping is shown as
Curve D. If the pump we had originally selected (with a
|43fn-in. impeller) were to operate unt}rottled, it would
run at 3,600 gpm and take 187.5 bhp. A pump with a
14-in. impeller would intersect the system-head cuwe

at 3,230 gpm and take 156.5 bhp, with a saving of al-

most 16.5%.
As a matter of fact, we could use a 13,3/n-in. impeller.
Its head-capacity curve would intersect Curve D at
3,100 gpm and take 147 bhp, now saving 2|.6%.
Important energy savings can be made if, at the time
of selecting the conditions of service, reasonable restraints are exercised to avod usng excessve safety
margins for obtaining the rated service conditions.
But what of existing installations in which the pump
or pumps have excessive margins? Is it too late to
achieve these savings? Far from it! It is possible to accurately establish the true system-head curve by running a
performance test once the pump has been installed and
operated. A reasonable margin can then be selected,
and tlrree options become available:
l. The existing impeller can be cut down to meet the
conditions of service required for the irstallatio.
2, A replacement impeller with the necessary re-

CrIINllOAl, DNClNll]ItINc oc,l,()DF,I{ 4. 002

duced diameter may be ordered from the pump manufacturer, The original impeller is then stored for future
use if friction losses are ultimately increased with time
or if greater capacities are required.
3. n certain cases, there may be two separate impeller designs available for the same pump-one of which
is of narrower width than the one originally furnished.
A narrow replacement impeller may then be ordered
from the pump manufacturer. Such a narrower impeller will have its best efficiency at a lower capacity than
the normal-width impeller; it may or may not need to

be of smaller diameter than the original impeller, depending on the degree to which excessive margin was
originally provided. The original impeller is put away
for possible future use.

Variable-speed operation
While the majority of motor-driven centrifugal
pumps are operated at constant speed, some pumps
take advantage of the possible savings in power con-

sumption provided by variable-speed operation.

Wound-rotor motors were once frequently used for
this purpose, but present practce generally nterposes a
variable-speed device such as a magnetic drive or a
hydraulic coupling between the pump and the electric
motor. Alternatively, we can use a variable-voltage,
variable-frequency motor control that has the great
advantage of mantanng essentally a constant motor
efficiency, regardless of operating speed. On the other
hand, a variable-speed device such as a hydraulic coupling has slip losses, so that its efficiency decreases directly as the rato of output speed to nput speed. Variable-frequency operation has an additional advantage:
It permits an electrc motor to operate above the synchronous speed as well as below it.
Variable-speed operation makes it possible to meet
the required conditions of service without throttling, by
reducing the pump operating speed. Variable-speed
drives-especially by means of variable-frequency
input-will be used more frequently in the future.
It is not necessarily too late to achieve power savings
in some existing installations by converting them to
variable-speed operation. To decide whether to make
such a modficaton, t s necessary to plot the actual
system-head curve, to calculate the speed required at
various capacities over the operating range, and to determne the motor horsepower output over ths range,
includirg losses, if any, incurred in variable-speed devices. The difference between this horsepower and the
pump brake horsepower at constant speed represents
potential power savings at these capacities.
It is then necessary to assign a predicted number of
hours of operation at various capacities, and to ca]culate the potential year|y savings in hp-h or kWh, These
savings are translated into cost savings and can be used
to determine whether the cost of the modfication to
variable-speed operation is .justified.

Run one pump instead of two

Many installations are provided with

so-called

"half-capacity" pumps, with two pumps operating in

parallel to deliver the required flow under full load. If
the service on which these pumps are installed is such

tlat the required flow varies over a considerable range,

important power savings may be available through

improved operating practices. Too often, both pumps
are kept on the line even when the demand drops to a
point where a single pump can carry the load (see Fig.
28). The amount of energy wasted in running two
pumps at half-load when a single pump can meet this
condition is significant. This can best be demonstrated
by referring to a so-called 100% pump curve, as shown

in Fig.

10a.

To simplify our analysis, let us neglect the question of

capacity or pressure margins and imagine that the

pumps carry their full load with throttling valves wide
open. We shall also assume that these pumps are operating at constant speed. Thus, full-load conditions are
met by running two half-capacity pumps, each operating at its lo07o capacity point and each taking l0OTo of
its rated brake horsepower.
If we want to reduce the flow to half load and still
maintain both pumps on the line, it will be necessary to
throttle the pump dscharge and create a new ystemhead curve. Under these conditions, each pump will
deliver 50% of its rated capacity at ||?% rated head,
much of which will have to be throttled. Each pump
will take 72.5% of its rated power consumption. Thus,
the total power consumption of two pumps operatng
under half-load conditions would be 145% of that required if a single pump were to be kept on line.
Instead, we could stop one of the pumps, and carry
the half-load requirement, with a single pump running
at 100Vo of rated capacity. The discharge would be
throttled considerably less than if both pumps were left
running. Power consumption would be |00Vo of the
rated value for a single pump. Running a single pump
when the process load is down to 50% results in a power
saving of 3|vo over running both pumps,
As a matter of fact, a single pump could carry much

more than the capacity corresponding to half-load,

since the head-capacity culve of a single pump might
intersect with the system-head curve at anywhere from
6O to 70% of load, depending on the exact shape of the
system-head and the head-capacity curves.
Other benefits are to be had from such an operating
practice. In the first place, if we assume 8,500 year|y
operating hours, of which 2O7o take place at flows of
5a7o of rnaximum flow or lower, each pump will operate
for only 7,650 h/yr instead of 8,500, extending the calendar life of all the running parts by over l17o.
Pumps that frequently operate at reduced capacities
do not have as long a life as purnps operated at close to
their best efficiency point. Thus, running only one

pump whenever it can handle the required flow will

add much more life to each pump than just the differ-

ence in operating hours.

Restoring internal clearances

The rate of wear of pump parts at nternal clearances
depends on many factors. To begin with, it increases in
some relation to the differential pressure across the
clearances. It also increases if the liquid pumped is corrosve or contains abrasve matter. On the other hand,
the rate is slower if hard, wear-resisting materals are
used for the parts subject to wear. Finally, wear can be

ctlE]\.{lcAl ENGINEERING ocToBER 4,

l9B2

i05

C-E,NTRIFUGAL PUNtPs

lniti,,

:\
H

'"ud

cutve

"ir"uud

10

a
a
cy
L.=

"r..'"

aerr_\O\

-".rur..Oa
''P/

3
o\
oa)
Ec
o.9

H6

Eh,

^ AL
oro]-

lncrease in leakage,

\
Al J

oO
-n
oh
xo'o
o
J

500
Capacity

accelerated f momentary contact between rotatng and

statonary parts occurs durng pump operaton.
Wlen the running clearances increase with wear, a
greater porton of the gross capacty of the pump s
slrort-circuited through the clearances and has to be
repumped. The e{fective or net capacity delivered by

the pump against a given head is reduced by an

amount equal to the increase in leakage.
While in theory the leakage varies approximately

with the square root of the differential pressure across a

runnng jont and, tlerefore, wth the square root of the
total head, it is sufficiently accurate to asume that the
increase in leakage remains constant at all heads. Fig.
33 shows the effect of increased leakage on tlre shape of
the head-capacity curve of a pump, Subtracting the
additional internal leakage from the initial capacity at
eacl head gves a new head-capacity curve after wear
las taken place.
We must compare the cost of restorng tle nternal
clearances after wear has taken place with the value of
the power savings that we may obtain from operating a

pump laving the original clearances. This cost is relatively easy to determine. We can obtain prces on new
parts and estmate the cost of labor requred to carry

ut

the task. But how about the savings?

These savings are not the same for every pump. Both
analytical and experimental data have indicated that
leakage losses vary with the specific speed of a pump.
Fig. 34 shows the relationship between the leakage
losses of double-suction pumps and tleir specific speeds.

1,000

,500 2,000 2,500 3,000

Specific speed
,l

If, for instance, we are dealing with a pump having a

specific speed of 2,500, the leakage loss in a new pump
will be about |Vo. Thus, when the internal clearances

will have increased to the point that this leakage will

have doubled, we can regain approximate|y |Vo in
power savings by restoring the pump clearances. But if
we are dealing with a pump having a specific speed of
750, it will have leakage losses of about SVo.If the clearances are restored after the pump has worn to the point
that its leakage losses have doubled, we can count on
5vo power savings.

Thus, restoring the clearances of the lower-specific-

speed type pumps yelds greater return n terms of tle

reduction of leakage losses. In addition, lower-specificspeed pumps generally have higher heads per stage
than higher-specific-speed pumps. All else being equal,

wear i increased with higher differential heads. Therefore, we will generally find more reason to renew clearances of high-head pumps, and more savings.

Stevcn Danatos, Editor

References
l. "Hydraulic
Ohio,

t975.

lnstitute Standards," l3th ed., Hydraulic Intitute, Cleveland,

2. Karassik, L J., Krutzch, W. C., Fraser, W. H., and Messina,J. P., "Pump
Handbook," McGraw-Hill, New York, l976.
3. Bus|, A. R., Fraser, W. tl., and Karassik, I. J., Coping Witl Pump Progres:
The Sources and Solutions of Centrilugal Pup Pulsations, Surgo ad
Vibrations, PumP Wotll Worthington, Mountainside. NJ., Sunmer 1975
and Marclr l976.
4. Fraser, W. H., Recirculation in Ce ntrifugal Pumps, Winter Annual \,Iating
of ASME, Nov. [6, l98l, ASME, New York.
5. Fraser, W. H., Flow Recirculation in Centrifugal Pumps, Turbomachinery
Syrnposium, Texas A&M Univenity, College Station,ltx., Dccember l98l.

The author
lgor.}. Karassik is chief consulting
enginrer, Worthington Div.. McGrawdison Co., 233 Mount Airy Road,
Bmking Ridge, NJ 07920. He u,as
appoined to 1he position in Deamber
l976 atter spcnding 42 yeam in various

moacitic at lvorthinston on

1he

."J.u-h, design and pplietion of

single- and multistage pumps. He has
written more than 450 articles on
pumps, and is the author or coauthor of
lour books and a handbook on pump.
He las a B.S. and M.S. from Carnegie
Institute of Technology, is a Life Fellow
of ASME, and a profesional engineer in
Nev .}eney. I le is also a member o[ Tau
Bea ?i, Pi Tau Sigma and Sigma Xi.

2-2-3-5-I:0B3

C]IlENtoAL ENGINDtrRrNC oc-l,oBDR 4, l9B2