The Influence of Viscosity on Centrifugal Pump Performance - 199_1

8/11/2019 The Influence of Viscosity on Centrifugal Pump Performance - 199_1

1/16

h

nfluence of

Pump

Viscosity on

Performance

PAPER

NO. A-45-57

entrifugal

By ARTHUR

T.

IPPEN

SUMMARY

HE wide use of centr ifugal pumps in

{he

oil industry dc

their performance characteristics for oil be pre

dICted with reasonable assurance. A systematic

study

of

performance changes with increasing viscosities

had not

been

undertaken so

far

under controlled laboratory conditions.

For

the

purpose of undertaking such a

study

theoretically

and

experi

mentally,

the Hydraulic

Laboratory of Lehigh University

and the

Cameron

Pump

Division of

the

Ingersoll-Rand

Company

cooper

a ted in

following

through

a comprehensive program of research

at Lehigh University

during

1944

and

1945. Over 200 per form

ance tests for viscosities up to 10,000 SSU were completed

on

four

variants of cen tr ifugal pumps , employing a special

t es t s tand

designed and bui lt under war conditions.

.

The

influence of viscosity-changes on

the

head, discharge

and

input

power characteri st ics of the pumps is demonst;ated

graphically and systematically for a wide range of viscosities

and

speeds.

The

usefulness of a special Reynolds number for pumps

is demonstrated

and test

results are correlated on that basis.

General conclusions are possible as to the influence of various

fea ture s of design, since several pumps of dif ferent specific

speeds were tested. A theoret ical analysis establishes def initely

th e variables to be considered

and

places

the

discussion for

the

problem on a sound scientific basis, especially with regard

to

disk

and

ring-losses.

INTRODUCTION

The

influence of viscosity

on the

performance of centrifugal

pumps has

no t

received

up to

the

present,

the systematic atten-

tion

of

the pump

engineers which

this

problem deserves

in

view

of

the

ever increasing use of the centrifugal

pump

for the trans-

port ofviscous liquids. A firs t pioneer effor t

and

so far the most

extensive one, was made by Professor

Daughertyl about twenty

years ago. Variouspapers

2

3 have appeared since then,

bu t they

were never based on more than a relatively small number of field

tests carried ou t

by

various investigators with greatly differing

pumps. These t es ts were used to derive correction curves for

efficiency by extensive extrapolation of

the

test information on

hand

and

must

clearly be recognized as a temporary means of con

sidering

the

viscosity influence.

In recognition of

the need

of a systematic approach

to

this

problem under controlled conditions,

the

Hydraulics Laboratory

of Lehigh

University and the

Ingersoll-Rand

Company

of Phil

l ipsburg, New Je sey, entered into a cooperative agreement

to

explore

the

behavior of centrifugal

pumps

when

pumping

oils. A

special

tes t s tand

was designed

and

constructed

under

war-time

.restrictions

and

over

two hundred

performance

runs

were

made

during 1944-1945

on

four

variants

of centrifugal

pumps

with

viscosities

ranging up

to

10,000 SSU.

The

influence of viscosity

1 Associate Professor

of

Hydraulics, Department

of

Civil Sani

tary Engineering, Massachusetts Institute

of

Technology. Assoc.

Member, A.S.C.E. .

Contributed by the Hydraulic Division for presentation at the

Annual Meeting

of

The American Society

of

Mechanical Engineers,

New York, N. Y., November

27

1945.

Discussion on this paper will

be

accepted until January 31, 1946.

3

changes

on

the head, capacity

and

input-power characteristics

was systematically explored for each

variant

of

pump and

for

several speeds.

The

usefulness of

the

Reynolds

number

for

the

presentation

ef

these characteristics was

demonstrated

convinc

ingly

and

all results are correlated on

that

basis.

The

data

pre

sented in

the

diagrams are those obtained

in

the present

investiga

tion only. No

attempt

was

m ade to

include information from

other

sources, since

it

was almost always found lacking

in

com

pleteness. This situation is analogous to

that

existing for

many

years

in

the field of pipe frict ion, until finally the pertinent

variables were definitely established

and

a final solution found.

is

to

be hoped

that the

experiments reported here will bring

forth

many

detailed experimental results

in

complete form from

the

files of various investigators, so that

the

conclusions drawn

from this work

may

be either confirmed or be modified

to

fit

into

a more general picture. However , it is felt

that the paper

repre

sents

the

first

attempt to

analyze

the test

results

on the

basis of

all

pertinent

var iables so

that

ind iv idual influences can be

isolated more easily.

Thus their

share in

the

overall effect of

the

viscosity

on the

performance of centrifugal

pumps

is

to

be recog

nized in relatively

true proportions.

may

be mentioned

that the paper had to

be severely limited

in presenting

the

information

on

hand.

The

descriptive

part

on

the

experiments

had to

be

cut

more

than may

be desirable from

the viewpoint of clarity

in

order to include the fundamental re

sults of greatest value to the engineer concerned

with

this pro

blem. The same limitations had to be imposed on the theo

retical

part

which covers only those aspects which are most useful

and

characteristic in explaining

the

results of

the

experiments as

stated

in graphical form.

EXPERIMENTAL PART

GENERAL OUTL INE OF TEST PROGRAM

The

stated purpose of

the

experimental study was to find, if

possible, a specific relationship between viscosity on one

hand

and efficiency, head, capacity, and power input on the

other

hand.

The

results for four different variants of pumps were to be corre

lated. The choice of pumps was naturally restricted by

the

power supply of

the

hydraulic laboratory,

by the

r ange of

the

torsion-dynamometer,

and

by

the

available storage space for

the

oils used

in

the tests.

a Specifications

o

umps Tested

Under

the

circumstances

thlf

four

pumps

listed in Table I were chosen as most suitable

and

as of widest possible use

in the pumping

of highly viscous oils.

All

vital

dimensions

and

normal performance

data

for

water are

listed

in

Tables I

and II . Pump No.1

was a single stage, single

suction

pump

which could be fitted with hydraulically identical

impellers of the closed

and

open

type. This pump

was relatively

small

and

of low specific speed.

The pump No.2

was a single

stage, double suction

pump

which could be

run with

impellers of

different diameter.

Thus

the specific speed was varied to a

value of almost 3000 with a small impeller, however not without

some sacrifice

in

efficiency. The experiments covered conse

quently a range of specific speeds between 1000 and 3000 which

is

the

range of most efficient operation for radial flow pumps.

L HIGH

UNiV RS TY

BETHLEHEM PENNSYLV NI


2/16

b Properties Liquids mployed The

program of testing

contemplated originally the use of water and of two oils as

experimental liquids. Fo r

the lat te r the

range of viscosities was

defined by the range of tcmperatures seasonably obtainable, i.e.,

from almost normal outside temperature upward

to

a maximum

produced by dissipating the power input. No special heating or

cooling

equipment

was provided. The

variation

in th e volume

circulated was sufficient to obtain the desired range of tempera

tures

and

viscosities. Two Black Oils, giving a medium slope in

the viscosity-temperature chart, were selected

and

are referred to

subsequently as Heavy Oil (HO) and Light Oil (La). A third

oil was added however, after the La-runs were taken, since it

appeared desirable

to

bridge

the gap

between

the

viscosity of

water and

that

of the Light Oil.

This so-called Thin Oil (TO) was produced by mixing the Light

Oil with a .

certain amount

of fuel oil.

The

Light Oil proved to

be extremely

stable

and it s viscosity remained

constant

for al -

most half a year of intermittent testing. The

Thin

Oil and th e

Heavy Oil were affected somewhat by the

pump

testing and fre

quent analyses were made t o adju st results for changes in vis

cosity whenever they exceeded the limits of permissible errors.

Fig. 1 presents

the

essential data on viscosities

and

specific

gravities for these oils.

c Schedule

Testing Program The experiments s tarted by

determining the performance of eaeh impeller of the two pumps

for water. All impel lers were

tested

as summarized

in Table

II

for three different speeds with the exception of the largest im

peller, which was run only for

the

two lower speeds, since

the

torque for the high speed exceeded the dynamometer capacity.

The

water tests checked very closely

the

performance of the

pumps reported from the Ingersoll-Rand Laboratory and served

in addition

the

purpose of establishing working procedures and of

finding o ut ab ou t

the

general performance of

the

test stand.

After someminor alterations were completed,

the

four sets of

runs

TABLE roo

I

PUMP

DIMENSIONS

Pump

Imp.

Fi le Type D

s

D

d

No.of

D

1

D

2

D

sh

B

2

~

2 S

R

No. No. No. Vanes

Av

-

1 1 IL 11 Single 4

2 7

2-1/2 7-1/4

15/16 1/2

3

7/8 .014

Closed Suction

1 2

IL

12

4

2

7

2-1/2

7-1/4

15/16

1/2

3

7/ 8

.014

Open

2 1 IL 21 .Double

8 6

7

5-3/16

8-5/8 2-1/2 1-15/16 7 1-7/16 .016

Suction

IB

IL

21 Double 8

6 7

5-3/16 8-5/8 2-1/2 1-15/16 7

1-7/16

.032

Clearance

10

IL 2

Half

8

6 7

5-3/:: 6

8-5/8.

2-1/2

1-15/16

7

23 .014

Width

32

Rings

2 2

IL

22 Double 8

6 7

5-3/16

11-5/8

2-1/2

1-7/16 7

1-7/16

.018

Suction

A

AilLE NO.

PU J Il

No.

Specific

Maximum

Liquids

Tested

ApprOXimate

Speed

for

Efficienoy

Speeds in

Water

for Water

RPM

L

1163

75.0

W

O

TO

2330

2875

3460

I

12

1163

74.0

W

O

T

2350 2875 3460

IL 21

2622

80.0

W

O

T

1230

1890

2320

i

I

IL

22

1991

85.5

W O

TO

1240

1880

Graphio

o e ~

otations

0

0

i

-

4


3/16

70 80 90

100

110

120 130 140 150

TEMPeRATURE DEGREES FAHRENHEIT

-

-

f - -

-

I---

TO

-

I - - -


4/16

C LIBR TION

OF METERS

FLOW DIAGRAMS

OF

CIRCUL TING SYSTEM

4

6

r

8

B

PERFORMANCE

O PU PS

FIG. 2. LINE-DIAGRAM

OF CIRCULATING SYSTEM

8

affect

the

readings.

The

speed was measured

with

a Chrono

Tachometer and with

a revolution counter electrically operated

Jo r

lito-minute intervals.

was

read

for every

test point and

comparisons were made

with

a Hassler

Tachometer

with results

well within

the

precision of

either

instrument.

The

tachometer

and

revolution counter

are

connected to a synchronous

motor

powered from a generator

mounted

on

the pump

shaft.

EXPERIMENTAL

PROCEDURES

a alibrations Measuring Equipment

All the

instruments

used were subjected

to

several calibrations during

the

experi

mental

period.

The

mercury gauges of course do not need

special calibration, except that

the

Bourdon gauge

tester

was

checked against

the

mercu ry columns. All connecting lines

were freed of

air and

particularly

the

discharge gauge was

. checked before

and after

each

run

for air,

by

the use of a special

water-air manometer,

the

latter

indicating easily.

any

errors of

0.01 ft. o f

water

column. Balance was a lways requi red within

this

limit and

was always obtained.

The

dynamometer bars were calibrated

by

careful

static

load

tests

and

showed a

constant torque pe r unit

deflection over

the

entire range.

Naturally the

percentage of error of the reading

for

any

run

was dependent

upon the torque

range covered for

any particular pump and

speed.

The

speed measurements were checked against simultaneous

measurements taken

with

other tachometers

and

were found to be

6

consistent within one-fifth of one per cent, which was

within the

required limits.

The

temperatures indicated

by the

dial thermometers were

verified by

the

use of a mercury

thermometer

certified

by the

Bureau of Standards.

The

readings were also checked during

runs by mercury thermometers

in adjacent

thermowells, which

had

been compared

to the s tandard

thermometers. Sticking of

the

dial

hands

was prevented

by

lightly

tapping

the glass cover.

The

main

problem

in

ascertaining

the

fundamental

quantities

for

the

performance of the

pumps

was encountered

with

the

l iquid meters. While the circula ting system for calibra ting

pur

poses is indicated

in

Fig. some of

the

phases of the calibrating

runs

merit

mentioning. Noticeable

temperature

differences in

the oil circulated had to be avoided . There fore the oil was

normally discharged from

the

line into

the upper

storage

tank

and

f rom here

by gravi ty back into the

lower

tank. The flow

was

then

gradually switched over

t o t he

swing spout,

thus

taking

the

oil

in the upper

storage

tank ou t

of circulation. As soon as

steady

flow

conditions were established again

through the

meter,

the flow

was deflected into th e measuring

tank

while a constant

level was

maintained i n t he

lower storage

tank by admitting

oil

from

the upper

storage

tank. This

procedure

naturally

required

considerable practice

bu t

worked

ou t

very satisfactorily.

The

results of

the meter

calibrations

are

given

in

Fig. 7

and

are

stated in

a form which deviates somewhat from

the

orthodox way.

The

discharge coefficient

for

the

4 by 2-in. Venturi

meter

is


5/16

1 10. 3.

G E N E R A L V I E W

OF T E S T

S E T U P

plotted against

th e

Reynolds

number

for

th e

4 in. diameter pipe

divided by th e discharge coefficient

CD

Similarly th e results for

th e 8 by 4.92 in. orifice meter are given in terms of a correction

coefficient for viscosity { plotted against

th e

pipe Reynolds

number for th e 8 in. pipe divided by the discharge coefficient.

T hi s l a tt e r q u an t it y can be calculatcd directly from

th e

differ-

ential manometer readings

an d

th c correction coefficient

J{m

is

thus obtained without trial an d error. Th e discharge coefficient

for water CD as introduccd is cqual

to

0.6185. Th e Venturi

FIG . 5 .

VENTURI AND ORIFICE MlilTERS

7

1 1 0 4 P U L L E Y SYST EM A ND D Y N AM O M ET E R

meter

curve is compared

to

th e

curve given in

th e F lu i d M e te r

Report of

th e

A.S.M.E.

an d

is een to lie considerably below

this

curvc. I t is shown extending down to an experimental value of

CD = 0.50. Th e orifice curve i compared to Johanson s curve

for a imilar orifice in a one in. diameter glass t ub e a nd shows

rcmarkable agreement. I t ma y even be argued

that

th e dis-

crepancy is due to

th e

slight difference in th e orifice pipe diameter

ratio which was 0.615 here while

Johanson s

curve holds for

0.595.

FIG .

MA IN

INSTRUMENT

PANEL


6/16

10 '0'

0

0

REM

F IG 7. FLUID METER

CALIBRATIONS

0

0

--

METER

COEFFICIENTS

1....

V SUS

1/

I --JOHAN

~ N S

culM

REYNOLDS NUMBER R

/

k..

EM

1/

~

/

M

1 -,;

/

/> SM i -(

0

0

0

o I

V

V

~

V

8X4 92 ORIFICE METER

K ~ ~ I l R o ;

1/

4 X2 VENTURI-METER .

tr YfJ

.

1/

1III

I I I

10

10 '

,

4

.5 0

60

12

70

,80

u

Cl

b ypic l

Performance Tests

Most

operational difficulties

were discovered

an d

ironed

ou t

by running

fi rst a full series of

experiments

with

water. Only minor improvements however

were necessary, and definite procedures for

the

oil runs

w e r ~

formulated.

Th e

dimensionless performance curves shown

in

Fig. 8 are those obtained for the medium specific speed impeller II ,

22 and t hey are

typical

of

the

range covered by individual tests.

The almost complete absence of viscous effects for th e normal

range of operation

with water

was indicated by all

water

runs

proving if

it

be necessary indeed, that the Reynolds

n u m b e r ~

normally encountered with

water

are

in

th e so-called rough

flow zone.

All performance

runs

with oil were carried

ou t by

recording all

pertinent instrument

readings

photographically by

mcans of a 35

mm.

Kodak

camera.

Thus

readings for individual points were

taken speedily and simultaneously for

permanent

reference.

Runs

of

ten

to

fifteen

test

points

were

normally

completed within

ten minutes, wl1ich fact considerably diminished the'increase in

temperature during any run.

Upon

tennination

of one

run

the

pump

was operated near normal discharge and was

run

continu

ously,

unti l the

dissipation of power

had

resulted in bringing the

temperature to

th e desired higher value. The experimental pro

cedure was

then

repeated.

The temperature

of

the

oil could be

varied for any of the higher speeds from

about

70

0

to 125

0

by

continuous running within one day.

Thus

five

to

eight perform

ance

runs

were obtained within a four- to eight-hour period.

Shut-off conditions were usually

tested

by closing down fairly fast

from a large discharge, thus retaining a fairly normal temperature

of

the

oil churned

in the

pump. At best, however,

the

shut-off

points

are only approximate, especially for

the

higher viscosities.

The temperature

changes in th e oil could

be

controlled

by ad

justing

the

volume of oil

in

circulation. I f

the

temperature

rise

was too fast

the entire

volume of oil could be utilized by circulat

ing

through

upper

and

lower storage

tanks.

A differential

tem

perature

between upper an d lower

tank

also

permitted

to arrest

the rise

in

temperature during any run

after

some practice.

Small increases

in temperature

were permitted,

say

2

0

to

3

of,

an d

their

influence was determined by tak ing additional points

immediately

after

each

run

over

the

same range. Corrections

to

a

constant temperature

and hence to a constant viscosity could

be made, therefore ,

in

th e analysi s. However, usual ly such

corrections were too small

to

influence the maximum efficiency in

location an d magnitude.

ANALYSIS

OF TEST

IlESULTS

a

omputation Procedures

As pointed

ou t

before, all

data

necessary for

the

performance calculations are contained on a 35

mm. film

strip

which could be projected,

point

by point, on a

screen made of

tracing

cloth.

The

readings were

taken

down by

an

observer

behind the

screen.

The

gauges showed

throughout

the

experiments

very steady

readings, fluctuations were largely

eliminated by

the

viscous action of th e oil

in the

connecting lines

and by

symmetrical constrictions

in the

mercury manometer

blocks. Two exposures were taken for every point in order

to

discover changes

in

readings an d to insure

that

al l the informa

tion

could be read clearly.

I f

poor

test

points were discovered

la te r i n p lo tt in g

th e results,

they

were invariably

traced

to

faulty readings from the film and were easily corrected by check

ing.

The

value of preserving

thus

on film

th e

original measure

ments cannot be overemphasized, especially when insufficiently

trained personnel

must

be

entrusted

with a

major portion

of

th e

computing work.

The

entire calculation procedure involving a great

many

steps

was careful ly worked out , so that a standard routine could be

followed

throughout and the evaluation

of

data

from

th e

film

to

the

final performance graph cou ld largely be left to personnel

without

technical training. Viscosit ies

and

specific gravities

vclocity-head corrections, and discharge coefficients were read

from graphs

plotted to

simple linear scales.

Since th e discharge manometer in many runs indicated

very

small differences due to the exclusive use of th e 8

by

4.92-in.

orifice meter, these readings were also

read

directly which

upon

comparison with the readings obtained from th e filIp-showed very

good agreement.

I t

was found practical t hen to take down also

the

speed

and

th e torque readings during th e r un so that some

calculations could

be s ta rt ed

immediately

after

th e

runs

were

taken, before

the

films were developed.

Pressure readings were no t eorrected for pipe-friction losses

between

pump

flange

and

piezometer connections.

This

refine

ment

would

have

introduced additional computing work to

an

extent unjustified by it s merits, the pressure connections in each

case being located only one pipe diameter from flanges.

I t

is

natural

also that

the

percentage error in the overall efficiency be

comes greater for higher viscosities, so that

the

above correction

stays usually within the permissible margin.

I t

may be stated

that

the

accuracy of

the

results is referred to the efficiency loss

8


7/16

_I

1

I

1

I I I

1

I I I

I

1

I

1

-

-

r r

ou

U )

?

-

-

.J

-

v

:

-

/

-

/

tr -

/

-

-

-

.. r

-

V

-

9

LEGEND: 0 1240 RPM

1880 RPM

-

1

I I I I I

I I I I I I I

1-

ISO

40

:0-

U

Z

w

40

U

-

-

w

30

20

0

200 400 600 80 0

1000 1200 14 J0 If

00

1800

FIG.

13.

E X A MP L E S OF H E A D - C A P A C I ~ Y

AN D

E F F I C IE N C y - C A P A C IT Y C U R V E S F O It V A R IO U S P U M P S AN D V IS C O S ITIES

th e

efficiency on

th e

basis

of

constant

o

departs markedly from

th e

maximum or peak efficiency for higher viscosities,

th e

method

proved very soonimpractical as can be seen from Fig. 13 an d was

therefore given up.

2 I t

ha s been suggested sometimes, to assume a constant

head, regardless of viscosity.

This

gives normal capacities which

are very soon much lower t h an t h eone corresponding to maximum

efficiency as ca n easily be verified on Fig. 13. However, for

viscosities up to 100 SSU either

method

ma y

be

used without

introducing large errors.

3.

I t

was found that if

th e

ratio Qo/VH

o

was assumed

to remain

constant

as calculated from th e water performance,

a

parabola

cou ld be p lo tted , which would inter sect

th e

head

discharge curves a t th e

point

of maximum efficiency, as

shown

in

Fig. 13. This proved

to

be especially helpful in view

of th e fact that for lower viscosities it is impossible to decide by

inspection

only on

a point of

maximum

efficiency

an d

correspond

ing normal discharge, head

an d

power. T hi s m e th od was used,

therefore, in plotting th e information presented in th e following

graphs, which represent

th e

essential results of

t h e s tu d y

for easy

reference.

These

graphs give

th e

complete Reynolds number

characteristics for th e practical range of operation for a ll four

pump

variants.

b

Influence of Viscosity on Performance

1.

Normal

Head

an d Capacity. A few genera l r emarks are in

order to explain Fig. 14

an d

15. All heads

an d

horsepower inputs

are given

in

terms of their corresponding water equivalents an d

are

plotted

against

RD.

Th e efficiency loss (100 - /100 was

plotted in preference to

th e

efficiency, since logarithmic scales are

used, so that accidental errors ar e shown in correct proportion.

All curves show th e water performance points on th e right-hand

side in black with special marks to identify the different speeds.

Th e

points for

Thin

Oil arc white, followed

by

Light Oil

points

in

black, an d finally th e points for Heavy Oil are white again. This

identification by liquids proved desirable, since differences in per

formance were discovered as

th e

liquid

was changed, even

though th e Reynolds numbers remained of th e same order of

magnitude.

Th e

explanation for these differences is given

later

with

th e

reasoning relative to th e shift of curves with speed. Th e

general trend of

th e

curves is self-explanatory

with

decreasing

heads

an d

with increasing power

i n pu t a n d

efficiency losses as th e

lleynolds number decreases. Th e striking influence of th e disk

friction losses becomes apparent by comparing Fig. 11 with Figs.

14

an d

15.

Th e

steep rise in efficiency losses is directly

traceable

to th e change from

turbulent

t o l a mi na r disk-friction losses. A

specia l curve for reduction in capacity was found unnecessary

14


13/16

since th e discharge Qo can

be

obtained from H

o

on the basis of

Qo = K

V i

wherein K = Qw/VH Th e latter relationship

was found

to

hold true for Reynolds numbers R

D

as low as 3000

4000.

Below these values of

R th e

capacity

decreased more

rapidly, th e head-capacity curves became very

steep

an d th e

point s of

maximum

efficiency ha d to

be

picked by inspection.

This, of course, results

in greater scattering

of th e points, since at

th e same

time

th e

general

accuracy

becomes lower. A

slight

s hi ft i n

th e

capacity

ma y

produce

an

extremely large change in

th e head. However , it is felt that

th e

lower limit of R

D

4000

will terminate almost an y range, that might come up ill actual

practice.

2. Input Power.

Th e

input power curves

represent

a correc

tion

factor by which th e water horsepower

input

BHP

w

as

cor

rected for specific gravity, is to

be

multiplied to get th e horsepower

input for oil

BHP

o

.

These curves indicate a rather large in

crease

in th e

latter for Reynolds numbers, for which th e head

an d

capacity

corrections

are almost

insignificant.

This

would

be

fur

ther

proof of th e

contention

that th e decrease

in

efficiency an d

th e

increase in power input for Reynolds

numbers

near th e

100,000

mark is

mainly

due to th e disk an d ring friction. I t should also be

noted

that pumps IL 11 an d

IL

22 show

th e

largest increase here,

while

IL

21

with

a small impeller

bu t

relatively large rings, an d

IL 12

with

an

open

impeller show smaller values.

Th e

rise

in

input horsepower for lower values of

is

m o re o r

less linear

in

a

log-log plot, corresponding t o t he change with

Reynolds

number

of al l

hydraulic

losses including

disk an d

ring friction. Fo r al l

pumps, differences in power input ar e found for constant Rey

nolds numbers as a funct ion of speed. T h is m a y be explained in

part

by

th e effect of

ring an d

stuffing-box friction

as

outlined

in

a previous section. Th e

running

temperature for

higher

speeds

increases,

this

i n t ur n causes a

reduction

of th e shearing stresscs

a n d t h er eb y a reduction also in th e

percentage

of

ring

losses in

terms

of total power

input.

Th e

larger

th e power output for a

given pump

th e

smaller will be

th e

ring-losses i n p er c en t o f in

pu t

power for th e

same Reynolds number.

t is obvious

then

that

these

differences

s ho ul d b e most

con

spicuous for

pump

IL 21 where

th e

output was very

small an d

that th e difference between low

an d medium

speed for pump IL 22

is

much

less, since

th e

impeller of

th e

latter

was 35

pe r

cent

larger in

diameter

for th e same inlet an d

ring

dimensions. t is

clear, furthermore, that th e results for pump IL 11 will show th e

influence

under

discussion even less, since th e rings h er e a re v er y

small

in proportion

to t he impeller

an d

therefore have but

little

effeet on th e

total

losses.

Th e

stuffing-box losses will

accentuate

th e phenomenon, since

IL

21

an d

IL

22

ar e

doublc-suction

pumps

while

IL

11

is of th e single-suction type. Th e disk-friction

losses i n c on tr as t

to

the ring-friction losses will in general be

larger for low specif ic speed pumps. They will therefore affect

th e

performance of IL

11

most o f al l

an d

will have a

minimum

effect on th e

behaviour

of IL 21.

This

of course, was one of th e

reasons for selecting

these pumps.

t should also be

mentioned

t h at t he curves for power

input

correction an d efficiency loss for

th e low speed a re p r ob a bl y no t of

practical

value, s ince such

speeds

are

seldom employed.

They

were included

in

these tests

mainly

to

bring ou t th e effect of speed an d i n o r de r

to

increase

th e

range of

Reynolds number

for

each

oil.

3 Maximum Overall Efficiency.

Th e

tendencies of th e head

an d input-power correction curves

ar e

essentially reflected in th e

plot of th e

ratios 100 -

e

/100

against

R ey no l ds n um b er .

These curves wil l show all th e discontinuities encountered with

each of

th e

o t he r t y pe s t o a somewhat larger scale. Th e question

ma y

be

asked,

wh y

th e efficieney loss was no t given as a fraet ion

of th e effieiency loss for water.

Th e

argument against

t h e l a tt e r

method would

be

that th e efficiency loss itself represents already,

15

to a

eertain extent

a dimensionless

factor

of res is tance for a given

type of pump similar

to the

pipe-friction factor j for a given

pipe with relative roughness e/d .5 Just as it would be unwise

to

di vide all pipe-friction factors b y t he special values of

f

obtained

for var ious values of e /d , so nothing could

be

gained by dividing

all efficiency losses

by

th e water efficiency. Since all efficiency

losses

remain

constant with

respect

to

Reynolds number

in th e

range of

water or

ai r performance, while

Reynolds number deter

mines

th e

behavior for

more

viscous liquids,

th e analogy

to

pipe

friction

phenomena

is close. Th e

relative

roughness

of

th e pipe

would find its counterpart

in

a specific

speed

modified to include a

relative-roughness

parameter.

However,

such

an undertaking

must

be postponed until such time w he n m or e e xp er im en ta l

material systematically sifted an d coordinated, is available .

Th e

efficiency-loss curves for low

Reynolds

numbers show a re

versal of th e

trend

at which th e losses increase. t was pointed

ou t

before that below

Reynolds numbers

of R

D

3000-4000

th e

capacity decreases at a much faster

rate

than above those values.

A spec ia l

capacity

correction .eurve should have been added,

which, however, was

no t

believed necessary

in vieWl:of

it s negligi

ble

practical

significance an d of th e

somewhat

uncertain evidence

which

permits

a wide interpretation

a s c an

eas ily be seen

in

Fig.

13.

wanted

it of course be

calculated

from th e

other

curves.

There

is, however,

an

interesting explanation

for

th e

lower

rate

of increase of

th e

efficiency losses, which t a ke s i n to account th e

heat exchange

in

th e

pumps.

no heat were developed

by

ring,

stuffing-box,

an d disk

friction, th e efficieney would

reaeh

in

significant values very fast. As it is, t he h ea t developed

when

heavy oils are

pumped

wi ll slow down this development con

sider ab ly . As t he capacity decreases

th e

cooling of

th e

pump

body

decreases, so

that

th e efficiency-loss

curve

will

approach

a

value

of

unity

only for

very

small

Reynolds numbers.

Th e

ultimate performance will be

such

that oil

ca n

be pumped only

as

it is warmed

up

by th e dissipation of

almost

th e

entire

power

in

pu t

i n to h ea t. Th e

fact

that this tendency became apparent

in

th e

curves also points

to the

limits of the

practical

use

to

which

centrifugal pumps may be gainfully employed in pumping viscous

liquids.

Because

it

was desired

to

show

th e

application

of a d if fe rent

Reynolds number

th e results for pumps IL

11 an d

IL 22

have

been plotted

against

R

s

283.7Qo/d

i

.p10 . This

as

is

shown in Fig. 16, results in changing th e

order

of magnitude of th e

Reynolds number by a factor of

te n

an d since

will decrease

somewhat with Reynolds number

th e curves wil l be stretched

over

a wider range. Bu t essentially

nothing

new is gained f rom

t hi s p lo t a nd th e difficulty

o f h a vi n g

an

unknown

quantity Qo

in

th e expression for R

s

makes

it s usefulness quite doubtful.

GENERAL CONCLUSIONS

GleNICHAL RgYNOLDS N U ~ g R CHARACTgRIS l ICS

Th e results of th e

experimental

an d analytical work described

so fa r show clearly that th e resistance of a centrifugal pump

ca n

be

represented

by

distinct

curves plotted against

Reynolds

num

be r R

D

2620 Nd

2

2

/p.1O. . Three curves ar e normally re

quired to

describe

th e behavior as

a

function

of

Reynolds number

RD:

1. Ratio

of n or ma l h ea d for oil H

o

to th e n or ma l h e ad for

water: Ho/H

w

2. Ratio of

normal

power input for oil

BHP

o

to th e

normal

power input for water corrected for specific gravity:

BHPo/ soBHP

w

.

3.

Th e efficiency

e

01

better

th e efficiency loss

100 - e

Th e

capacity

correct ion for th e practical

range

of

operation


14/16

REYNOLDS NUMBER RD 2620 ~ ~ O

RESULTS PLOTTED

AGAINST

R FOR PUMP IL

1

IG. 1 4.

LEGENDS

LIQUI S

SPEEDS

W

LOW SP

0

0

Q I.lIEDlUl.II SPEED

LO

HIGH

SPEE

0

HO

r-..

IQQ. e

12

~

. HPw

~

.

100

ILi I

0

:S:-BHPw

.

o

010

1.00

10

2

010 lOO

lOO

O O

. . . 0.60

11

04 0 fIef

I I

_______

0 l 0 l

rio iJ

2 2

100

08 0

--:--1 0.60

:n :

o

0.40 flo

u

iJ

TIO

2 2

1.00

0.80

: IJ 0.60

I I

o

Z I I

.. 0 l 0 l

= ~ \ b VI

P 02 0 20 0

Hw

lOO-e

IL 21

-J,v

f 100

f or S SU

-

I

r-o--,.

BHPo

I

.BRJlW

010 ICO

10

...... 0.60

i:II

- .. .. .. .- . 0

o

z

~ I

0

O.

i

63

.

7

-Q

Q

,

REYNOLDS NUMBER R

s

=

10 V 0

FIG. 16.

RESULTS PLOTTED AGAINST

IL

II r IL

22

j ~

:1_1 I ~ i::t ' - I , ~ H o i i dj-H

I

I

-= I

i

.

f i ~

I

i=t HJ ~ H w

u..uJJ

I

_L=I. .I__ i i .

J_j j H

I

- -

1

- - ~ - + -

I I

I I I I . I I I

I

-----

r--

I

-+-, r ~

I

100

III I

0

-

---r-

I IL22

BHPo

II

I

SoBHPw

1

080

06 0

04 0 ,fl -

__

o ( I J ~

z

1

2

2

g

= -

0 10

10

R

s

GENERAL R EY NO LD S N UM BE R CHARACTERISTICS

, , ,

, ,

=l

, , ,

I

-----.::::t

Hw

,

,

,

::s:::::l

1

, ,

I

i ::::

___

I

I

~

-----

IL II

t

II I I

IL 22

Jll:l

r SoBHPw

I

L II

IL 2

,

4

010

1.00

10

1.00

08 0

: ....o 60

~ 0 4 cPL

z

J : I ~

2

--=---

FIG. 17. GENERAL

REYNOLDS

NUMBER

RELATIONS FOR SPECIFIC SPEEDS

OF 1000

TO

2000


16/16

may increase

at

first above that produced for water .

This

is

believed mainly due to the large effect of disk act ion for such

pumps, since the pumping by disk

action

here becomes a con

siderable percentage of

the

total capacity. Further proof for

this contention is supplied

by the

curves for

the

open impeller

of the same pump, which show a lower head and lower power in

the same range of Reynolds number.

ring losses are to form a small part of t he total losses, ab

normally low speeds should be avoided.

The

useful

output

of a

given

pump

is increased considerably for viscous conditions of

flow by increasing the speed, since the influence of ring losses is

diminished. This tendency is exemplified by the curves for the

higher specific speed pump IL 21, which shows

the

greatest im

provement with higher speeds, since ring and stuffing-box losses

form a large percentage of the total losses for

the

ower speeds.

Individual losses, which

are

due

to hydraul ic

friction, to

disk action,

to

ring and stuffing-box friction, are

no t

easily

separated at present, bu t will vary apparently within reason in

a proportional

manner

for

pumps

of conventional design.

t

may be assumed that a decrease in disk friction for a higher

specific speed is compensated for in part

by the

relative increase

in ring f rict ion for this type of pump. However,

there

is sti ll a

net

gain

in

efficiency, when compared on

the

basis of

constant

lleynolds number, and considering only the curves of Fig.

7

Compared on a percentage basis

the

efficiency losses differ less

for low Reynolds numbers

than

for the water performance runs,

since the curves converge rather rapidly for a cer tain r ange of

R

D

values.

A considerable increase in the running temperature for a given

pump

is caused

by

ring and stuffing:box friction. While the

evidence on hand does not permit more than a qualitative esti

mate

of the influence of the pump

temperature

on the per

formance, it may be s tated that for a given, Reynolds number

and a given speed the efficiency is , somewhat higher

and

the

head produced is increased, if the running temperature is high.

In

therangeof R

D

where Heavy Qi(HO) and

Light

Oil (LO)

points overlap,

the

curves show a distinctly higher head for

the

ormer, since it s

temperature

is high here while t he Light Oil

is cold.

These

remarks are made at this t ime with the intent

of calling

a tt en tion to the

influence, of

the

running

temperature

rather than with the idea of suggesting definite answers.

Attempts

have been made during

the

evaluation of the data

to analyze the individual losses separately

and

t o deduct from

the power input the influence of disk and r ing friction. While

expressions for disk friction give probably the order of

magnitude

of these losses correctly, the ring friction is not as easily calcu

lated, since

it

is influenced to a decisive

d e g r e ~

by large tem

perature changes in the ring space. For

the

t ime being, there

fore, the attempt at isolation of the hydraulic losses from the

input power was discontinued. Bu t it is clear

that

eventually

this phase of the problem will have to be solved, if further insight

into the relat ive dis tr ibut ion of these losses is

to

be gltined.

This will be a very fruitful subject of further experimental studies.

R g c o M ~ m N T I O N s

On

the

basis of

the

information

made

available

in

this

paper

it is possible to propose some very definite

future

steps towards

it final solution of the problem under disyussion.

A considerable amount of detailed information, which up

to

date seemed irrelevant, should be disclosed

by the

profession,

as a result of the material published here.

This

material should

be critically sifted

and

coordinated

by

a

central

agency on

the

basis of all

the

pertinent variables involved.

2 The information embodied i n th e performance curves and

data, of which the results presented form

an

important bu t rela

tively small part, deserves

further

analysis.

The

influence of

viscosity on part load

and

overload operation should be cleared

to

a ce rt ain ex tent . Since disk-

and

ring-friction losses remain

more or less fixed regardless of discharge, the detailed analysis

of

input and output

curves should yield valuable information

towards separation of these losses from the hydraulic losses.

3. The problem of ring frict ion shou ld be

studied further

analytically and experimentally. Applications towards -

proved stuffing-box design

and

use of mechanical seals may be

included here.

4.

Analytical studies should be undertaken to determine the

influence of relative roughness

and

of specific speed as variables

influencing the efficiency, so

that

eventually a set of universal

performance curves

may

be

plotted

including these quantities

in addition to Reynolds number.

5

Additional performance tests should be

made

with

pumps

of different design, especially with pumps of higher specific

speeds. Such tes ts could be

made

on a

much

more economical

basis now, since the pertinent variables and t rends are estab

lished Effects of changes in design should be systematically

studied.

ACKNOWLEDGMENT

The

study reported here represents an example of academic

industrial cooperation, which was sustained by the active interest

of

the

administrative officers of Lehigh University

and

of

the

Ingersoll-Rand Company in this work.

In

this connection the

author is especially indebted to Mr. W. M.

Stanton

of Ingersoll

Rand

and to Prof. H. Sutherland of Lehigh Unwersity.

Inti

mately

connected

with

the technical phases of the work were

Messrs. A. P. Brocklebank

and

H. Hornschuch, to whom the

author wishes to express his particular appreciation. Mr. Ming

Lung

Pei was responsible for a great deal of analytical work.

Thanks are due to a considerable nUmber of other associates

connected at various periods with thi s work and to the loyal

cooperation of

the

staff members of

the

Hydraulic

Laboratory

of Lehigh University.

BIBLIOGRAPHY

1. R. L. Daugherty,

A

Further Investigation of th e Per

formance of Centrifugal Pumps

when

Pumping Oils. Bull. 130,

Goulds Pumps,

Inc.,

1926.

2. A.

J.

Stepanoff,

Pumping

Viscous Oils with

Centrifugal

Pumps.

Oil and Gas Journal May,

1940.

3.

N.

Tetlow,

A Survey

of Modern Centrifugal

Pump

Prac

tice

for Oilfield

an d

Oil

Refinery Services.

Inst. of Mechanical

Engineers, 1942, pp . 121-134.

4. A. H. Church, Cent ri fugal Pumps a nd

Blowers. John

Wiley

Sons, New

York,

1944.

5.

L.

F.

Moody,

Fr ict ion Factors for Pipe Flow. Trans.

A S M E

Vol. 8 , 1944,

p.

671.

6.

Th .

von Karman, Laminar and

Turbulent

Friction. Z

f. Ano Math. und Mech.

Vol. 1, No.4, 1921,

pp . 244-249.

7. F. Schultz-Grunow, Frictional

Resistance

of Disks

Rotating

in

Housings.

Z f. Ano ilIath. und Mech. Vol. 15 , No.4, 1935,

pp . 191-204.

8. A.

H.

Gibson, Hydraul ics and It s Applications. 4t h edi

tion, 1930, Macmillan Co., pp . 186-191.

9.

E.

Schneckenberg, Durchf luss von Vasse r durch konzen

trische

und

excentrische Drosselspalte mi t und ohne Ringnuten.

f

Ano

lv/ath.

und

Mech.

Vol. 11,

No.1,

1931,

pp .

27-40.

10. S. Suzuki, On th e Leakage of

Water

Through

Clearance

Space. Journal

of

Fac. of Eno. Univ.

of

Tokyo Vol. 18, No.2 ,

1929, p. 71.

1

R. J.

Cornish, Flow of

Water

Through

Fine

Clearances with

Relative Motion of th e Boundaries. Proc. Royal Soc. 140 A, 1933,

pp .

27-240.; (Note : mispr in t

p.

237,

Fig.

7,

dp

should read

p. p dx

m 3 p ~ .

p.

dx

12. S. Goldstein, T he Stability of Viscous Fluid Flow Between

Rotating Cylinders.

Proc. Cambridoe Philosoph. Soc.

Vol. 23,

Jan., 1937.

18

Documents

The Influence of Viscosity on Centrifugal Pump Performance - 199_1