3ltl)aca,

Hem ^atk

SOUGHT WITH THE INCOME OF THE

SAGE ENDOWMENT FUND

THE GIFT OF

HENRY

W. SAGE

1891

'

The date shows wh^n

this

volume was taken.

No. and give to

To renew this book copy

th(* call

the hblarikn.

HOME USE RULES

:^:
.
;
*

All'fiooks subject to recall

All borrowers

must

regis-

"

ter in the library to borrow'

u
^_
_*yJ......
....
'"

books for home use.

]A11

^turned
.

year

bopks must be rea't end of college for and inspectiion

repairs.

'

Limited books must be returned within the four weeklimit and not renewed.
I

'

'.

,./,

y"

'

Students must return all books before leaving town. Officers should arrange for the return of books want^ed during thei&absence from
town.

Volumes of periodicals and of pamjihlets are held

.

'

'

in the library as

much

as

possible.
.

For fecial pur-

a'
;

poses they are given out for limited time.

Borrowers should! not use

their library privileges for
-

thebenefitof other#fcersons.

Books of special value and gift books, when the

,

giver wishes it, are allowed to circulate.

not

port

Readers are asked to recases of books all

or

marked

mutilated^^^^^

Do not deface

books by marks and writing.

Cornell University Library

arV19451
Water hammer
in

hydraulic pipe lines

3 1924 031 260 254

olln,anx

The
tine

original of

tliis

book

is in

Cornell University Library.

There are no known copyright

restrictions in
text.

the United States on the use of the

http://www.archive.org/details/cu31924031260254

WATER HAMMER IN HYDRAULIC PIPE LINES

BY TEE SAME AUTHOR

HYDRAULICS:
And
Its Applications.

Price, 18s.

ne,t.

This is the test book to date ou the subject ; it is, in fact, better than . all the others put together. We can thoroughly recommend the
'
.

book.'

Engineering Times.

consider the work one of the most satisfactory text-books on

'We

Hydraulics World.

extant.'

Mechanical
CO.

ARCHIBALD CONSTABLE &

LIMITED

WATER HAMMEE
of the Rise or Fall in Pressure in a

IN

HYDEAULIC PIPE LINES

Being a Theoretical and Experimental Investigation
Pipe Line, caused

by the gradual or sudden

closing or opening of a

Valve; with a Chapter on the Speed Regulation

of

Hydraulic Turbines

BY
A. H.

GIBSON,

M.Sc., Assoc.

Mem.

Inst. C.E.

SE^fIOB

DBMONSTKATOK AND ASSISTANT LBCTUEEE IN ENQINEBEINO IN THE MANOHESTEE UNIVBESITT

LONDON
ARCHIBALD CONSTABLE &
10

CO.,

LTD.

OEANGE STREET, LEICESTEE SQUARE

1908

D.

VAN NOSTRAND COMPANY

3WW YORK

-g-t-s^

1^

A.3TEo5'lo

PEEFACE
Although
the subject the study of Hydraulics has been very

much

to the fore during recent years, one particular branch of

viz.,

that dealing with the problems involved

during the acceleration or retardation of a water column,

would seem to have been almost entirely neglected.

In
fact,

except for

an experimental investigation by
translation,

Joukowsky, of which we have no English

which only deals with sudden stoppage
available literature dealing with the matter.

and

of flow,

we have no

In the majority of text-books on Hydraulics the treat-

ment which the question

but
is

receives

is

not only of the scantiest

is

as inaccurate as scanty,

and as the whole subject

not without

most interesting

and

is

certainly

practical

importance, perhaps

it is

unnecessary to apologise for the

following account of a theoretical and experimental investigation carried out

by the

author.

The experimental portion of the work was done in the

Engineering Laboratories
of

the

Manchester University,

and the thanks

of the work, for

of the author are

due to the University

authorities for their ready acquiescence in this publication

and

to Mr. J. Hall of the Engineering StafiF

much

help in the preparation of the apparatus and in

carrying out the experiments.

A. H.

GIBSON.

Manchester, September

1908.

CONTENTS
PART.
1.

Introduction,

...
.

CHAPTER

.1
. . . . . . . . . .

PAOB

2.

General theory of gradual closure,

3.
4. 5.

Description of experimental apparatus,

Experimental results,

Gradual opening

of a valve,

.18 .22 .24

CHAPTER
1.

II

Application to the theory of turbine regulation,

27
31

2.

Speed regulation assuming uniform pipe-line acceleration,

Effect of a stand pipe on speed regulation,
.

3.

.33

CHAPTER
1.

III
. .

Sudden stoppage

of

motion

theory,
.

.36
41

2. 3.
4.

Valve closed suddenly, but not instantaneously,

Experimental results,
.
.

.42
.

Sudden closure

in non-uniform pipe-line, of a valve,

.
.

45

5. 6.

Sudden opening

.48
.

Sudden opening neglecting

effect of elasticity,

50

CHAPTER IV
1.

Further phenomena connected with pipe flow,

Appendix,
Bibliography,

-.... ......
.

53
57

60

CHAPTER
Introduction

General Theory of Gradual Closure Description of Experimental Apparatus Experimental Results Gradual open-

ing of a Valve.

1.

Introduction.

It

is

a matter of general knowledge, that any retardation

or acceleration of a water column, such as

may

be caused by

the closing or opening of a valve at the outlet from a pipe

line, is

accompanied by a

rise or fall in the pressure

behind

the valve.

In certain cases this change in pressure

equally to every portion of the pipe
line.

is

transmitted

When
may be
and

such a change takes the form of a sudden rise in

pressure, this is

known

as

Water Hammer.

Its

magnitude

large with even comparatively low velocities of flow,

of importance, not only in so far as it affects the

it is

safety of a pipe line, but also in its effect on the ease of

regulation of any prime

mover supplied by the

latter.

In,

this latter respect the fall in pressure

caused by opening the

regulating sluice

is

of even greater importance than a corre-

sponding

rise in pressure, since the latter

may

be readily

guarded against by the provision of an adequate relief valve

or pressure regulator.

Although the
of a valve has

rise in pressure

caused by the sudden closing

been investigated by more than one experi-

menter, the pressure change due to a gradual closure, or to a gradual opening, has not, to the author's knowledge, been

WATER HAMMER IN HYDRAULIC PIPE LINES

Further, those formulae

the subject of previous investigation.

which have, in one or two instances, been educed as giving

such pressures,

may

be shown to be based on assumptions

which are fundamentally unsound, and when tested show

results

which are in general widely at variance with those

which led to the following theoretical
of the subject.

obtained experimentally.
It

was these

facts

and experimental investigation

2.

Gradual Closure of a Valve

a rigid body of weight

General Theory.
has
its

If

lbs.

velocity in
ft.

any
of

given direction changed by an amount 8v

per

sec^n^
gravita-

time

seconds,

its

acceleration

is

Sw/Si,

and the^change

momentum
tional units.

per second in this direction

is

-^

in

To produce
force
this

this acceleration or

change in momentum, a

F must be applied to the body in the direction in which

change takes place,
:

its

magnitude being given by the

formula

i^^^.^lbs.
g
ot

As an example

of this,

if

water be flowing along a rigid

is

pipe whose cross-sectional area at any particular point

sq. ft.,

and

if -j-

be

its acceleration,

the force necessary to

this point, of

accelerate a short element of the

column at
,,

length

Sx, will

equal

a.6x.-^
dt

<>

dv

lbs.,

all

g dimensions being in feet and

w being the weight (62-4: lbs.)

of 1 cubic foot of water.

This force can only be produced by a difference of pressure

GEADUAL CLOSUKE
at the

3
this pressure

two ends

of the element, and,

ft.,

if

Sy be

difference in lbs. per sq.

we have
bt g the difference of pressure at the

8p=-.Sx..
In a pipe of length
I ft.

two ends, due to acceleration

P'
If the

of the

column ==ESp'=j?', so that

(1)

= ~\

jf^^'^^^-'^^'^^l-^^
if

pipe be of uniform area, and

water be assumed to

be an incompressible fluid so that, no matter what changes

of pressure

may

occur throughout

its

mass, each particle has

the same velocity at any given instant,

we

get, at this instant

p
Where

=
g

dt

lbs.

per

sq. ft.

this measures the

change in pressure at the outlet

is

valve, it will be positive

when the motion

being retarded

and negative when being

accelerated, so that the equation

must be written
,

wl dv

" ~~g'~ii

1,

J.,

,,,,

^^"^

^^
is

^
super-

This pressure difference at the two ends of the pipe

posed on that due to steady flow with the velocity obtaining

at the given
instant.

Thus

if

the loss of pressure from

ft.

entrance to exit, due to steady flow with velocity v

sec,
is

per
re-

-( l+=i-|
^g\

m/

lbs.

per

sq. ft.,*

and

if

and -^ are
dt
feet,

Here

_ measures the loss of pressure head in

due to change

-^ measures the loss of head Igm. In this expression in is the hydraulic mean due depth of the pipe and in a circular pipe has the value (diam. -r4), while /is a numerical co-eflSoient whose value depends on the physical conditions of the pipe walls and to a less extent on the diameter of the pipe and on the velocity of flow. See. Jlydraidics, Gibson,

from pressure to kinetic energy, while

to pipe friction.

pp. 202-204.

WATER HAMMER

IN HYDRAULIC PIPE LINES

spectively the velocity and the acceleration at a given- instant,

the pressure at the valve will be less than the statical

pressure under conditions of no flow,

by an amount

Where

-^
at

is

constant the

state

of

affairs

is

shown

graphically in Fig.

Here

AC=ps

represents the statical

GEADUAL CLOSUEE
valve at the instant d
is

5
^(i,

thus given by the ordinate,

I

\pd=

-JL+n+ZiH

while the actual pressure

is

given by the ordinate pi.

Under these
has
its

conditions the pressure in the pipe evidently

maximum

value BQ, at the instant the valve reaches

its seat, this, at

the valve, being given by

P=Pb-j(j^^
and thus being independent

lbs.

per

sq. ft.

of frictional losses in the pipe.

Whether the pipe

affairs is

line is
is

uniform in section or

not,

and

whether the acceleration

uniform or

variable,' the state of

represented by the general equation,

* As is proved in any treatise on Hydraulics, during the steady non-sinuous flow of any non-visoous incompressible fluid the energy

per lb. remains constant, as expressed by the relationship

constant.

w 2g This theorem is called after its discoverer, Bernoulli. In the case of a viscous fluid such as water, however, this is not quite true, since energy is lost in overcoming viscous or frictional resistances. If, as is commonly the case, the work done against such
is is

^ +-_ + z=

resistances

expressed as ^

per
2g

lb.

per unit length of a pipe through

which flow

taking place, the equation becomes

dx\w
that the acceleration
acceleration
is
if

2gm

If in addition the velocity at

is

any point varies with the time, so

the force per lb. necessary to produce this

this force while

^,

and the work; done by

Sa; is

motion

dt

takes place through a distance

-^

Sx.

g dt Since this work is done at the expense of the energy in the water, the equation as finally modified for friction and for acceleration becomes

dx\ w

2g

\g

dt

2gmJ

WATEE HAMMER IN HYDRAULIC

is
ft.,

PIPE LINES
datum
and -j

where z
level in

the height of some point in the iiuid above

x being

its

distance from some

;

arbitrary point,

measured in the direction of flow

v is the velocity

the rate of change with time of the velocity, at this point.

This, which
is

Bernoulli's equation modified to suit the"

accelerated motion of a viscous fluid, expresses the fact that

the rate of change of energy (pressure per

lb. of

+ kinetic + potential)
is

water in the direction of flow,

equal to the force

necessary to produce the required acceleration, minus the

force equivalent to pipe friction.

Integrating both sides of

(2')

with respect to x we get

\sY
'

+iLV.=-ir|.^.-Xrw.+c.
'

^Whenever

g } adt the acceleration is a

2(jr

2gmJ

...

(3) ^

<,

known

function of the time

or of the dist9,nce travelled

by a

particle,

equation (3)

may
flow
of

be solved and the pressure at any point obtained.

As an example,
through a pipe
length
I.

consider the

uniformly

retarded
area
1

of

uniform

cross-sectional

and

Let -3^= a, and

Cut

let

the sufSxes v and

refer to

the pipe immediately behind the valve and to the inlet at the top respectively.
(Fig. 2),

Let x be measured from the

line.

inlet

and

let Va

be the velocity in the pipe

Then

at the inlet,

where x=0, we have p=j>i, 2=i,

[''v\dx=0, so that Jo

c=-^+^+, w 2g

from

(3) '
^

when x=l,
Jo
...
I

i.e.

behind the valve,

j
-

we have p=Pm
7

2='*j

V^dx=Val,

dx=a\ dxa.l;
di (4) ^ '

?l+f+z,=^..l+Pl+f+,,-fl w 2g w 2g 2gm g
z^ is

,But ti-|-ai+--J w Ig

the head equivalent of the statical

GEADUAL CLOSURE
thkt

pressure pa at the vaive with no flow through the pipe, so

we

get

,-^=??+_?!5!(l+=^"ifeet
or

of water,

(5)

p,=p,+^|aZ-?!|^l +|)}lbs.

per

sq. ft.,

the result previously obtained from general considerations.

Obviously this expression has

its

maximum

value

when

Va=0,

i.e.

at the instant the valve reaches its seat.

In order to get uniform retardation of a column by closing

a valve at its lower end, the rate of closure of this valve would

Fig.

2.

however need
effective valve
coefficient

to be

somewhat complicated.

If

is

the pipe

area, and a the effective valve area at any instant (the

area

is

the actual area multiplied by the

of

discharge),

and

if

v^

is

the corresponding

velocity of
pipe-flow,

efflux,

being the corresponding velocity of

so that

we have %=-^-5,

m- aX"' dt^

'

dt}'

WATER HAMMER IN HYDRAULIC

If a, is the effective valve area

PIPE LINES
to

when the valve begins

close, or if Vi is

the corresponding velocity of pipe-flow, the

t

value of a after

seconds

is

given by

'^

= '*|\/2.l+2gh--^{v,+at) m
I.

In an experiment carried out by Mr.

P. Church,! on a

pipe line 8 inches diameter, and 2395 feet long, fitted with a
nozzle 2 inches diameter at its lower end and discharging into

the atmosphere, the nozzle was closed in 25 seconds, so as,

stated, to give

it is

uniform retardation in the pipe.

lbs.

The

statical

head at the valve was 302 feet (131

per

sq. in.),

and the
lbs. lbs.

pressure at the nozzle during steady flow was 108

sq. in.

per per

The maximum pressure attained was 143

j?^ ps=12
lbs.

sq. in., so that

per

sq. in.

Here, assuming a co-efficient of velocity of '985, the velocity

of efflux

would be -985 V2g'X 108x2-31 = 125 ft. per sec, so that the velocity in the pipe would be 7 '81 ft. per sec. This makes a=-312 ft. per sec. per sec, and makes
* If

p=Po, z=2o on the exit side

lb.

energy per

of the valve, assuming that the remains constant during the passage through the valve,

we have ^^ + 5hz = +o::+!!o, w Zg w Zg

al

so that (5) above becomes

+15- =

jg

p,

Va
TiT

fi

Ig

~-

Putting

Pi~P(i

=11,

where h

is

the difference of level

of the free surfaces

on the two sides

of the valve, this

becomes

Vo=>/ 2al + 2gli-i-

m Va^.
-^, and
since =! + a,

From

this

we

get no=''-^=

'"'"'
.

"

where t is the time since the valve began to close, this gives us both the velocity of efflux and the valve opening at this instant, t Journal o/FranUin Institute. April and May 1890.

GKADUAL CLOSURE
62-4 X -312x2395 -"11= w = ,,,^ per sq. ft. 1455 lbs. ^^ = 10'1 lbs. per sq. in.,
J

as

compared with the observed value 12

lbs.

per

sq. in.

Uniform Closing of a Valve.

The most
case to

useful case in practice, however, appears to be

is

that in which the valve

closed uniformly,

and

it is this

which the present investigation has been more

pipe-line, of

particularly applied.

Let the

uniform area a

sq. ft.,

discharge at

its

lower end through a valve into a chamber where the pressure

is

uniformly p^

lbs.

per

sq. ft.

Let times be measured

its seat,

backif

ward from the instant the valve reaches

the valve be closed uniformly in

so that,

seconds, using the

as giving
closure
is

same

notation as before

we have ao=ai-m
t

the valve
complete.
(3)

opening

at

an

instant

sees,

before
line.

Let

Va

be the velocity in the pipe

Equation

now

becomes

w
St

2g

gjodt
f

2gmJ

(6)

being negative.

p=Pi, z=^u

Also

since,

when x=0,

7 j ^dt'^^-^'
fd'o

"1

2g
i.e.

Again, when x=l,

on the outlet side

of the valve,

neglecting losses in the valve,

we have
dv)
I -Y-,-dx=t dt ] ,dt

1)=Vo,

v'^dx

Val,

10

WATEE HAMMEE IN HYDEAULIC

PIPE LINES
(7)

so that

"VV+=^+.-^,^lfe-^|

Writing
difFereiMje

^+^+2, -^= A w w 2g
of statical

feet,

where h

is

the

head on the two sides

(7)

of the valve

with no flow taking place, equation

becomes
^ ^

2g

g\
its

dt

2m J

On

substituting for -^,

value

i*^ + ''o-^[>
;",

^^^^

bee. .. + .*.=|{-:(l + --^)-,*}....(9>

and on dividing throughout by the
co-efficient of

we get

]vbv^c
21 dtta

Icja^-

kaj

If /* is small, so that the '

term i-( .)

m\

is

small in com6, c,

ft

parison with unity, and

ki,

may

therefore be neglected,

and

become constants having the values

0=

21 da.

a dt

3^: c=2gh; ' "

'

k=

2Z

This will always be the case as the valve gets close to

its seat,

and when in consequence the hammer

able.

effect is

most

notice-

In such a case both sides of (10) become integrable.

this in the

Writing

form
^"

-f

U/

,
,

b'V

+ by^

D^,^logJ.
ha,

(11)
^ '

UNIFOEM CLOSUEE
and determining
o

11

from the consideration that when < =7",

v,

has a

known

value

we have

loge

V4
r
or
,

llogj!^.tM=^log.^....(12)
m
^qv
r

r+v^J

hay

-i^

'v.;

_2
qv\
t

ft.

per sec

(13)

where,-=y+^-4
at

,=y.+ ^ + |; .=2^0+^
efflux,

This gives the velocity of

from which the velocity

Va

any instant
(

Mrithin the

range of valve opening over which

is
I

negligibly small,

may be

readily obtained. *

Writing p^ for the difference

of the valve at

of pressure

on the two sides

any instant during

closure,

we

get, as in

equation

(5),

^/

and, on substituting for

from

(8) this

becomes

Where J I _2

is
J

not negligible the treatment follows the lines

outlined on page 28. This case is, however, not of great practical importance, since the rise in pressure before the valve gets near to its seat, and hence before the state of affairs hypothecated obtains, is usually very small.

12

WATER HAMMER IN HYDRAULIC

^,=^,+-^|?^-^-^a|
lbs.

PIPE LINES
. . .

per

sq. ft.,
is

(14)

Evidently this has

its

maximum
(13),

value

when v^
i.e.

maximum

and

therefore,

from

the valve reaches

limiting value
q,

its

when t seat. At
becomes

vanishes,

at the instant

this instant v^ attains the

while

Vg,

zero, so that
lbs.

we get

{:Pv)max=Ps+{^-gh\
But

per

sq. ft.

y^-gh=-+-+-^c+--gh

It follows that the rise in pressure

behind the valve at the

instant

when

closure

is

complete, above that obtaining with

is

no flow through the pipe,

given by

where

a, is the
is

maximum

eflfec'tive

valve opening in sq.

ft.,

and

the time in seconds, taken to close the valve

uniformly.

Pipe of varying cross-sedimal areas.

If

the pipe

line,
Ix

instead of being uniform in section,

of sectional area ax, a length
:

consists of a length
tty,

ly

of area

and so

on,

we

have, at any instant

Velocity

in section

x=Vx=''^

..AcceW.-.,

.=*.=J_(./^+..|}

It

may

then be readily shown that the

maximum

rise in

UNIFOKM CLOSUKE
pressure behind the valve, at the instant of closure,
is

13 given

So
These

far the effects of the elasticity of the pipe line

and

of

the compressibility of the water column have been neglected.

factors,

however, tend to reduce the

to

maximum pressure
effect is likely to

attained,

and the extent

which their

modify the

results obtained above,

may be

seen as follows

Bernoulli's Equation as modified for the elasticity of the water

column.

In any elastic

fluid,

where

is

not constant but depends

equation
(2), p.
5,

on the pressure, on neglecting

friction,

becomes
d /

di\jw2g'^
Now
in water,
if

f^i^iA _ _ J_ ^
)
g

{o'\

df^'^'
mass at atmo-

V be the volume
we have

of unit

spheric pressure {^0),

n-j?-

^~XV

where

is

the modulus of compressibility of the water

of about 43,200,000 lbs. per sq. foot.

and has a mean value

From

this

we have

~^ = 1 + J.
p
lbs.

Pr, if

is

the volume

corresponding to a pressure
spheric,
If w'

per

sq. foot

above atmo-

p=l-|.
and

^^^^'^^

r^^'^-^'^
^^^

are the corresponding ^weights per cubic foot

1

this gives us

^=

- -|,

ui

"^

I^V

From

this

we get [ll'=Ll(l-^)dp=l^ w 2Kw' wj\ A J J w

14

WATER HAMMER IN HYDRAULIC

So that equation
(2')

PIPE LINES

becomes

1+^1+^2

jl=- Let +c.

(3')

the term

^5^* 2aw

representing the resilience, or strain energy

of the liquid, per lb.

Where

the liquid

is

being compressed in an elastic pipe

this requires further modification because of the fact that

part of the energy of the water

pipe.

is

expended in stretching the

the efiect of

This modifies the apparent value of K, and,

the pipe supports be neglected,
follows
:

if

may

be taken into account as

Effect of elasticity of pipe line.

Suppose the pipe to be

tively small thickness
it is
t

of radius r

ft.,

and of comparawhich
sq. ft.,

ft.,

and

let the material of

composed have a modulus

ratio -.
(T

of elasticity

E lbs.

per

and a Poisson's

Then

if

at

any section

of the pipe the increase in pressure

due to retardation
*

is p' lbs.

per

sq. ft.,

the increase in the

ft.,

If

the rise in pressure

is

p'

lbs.

per

sq.

the increase in
is

resilience per cubic foot is

- ,

and therefore per

lb.

-J^

For

consider a cubical block whose sides are of unit length, subjected to a pressure which increases steadily from zero to p'. The change in volume of the block is p' -r K, and therefore the change in length of each

But during the period of compression the mean vaXai of the pressure is half the final pressure p, so that the work done by each of the three opposite pairs of pressures is
side isp' -i-ZK (very approx.).

2
.

ZK
.

6K'
| $

Total work done on block, i. e. resilience stored in block

_ p^
2K'

EFFECT OF ELASTICITY
circumferential stress in the pipe walls
is

15

^
V

and

in the

longitudinal stress
If,

is

-^

lbs.

per

sq. ft.

then,

Sa;

is

the change in length of an element of the

x,

pipe at this point, whose original length was

we have

Sx_ p'r

p'r

tBV
The change given by
in the

2(rJ

volume

of this element is therefore

7r{(r+Sr)'(a;+Sa!) r^'a;},
,,.
c.

>

fneglecting small quantities

I of

the second order.

="-KS('-i)+S('-l)}

F~2m
But the
while
its

<Ty
of

actual

new volume

the liquid

Jir^xf 1

^
\\,

apparent

new volume =7rr

''a;-!

l^^=( 5

so that the effective value of K,

is

which

will be

denoted by K',

given by the relation

For iron pipes

o-

may

be taken as 3'6 approximately.

16

WATEE HAMMEE IN HYDEAULIC

From
equation
(3')

PIPE LINES
is

it

appears that where flow

taking
of the

place so that Bernoulli's equation holds, the ratio

resilient

energy per

lb.

to the pressure energy per

lb.

being

^ this will in all cases be so small as to be negligible.

In the author's experiments the value of K' was found to

be 251,000 100
lbs.

lbs.

per

sq. inch,* so
is

that where

is as

great as

per

sq.

inch this ratio

only 'OOOa.
is,

A
ever,

further effect of the elasticity of the column

how-

more important
is

and,

in fact,

is

all-important where

the time of closing

very short.
fact that, because of the elasticity the

This
efifect

is

due to the

of any change in the pressure, velocity, or acceleration

at the valve, will not be felt at a distance x along the pipe,

until an interval of time x-i-Vp seconds has elapsed.

Here

Vp
to

is

the velocity of propagation of pressure waves along the

water column and has a value generally ranging from 4200

4600
If at

ft.

per second.
t

an instant

seconds before
is

it

reaches its seat the

<^(t),

retardation at the valve

represented by
therefore be

the retar-

dation at a distance x will the valve at an instant

*
x-i-

the same as at
will be

Vp seconds sooner, and

As

will be seen later, the velocity of propagation of pressure

is

waves through water

is

given by K,

=\/ V

to

ft.

per second, where K'

In these experiments this velocity was in lbs. per sq. foot. measured by closing the outlet valve suddenly and timing the oscillations of the pencil lever of a Crosby indicator mounted on the pipe. In this way it was found that the velocity of propagation was 4310 ft. per second, this making ^'=251,000 lbs. per sq. inch. Assuming ^=300,000 lbs. per sq. inch, and taking r=Z'Zt as in this
experimental pipe
line,

this

makes ^=10'

lbs.

per sq. inch.

As

these pipes are cast-iron, and have spigot and faucet joints with lead caulking, this would appear a very probable value for E.

EFFECT OF ELASTICITY
represented by

17

<^(<+^Y

This

is

true at all times, after

closure.

an instant =^ seconds from the beginning of the

If

the time of closing

is

greater than l-^V^,

therefore at the instant of closure, on equating

acceleration) over the whole length
of

we have 2 (massx

pipe to the force

o'-'^Af^'-'^--'^

producing this acceleration

^^

c^~

Now

<t>{t=:o)

is

the retardation at the valve at the instant of

closing, so that the second term,

which

is

negative,

marks

the effect of elasticity in reducing the

attained at the valve.

maximum

pressure

This

may be

evaluated for the case

under consideration by substitution for

substitution in this expression
for;''

-^

from

(8)

and by

in terms of <from (13).*

The

expression, however,

is

too cumbrous for any useful

purpose, and simply serves to

show that

this effect is pro-

portionately greater as the length of pipe increases and as

the time of closure diminishes.

It is usually of importance

for values of T<4:l-i-Vp, but for greater values of

Tmay in

general be neglected without serious error.

In such cases formula (15) requires no correction, and the

first of

the present series of experiments was carried out

of testing its accuracy.
is

with a view
If the

retardation at the valve

uniform, <f>(t+ j^)

= <j>{t) =

-^=a,
dt

at all times after x-i-Vp seconds from the beginning of

If the retardation increases uniformly

the closure.

((>(t+

(dv^\

,^

-^,andwege't' = !^^//'^^

+^

J- 1,

the second term being negative.

18

WATER HAMMER IN HYDRAULIC

3.

PIPE LINES

Description of Apparatus.

by the
of the

The experimental portion of the work was made possible provision, at the Whitworth Engineering Laboratories
Manchester University, of a Cast-Iron Pressure Main,
ft.

3"75 ins. in diameter and approximately 560

long, this

Fio.

3.

bringing water for experimental purposes into the laboratory

from an elevated tank 107

ft.

above the laboratory

floor.
ft.

The main

consists, in chief, of

a vertical leg about 110

ft.

long and a horizontal leg 450

long,

and has eight

right-

angled bends in the course of

portion of the pipe
is

its

length.

The

horizontal
ft.

buried, except for the 80

nearest

to the outlet, the vertical portion being clamped, at intervals, to a wall, but otherwise being free.

The

joints throughout

are of the spigot and faucet type, with lead caulking,

and the

thickness of the metal

is

approximately |

in.

DESCRIPTION OF APPAEATUS
The main
a point 18
is
ft.

19
is

blanked

off at its

lower end, and

tapped at
ins. ins.

from
ft.

diameter and 8
in diameter,

end by a wrought-iron pipe 2^ long, which carries an outlet valve, 2J

this

and

of the design

shown

in Fig. 3.

This,

which

was used in the experiments, has a brass

metal-to-metal on the
flat

disc valve,

bedding
of the

brass seat.

The pressure

water tends to keep the valve

off its seat.

After passing this valve the water rises through a short

vertical leg, passes
fig.

through a second valve, shown at

motion

in

14, this being

is

kept wide open during the experiments,

its
is

and

led into a tank where

steadied

by

baffles.

It is

then allowed to escape over a rectangular notch 4

ins.

wide into a lower calibrated tank.

vided with a
float

The upper tank

scale,

is

pro-

and graduated

from which the

can

quantity per minute passing the notch

ascertained.

be

directly

This
is

scale,

which

is

in general use for laboratory purposes,

checked at frequent intervals, and

may

be taken as giving

results

which are correct within

pressures, a

%.

To measure
A,
5
fig.
ft.

Crosby Steam Engine Indicator

outlet pipe at a point about

4, is

mounted on the
pipe.

from the large

carries a pulley

The drum

of this

indicator

derives its motion from a cord coiled around the spindle B,

which

and

is

driven by a leather band from

a light pulley mounted on the valve spindle S.

R is geared to a
which
carries a

light cylindrical

smoked
is

paper.

The spindle drum B, 5J ins. diameter, As the valve spindle rotates,

by a
fine pointer carried

a wave diagram

traced out on this

by
5

the tuning fork G, which makes 100 complete vibrations

per second.
sees.,

Where
fork

the time of valve closing exceeds about

replaced

this

is

by a

horizontal

pendulum

making eight beats per second.

20

WATER HAMMER IN HYDRAULIC

By

PIPE LINES

counting the number of waves corresponding to the

trayel of the valve, the time of closure can be estimated to

the nearest -005 second in the shorter, and to the nearest

05

second in the longer experiments, while by counting the

DESCRIPTION OF APPAEATUS
number
of

21

waves per unit length

of the paper, the rate of

closing at

any instant can be estimated with considerable

accuracy.

In each experiment, the valve spindle was rotated by hand

by means

of a lever 2 feet long, this being

possible.
it

done as steadily

and uniformly as

After some practice,

became possible to obtain wave

the

diagrams in this way which showed that the valve motion

was

sensibly

uniform

over

whole

range

of

valve

opening.

As a check on the accuracy

two mercury gauges are
a ^
in.

of the indicator,

and

to deterpipei,

mine accurately the resistance to steady flow along the

fitted side

by

side,

one coupled to

lead pipe leading directly to the elevated supply

tank, and thus measuring the statical head, and the second

coupled to the foot of the main near the outlet at P.

third mercury gauge at

enables

the

pressure

usually small

at the

outlet side of the valve, to be deter-

mined.

During

all

experiments the gauge at

P was kept closed.

is

The pitch

of the thread on the screwed valve spindle

125 inch, while measurements showed that the discharge

area for one turn of the spindle was '00680

sq. ft.

The

mean

of a

number

of consistent experiments

with

difiierent

heads and valve openings gave a mean value of '932 for the
co-efficient of discharge,

thus giving an effective discharge

area of "OOGSi sq. area of the main

ft.

for

one turn of the spindle.

As

the

is

approximately '0767

sq. ft., this gives

a ratio of pipe area: effective valve area =12-1

for one
in. all

turn of the spindle, and this value has been used

calculations.

As an examination

of equation (16), p. 13, will show, the

22

WATEE HAMMEK IN HYDKAULIC

length of 2^
in.

PIPE LINES

effect of the short 5-fbot

is

pipe at the outlet

8-foot length

practically the

same as that of an additional

of 3 '75 in. pipe, so that in calculating the

hammer

pressures,

formula (15)

has been used,

being taken as 550

feet.
line,

Although the central portion of the pipe

position, did not
is

being in
this

admit of very accurate measurement,

probably within 2 per cent, of the correct value.

4.

Experimental Results.
Appendix (Table
I.)

These are tabulated as a whole in the

The

following, however, gives a brief resvmi of the experi:

ments

Series

A.

Valve open

'62 of a

complete turn;

= 19 '5.

:^=28-4
i,

ft.

= 12-3 lbs. per sq.

per
sec.

inch.

= 3'03 ft.

A= 104-6

ft.

Experimental results plotted at

dotted curve
Series B.

A (Fig. 5). The A A' shows the calculated results.

Valve open -349 of a complete turn; _=34'7.

^=6-2 1bs. persq. in.

= 2-155

ft. ft.

per

sec.

A= 104-6

Experimental results plotted at

B (Fig. 5).

The

dotted curve BB' shows results as obtained

by

calculation.

<jo

80

70

60-

Hi

_z

^:t

a
if)

a
0-

40-

3;

DC
1/1

50-

:i:

iS

3:

^^
^^
ao^xJi.

==^:;

/o
^c-

-B-

E
10
18

IS

-nipie.

OF c-QSiN<^ Vjilye: in

SeconiPS

Fia. 5. open. those obtaining with steady flow with valve Pressures attained behind valve, in exols of

EXPERIMENTAL KESULTS
Series G.

23

Valve open '167

of a complete turn;

= 72"5.

^-^=1"6

lbs.

per

sq. in.

r= 1-095 ^=105-5

ft.

per

sec.

ft.

Experimental results plotted at

(Fig. 5).

The

dotted curve C'C shows results as obtained

by

calculation.
VfiLVEL

Closed

Pig.

6.

Time

of closing, 1*25 seconds.

increase over pressure obtaining with steady flow = 47 lbs. per sq. in.

Maximum

D.

Valve open Vil complete turns;

=8*6;

^=
2gm
?;

18-0

ft.

=7-8
per

lbs.

per

sq. in.

= 2'405 ft.
ft.

sec.

h=104:
Dotted

Experimental
line

results

plotted at

(Fig.

5).

D'D' shows results as obtained

by-

calculation.

24

WATER HAMMER IN HYPRAULIC PIPE LINES

From
^
c

these

results,

which
to

cover

a ramge of

valms of Vo,from 1-095

3'03 feei per second, a, from 8 '6 to a range of values of the ratio a
:

S"

78'6,
to

and

ha/ve times

of closing varying

from

22

13 seconds, ample verification of the validity of

is

^ M g 5

forrrmla (15)

obtained.

Fig. 6 shows a typical indicator diagram,

and

Fig. 7 the corresponding

wave diagram
of

as obtained

from one of the experiments

Series

I.

g
ft

When
pressure

the valve has reached

falls

its seat

the

below that corresponding to

backwards and
This portion of

the statical head, then rises again, a wave

0)

a
o<

of

pressure being

reflected

2
"^

forwards along the pipe.

J S g a
o
11

the phenomenon, due solely to the elasticity

of the

I ^
j^

water column, will be explained

later.

If the indicator

drum be
of
is

rotated uniformly

and independently

the valve spindle,

'

diagram similar to Fig. 8

obtained.

5.

Gradual opening of Valve.

be gradually opened, so that

g>

If the valve

J
g

both valve area and velocity of pipe flow

increase

with time,

equation

(6.)

becomes
^
'

H M
.2

w
If

2g
the valve

g}dt
is

2gmj>,

opening uniformly, on pro-

Deeding as before (9) becomes

GRADUAL OPENING
and on making the assumption that (^j J
becomes
f .fl' jv^'+bv^ c
fdf +D = -]-\-a kji
1 is

25

small, this

When
Then

<=0,

let

v^^v^, the flow being steady, and let =,.

ao=a,+^=a,+Kt,
at

where

K=^. at
VflSi+Zif/

^Vo+q from which we get

rJ

"'

ft.

per second,

_^_

where
p.

r, j,

m and ^ have

the meanings attached to them on

11 viz.,

This gives the velocity of efiSux after an interval of

seconds from the commencement of the motion, in terms of

SJftLVE.

CLOSED.

Fig.

8.

the valve opening and velocity of eflSux at the latter instant.

This only holds so long as the valve

valve to be stopped at an instant
*

is

opening.
its

Suppose the
is ',

when

opening
10.

Here

6=

and c=2oA as on page

26

WATER HAMMER IN HYDRAULIC

a,! -i-a^n.

PIPE LIPTES
v^'.

the velocity of efflux at this instant being calculated to be

Let

From this time onwards -3^=1

-3^1

and equation

(9') be-

comes^=-[V{l+.l}_,,]
So that

where c =:

/A-+^-r/*
;

^3-

m
Integrating

m
1
,

we get
^/c'+^o
,

j_

t
ifc''

2^/e' ^ V?-o

When
Using

f=0,

i.e.

immediately the valve comes to

rest, o=*'o'-

this to determine

D, we

finally get

as the velocity of efflux after

seconds from the stoppage of

the valve.
It will

be noted that as

increases, this tends to the limit

CHAPTEK

II

Application of the foregoing work to the theory of turbine regulation Regulation of a Pelton wheel of a Pressure turbine Speed regulation assuming uniform acceleration in the pipe line Effect of a stand pipe on speed regulation.

1.

Application of the foregoing

work

to the Theory

OF Turbine Eegulation. In the ease

line is well

of a Pelton

wheel

installation,

where the pipe

velocity of

designed so as to give a

maximum

flow of about 5 feet per second, the term (-t) \ a /

JL

will in

general be negligibly small over the whole range of nozzle

opening, and

may be treated as being so in equation (9'), p.

error.

24,

without sensible

For example, consider a Pelton wheel working under a

head of 200
diiimeter
ft.,

and supplied through a

ft.

pipe-line 2

ft.

in

and 500

long

(/= -006).

Here /?-= -006x500x2 = 6

and, if=5,

2gm

^=2-34

ft.

The
975

velocity of efflux will then be approximately equal to

^2^(200- 2 -34) = 110

.

ft.

per

sec.

\a/n
.-.

iiir

_L. 22

{(^YM =^=-0124. \.\a/ mjmaa

is little

In this case the energy lost in the pipe line

more

28

WATEK HAMMEK IN HYDRAULIC

be
still

PIPE LINES

than one per cent, of the energy of discharge, while this value
will

further reduced as the valve closes.

is

Where

the valve

replaced

by the regulating

sluice, or
still

the gates of a pressure turbine, the friction term will

general be negligibly small where the ratio

in

Z-e-A is less

than

about

4.

'

Where

this ratio is large, as

is

where the supply, under a which should never be

comparatively low head,

brought a considerable distance

of affairs

through a pipe line

allowed in good practice

into account.

a state
24,

this

term must, however, be taken

In such a case the

p.

difficulty in integrating equations (9)

10,

or

(9') p.

may be overcome by

dividing the

period during which the gates are in motion, into a series of

short intervals during each of which the friction term

may

be taken as sensibly constant

(a,

having

its

mean value

throughout the interval though varying from interval to

interval).
If

the velocity at any instant be known, the

velocity at the beginning or end of the interval containing

that instant, and thus the velocity at the beginning or end

of the preceding or following interval

Proceeding in this
the whole period

may thus be obtained. way the velocity' at any instant during may be obtained. This case is, however,
fitted as

not so important, since in such an installation a stand pipe

would invariably be

near to the turbine as practic-

able, the effect of this being to greatly reduce the accelera-

tion along the supply main.

Neglecting this for the time being,

it

may

be postulated

that for successful governing on an increasing load,* the

*

By

at the turbine,
load.

the provision of an adequate relief valve or pressure regulator no difficulty is experienced in governing on a falling

TUKBINE KEGULATION
inertia of the rotating parts of the turbine

29

must be

so great

that the energy rendered available during a certain (fixed)

reduction in speed (the

strictly limited

maximum permissible

reduction being

by the conditions

of service), together with

the energy entering the turbine during the period of pipe

line acceleration, will equal the

demand

for energy during

that period by the driven machine.

Two

cases

may

be considered
(a) Pelton Wheel.

Here the energy entering the wheel per second, when the
jet velocity is v^
ft.

per

sec.

=-^-5-

ft. lbs.

Suppose the wheel to be running steadily under a load ft. lbs. per second, which necessitates an energy supply of and suppose the load to be suddenly increased and to

demand ft. lbs. per second. The original amount of energy enter-\ _ j^_ a(,%'
ing wheel per second
-'

,,

^9
If

Next
uniform,

let

^,
at

the rate of nozzle opening, be known.

in terms of
asi

is

known

and

t,

from

(18), p. 26, so

that, equating

^^
^9

to

M, we get the time

after

which the

supply

is

once more equal to the demand.

the

Let

this interval of time=ti.. this

interval,

Then, during

the

excess

of

energy

demanded over that supplied by the water

=/:
and
this

-dt-ML
2^

ft. lbs.

must be capable of being supplied by the kinetic given out by the rotating parts during a fixed reducenergy
tion of speed.

30

WATER HAMMER IN HYDRAULIC

Thus
if

PIPE LINES
is

(0= angular

vel. of rotation

when

load

thrown on

(radians per second)

o)i= reduced angular velocity,

1

= moment =
Wr'
9

of inertia of

wheel
of gyration

where r=its radius

we have
(5)

Presswre Turhine.

Exactly the same reasoning and method of solution holds

for this class of turbine, except that the

energy entering the

wheel
is

is

now

partly in the potential form, and, per second,

given by

where the
runner.

suffix

refers to the state of affairs at the exit

from the guide vanes and at the entrance to the turbine

Now

if

the datum plane be taken as passing through the

if

turbine 2o=Oj ^^nd

in addition

we

assume, as

is

the case

with a well-designed turbine running steadily, that

^
1)

bears

a definite ratio to

-?-,

7} V so that -5-= J, -5-,

1g

we have
>.*

2g

Energy entering wheel per second=-5-5-| l+ki

*

Here

^ has the following approximate values

:

'

TUEBINE KEGULATION
N.B.

31

Vo is

no longer the velocity of

7)

efflux

due to

head h but to a head

hw
running quite steadily

The

facts that the turbine is not

during this period, and that, owing to the variation in the

velocity of inflow, its efficiency
is

not constant, prevent any

accurate calculations being made, however, and

when

it is

remembered how comparatively far our preliminary data may be from being accurate, and that the expressions resulting from this method of treatment are most unwieldy,
it

appears preferable on
to simplify the

all

counts, from a practical point of

view,

treatment
is

by assuming that the

acceleration in the pipe line

uniform.

On
is

this assumption,

which cannot be widely at variance

with the conditions actually obtaining in practice, and which

probably more justifiable than that of uniform valve

opening in the case of a modern turbine governed through

an hydraulic

relay,
:

the whole

question

may

be readily

examined as follows
2.

Speed Eegulation assuming Uniform Pipe-Line

Acceleration.
sq. ft.

Let a=area of penstock, at entrance to turbine casing,

A = supply head in

feet.

?)i= velocity of steady flow through pipe line under the

least probable load.

?>2= velocity of steady flow through pipe line under the

greatest probable load.

Ag= pressure head at entrance to turbine casing, in feet. Then when turbine is running steadily under least load,

the energy entering casing per second

= &2-im,{h-'"^\

ft. lbs.

32

WATER HAMMER IN HYDRAULIC

And when running
steadily

PIPE LINES
this

under greatest load

energy per second

= Q2-4.a/vJh-^J^]
I

itAha.

2gmj

During the
the value

transition period, after the extra load has been

thrown on and before the velocity of pipe flow has attained

v^'

necessary for giving the required supply, the

is

water column
a

being accelerated, and during this period,

if

is the uniform acceleration,

t

and v the

velocity,
:

we

have,

at an instant

seconds after throwing on the load

^^^

Energy entering wheel per sec. = 62-4a-[ A

[ft. lbs.

But v=Vi+ai, so that this becomes Energy entering wheel per sec.

= 62M^, + at)h-"^-^^4^^]
If
<j

ft. lbs.

is

the length of the transition period, in seconds, so

that (fa l)-^<l=a,

we have the

total

energy entering the

wheel from the penstock during this period given by

62.4<./;.[(.,+<.0{-|-&i^}]^Ub.
an expression which
terms involving the
small, in
is

readily integrated.

Usually the

co-efficient of friction /, are negligibly

which case we get

transition period
dt
ft. lbs.

Energy entering wheel during

= 62-4a/"*'r(.+a<){A--}1
=E^
If
ft. lbs.

the

maximum

value of

n,

consistent with good speed

regulation, be

known from

experience of existing plants, the

foregoing expression

may

be solved numerically ; and so far

as the data at the author's disposal go, these tend to

show

TUEBINE EEGULATION
that the acceleration should not exceed the value given

33

by

the formula

a=2'4-=V

ft.

per

sec.

per

sec.

Now,

for satisfactory governing, the total

demand during

the transition period must not be greater than the supply

entering at the penstock together with, the energy rendered

available

by the slowing
and and
la^

of the flywheel

and rotating

parts,

so that

if (Oi

are the angular velocities before throwing

on the

load,

after the speed has settled

down under

the

new

load,

we have
is

where

e,

the

demand

for energy, per second,

under the

increased load.

3.

Efebct of Stand Pipe on Speed Regulation.

pipe, area

Suppose a stand
stand pipe.

sq.

ft.,

to be fitted at the
is

entrance to the turbine casing, so that hg

the head in the

Then with steady

flow at vel.

Vi

As=A ^|l+-^|
^s=h-^-{l-\
f

ft.

and with steady flow at

vel.

v^

ft.

So
|"l

that,

during the transition period, a volume of water

"^'"l

4-^11^'"

cub.

ft.

has flowed into the turbine

casing from the stand pipe under a

mean head

h' feet,

and

has carried in an amount of energy approximately equal to

= Es it.

Ihs.

Here h'^h-'^^L+^^i + i-j approx.

c

34

WATER HAMMEE IN HYDRAULIC

In this case then,

PIPE LINES

we have,

for satisfactory speed regulation,

an expression from which

may be

obtained

if Wi,

toj, a,

and

are known, or from which the necessary value of

if

may

be obtained

the other factors are known.

As an example, consider a turbine supplied under a head of 80 feet through a penstock 4 feet in diameter and 250 feet long, and working under a normal load of 400 b.h.p. stand pipe is to be designed to keep the speed -within 4 % of the normal (250 revs.), under an increase of load up to 600 b.h.p., the turbine being fitted with a fly-wheel whose

moment

of inertia
?

= 7000 foot

lb. units.

Assume the

efficiency to

be constant, and equal to

'80.

This

necessitates a supply of energy-)

"8 under normal load Also since the energy then entering the"! _ c:o-a^ (j, _ ?i!/?\ ^\ / 2ymj wheel casing per second where a=Jl2'57 sq. ft. ; h = 80 ft. ; /='0O4 say; this gives us on substitution and reduction
. .
:

.J

^ 400x550 ^ ^^

^^_

^^^

^^

Vi

= 4"41

ft.

per

sec.

while with steady flow under the increased load 'U2 = 6"64 ft. per sec.

Now in the first case

the head at the stand pipe=fe - J-! 1

^

+<

2g l = 80--303{l

m]

l-ffc.

+ l}

= 79-394
while in the second case the head
is

ft.

h- ^-j 1 +!-- \
2g\.

TO

= 80- 2 X -685 = 78-630 ft.

Fall in level at stand pipe

-764
ft.,

ft.

The mean height

energy leaving
it

in stand pipe = 79

so that if J.

is its area,

the

during the transition = -764 X 79 X 62-4 x^

ft. lbs.

=3760^

ft. lbs.

Now putting
after

a = 2-4^
b

= 2-4 x -- = -768

f.s.s.,

so that the velocity

is

an interval of

^50 seconds from throwing on the load

4-41

+ 768(

TUEBINE KEGULATION
ft.

35

per sec, we get the energy entering the casing from the penstock during the transition period of tj seconds (J?i, p. 32), given by
62-4a f\i-4:l

+ 768t)|80 - '2^1^^\dt
dt foot lbs.
ft.

foot lbs.

= 784 /''i(327 + 57)

J

= 784{327ti + 28-5fi2}
On
putting
1^

lbs.

=
(

'768

= 2 '9, this expression reduces to

784{950 + 83} = 800,000 ft. lbs. If now <B is the mean angular velocity of the runner in radians per second, and if 8E is the energy given out by the wheel in slowing down through So), it is easily shown that
8(0

_ SM
"04
;

But

=
0)

and

a>

= 277

X 250
'

so that

hE=

60 04 X 7000 X
9

ff2

X 625

Now under

= 192,000 ft. lbs. the increased load, the demand per second
600x550^^^
8
'

is

OQ

f^

H,3

so that for effective regulation

we have
x
ij

37604 + 800,000 + 192,000 = 412,500

.-.

= 1,196,000

4 = 54-3 sq.ft.

corresponding to a diameter at the top, of 8 ft. 4 inches. This stand pipe would probably take the form of a vertical pipe 4 ft. in diameter and carrying a circular cistern at the top 8 ft. 4 in. diameter. The top of this would be about 82 ft. above the centre of the turbine,

and

its

bottom about 77

ft.

above the same

level.

The method here outlined is equally applicable to the case of a Pelton Wheel or a Pressure Turbine, and while the results
obtained are admittedly only rough approximations, yet they
are suflBciently near for
all practical

purposes.

CHAPTER III Sudden stoppage of Motion Theory Valve closed

quickly but not instantaneously Experimental results Sudden closure in a nonuniform Pipe Line Sudden opening of a Valve Sudden opening, neglecting effect of elasticity.

1.

Sudden Stoppage of Motion

have

Ideal

Case.

If a column of water, flowing with velocity v along a

uniform pipe (supposed

rigid),

its

motion checked by

the instantaneous closure of a rigid valve, the phenomena

experienced are due entirely to the elasticity of the column,

and are analogous to those obtaining in the case of the

tudinal impact of an elastic bar against a rigid wall.

longi-

At

the instant of impact, the motion of the layer in conis

tact with the valve

suddenly stopped, and

resilience, or

its

kinetic

energy

is

changed into

energy of

strain,

with

a consequent sudden rise in pressure.

rise in pressure is

This stoppage and

almost instantaneously transmitted to the

on, the state of zero velocity
p'

adjacent layer,

and so

and

maximum
v)

pressure (this at any point being

above the

pressure obtaining at that point with steady flow at velocity

being propagated as a pressure-wave along the pipe, with

Fp.

velocity

[P^
i.e.

is

the same as the velocity of

ft.

sound

through water,
slightly

about 4700

per second, depending

on temperature.]
i

This wave reaches the open end of the pipe after

where t^l-~Vp.
is

seconds,

At

this instant the

whole of the column

instantaneously at rest in a state of compression.

SUDDEN STOPPAGE OF MOTION

At the open
tained,

37
is

end, however, a constant pressure p^

main-

and

in consequence the strain energy of the

end

layer

is

reconverted into kinetic energy, this (neglecting

its

losses),

rebounding with

original velocity v
this point

and with the

of

normal pressure obtaining at

under a state

steady flow towards the open end with this velocity.

This state of normal pressure and of

velocity

{v)

is

then propagated as a wave towards the valve, reaching the

latter after a second interval

l-^Vp seconds.
is

At

this latter
is

instant the whole of the column

unstrained and
v.

moving
instant

towards the open end with velocity

At the same
is
{jp'

the motion of the layer nearest the valve

stopped,

and a

wave

of zero velocity

and

of pressure

below the pressure

obtaining at the point at the instant before the stoppage of

the motion, or

p below

the pressure at the point with no

is

flow through the pipe)

Reflected

transmitted along the pipe to be

of

from the open end as a wave

v towards the valve.

normal pressure

and velocity

When

this
is

wave reaches
and the

the valve, U-i-Vp seconds after the latter

ditions are the

closed, the con-

same

as at the beginning of the cycle

whole

is

repeated.
ideal conditions the state of affairs behind the

Under such

valve, as regards pressure,

would be represented by such a

an
elastic,

diagram as Fig. 9 A, the

cycle, in the case of

non-

viscous fluid, being repeated indefinitely.

At any other

point in the pipe, at a distance

l^

from the open end, the

pressure-time diagram would appear as in Fig. 9 B.

Actually the valve does not close instantaneously, while

the stretching of the pipe and the viscosity of the water
cause the

maximum

pressure attained to be less than in the

also

ideal rigid pipe,

and

cause the

pressure waves to

diminish rapidly in amplitude.

The

state of affairs is then

38

WATER HAMMER IN HYDRAULIC

which

PIPE LINES

as indicated in Fig. 10, of

represents a diagram

from behind the valve, and dpen end of the pipe

line

B from a point

15 feet from the

experimented on by the author.

In this case the valve was closed in -07 seconds, and the

StTR-ric

ffTflOS

STKTie

"tmos
Fig.
9.

Press

vibrations died out so that the motion of the pencil of the

indicator

became imperceptible,

after

about 30 complete

oscillations.

Magnitude of rise in pressure at

If p'

the

sudden closing of a Valve.

per
sq. ft.,

be the

rise in pressure in lbs.

and

if

be

the velocity of flow at the instant of stoppage (supposed

SUDDEN CLOSUKE
instantaneous),

39

we

have, assuming the pipe line rigid, on

lb.

equating the loss of kinetic energy per

resilience
:

to the increase in

2g~'2Kw

^=^Jl
Putting
this

fKw
(19)
ft.
;

K= 300,000 X 144 lbs. per sq.

p=9160

w= 62'4
ft.

gr=32-2

becomes

lbs. per sq.

= 63-7 V lbs.

per

sq. inch.

A closer approximation

to the actual rise in pressure

may

be obtained by assuming that while the pipe

line is rigid

ST/trie

Pj^a&s

L,^

ffrnos

PtlS&,

Hthos,

._L ^
is

i___j_ t
Fig. 10.

that the motion

is felt

instantaneously stopped, yet the elasticity

in its effect

on the value of K,

this

adopting the value

K' where

1^=^+^(5-1).

P. 15.

In the experimental pipe line this makes ^'=251,000 58 '4 v lbs. per sq. inch. 144 lbs. per sq. ft., and makes

y=

40

WATEE HAMMER IN HYDEAULIC

The following demonstration shows how the

PIPE LINES
elasticity of the

pipe line and water column

may
Vm,
o,

be taken fully into account.

Let K' and E have the meanings already attached to them,

and

let

and Wm,
of,

Fand

and a^, be the weights

in,

of

unit volume

the velocities of wave propagation

and

the sectional areas of the water column and metal of the

pipe wall respectively.

Then, with instantaneous closure the ends of the water

and metal columns move, at impact, with a common

u,

velocity

and waves, respectively

of compression

and

of extension,

travel along the water

column and the pipe

wall.
St,

Hence, after a very short interval of time

lengths FSi

and Vm^t

of the water
u,

column and

of the pipe will be

moving
:

with velocity

momentum {wa V+ WmO-m VrnflJ^t = Wa VvU

and the equation
1

of

gives us

='ij^waF'

waV
Each element
passes
of the
it,

of the

column and

of the pipe, as the__wave

u,

takes suddenly the velocity

while each element

water column takes the compression

^^^ and

there-

fore the stress (' m)\/

and each element

stress u.

of the pipe

takes the extension

^==^

and the

Substituting for m

we have

9 the pressure rise in the water

^/
^

^"'^
'

given by

/=*"],
^

waF \x/
U)mamym,]__

H^s.

per

sq.

ft.

Since

V=J^ and V
Zi?

V,^=J^ * this may be written V

Win

unit cross - sectional area to impinge with velocity v in the direction of its axis, against a rigid wall. After
of

* Imagine a bar

SUDDEN CLOSUKE

41

The
longitudinal stress
action,
/,

j=

lbs.

per

sq.

ft.

v/A^^sl;/hammer
which equals u
,

(=)

produced in the pipe walls by

/ '"''m^ then becomes, on

substitution,
{

__
/_2_
,

1
\

/='i

IV Ewm'^
pressure
is

H^s. per sq. g iT'wJ

ft.

while since the circumferential stress in a pipe exposed to

twice the
longitudinal
stress,

the

maximum
is

increment of stress in the metal, due to this action,

to 2 /.

equal

2.

Valve shut suddenly, but not instantaneously.

of closure of a valve becomes less andr less,
rise in pressure

As the time
the

maximum

will evidently

tend to the

limit given

by formula

20, p. 41.
is

Now

if

the time of closure

so short that

= > T= ^,

a very short interval St seconds, a mass "'

"*

has been brought to

9
and, if p is the (uniform) pressure on the end of the bar during this interval we have, equating the force x time, to the ehan'ge of
rest,

momentum

^8< = '2LZ!f

orp = '^

V^-

V lbs.

But p = vk/
get

, so

that, equating these

two expressions we

V~ = \/

it.

per> second.

42
the

WATER HAMMER IN HYDRAULIC

disturbance
x,

PIPE LINES
travelled

initiated

at

the valve

has

distance

and has not arrived at the open end when the

the retardation
/

latter reaches its seat.

In this case
(17), p. 17,

if

is

uniform

= a),

equation

becomes v

adx=-^ lbs.
gjo
9

Buta=^=?!^,

sothat

9
this being the value obtaining

^9
with instantaneous stoppage.
closing,
if

It follows that

whatever the law of valve

less

this is

completed in a

time than l-i-Vp the pressure rise will be

the same as with instantaneous closure.

For values of

T between

l-i-

Vp and

ll-i-Vp the falling off in

it is

pressure will usually be comparatively small, so that

in general sufficiently accurate for all practical purposes to

count as 'sudden,' any stoppage occupying a shorter time

than

this.

For values

of

T>'2,l-i-Vp,

formula

(1),

p.

3,

when
is

modified for the effect of elasticity as indicated on p. 17,

to be used, while this elasticity correction

may

reasonably

be neglected where r>4Z-i-F^.

3.

Experimental Results with sudden closing OP THE Valve.

iT'

Taking
sq. ft.
;

= 25 1,000x144;
(p. 16), in

.=10,000,000x144
formula
(20), this

lbs.

per

=1'275;

becomes

^=7780
=54'0

V lbs.

per sq. foot

V lbs. per sq. inch.

EXPERIMENTAL RESULTS

43

For comparison the results observed by the author are shown in Table I. against those obtained by using this formula.

Table L

Experiment

44

WATER HAMMEE IN HYDEAULIC

From
-13

PIPE LINES

these results it appears that so long as

is

less

than

seconds (l-~Fp=-l3) the calculated and observed

pressures are in every case in close agreement.

This agree"21 seconds,

ment
while

is

substantially maintained until

T= about
is

when T='26

seconds, the

mean

error involved in

using the uncorrected formula in this case

about 14%.

Other Experimental Besults.

In a series of experiments carried out by M. Joukowsky *

on cast-iron pipes

of 4"

and

6" diameter, having lengths of

1050 and 1066 feet respectively, the time of valve closing

being '03 seconds in each case, the observed
agrees closely with the formula, p=57'V.
are
rise in pressure

The

following

some

of the results obtained

by interpolation from the

plotted results of these experiments.

4-INCH Pipe.
Vel.

ft.

per sec

SUDDEN CLOSURE
4.

45

Sudden Closure of a Valve

in a Pipe Line of non-unifokm section.

In such a case the phenomena become very complicated.

Let

li, l^, Is,

etc.,

be the lengths of successive sections of the

pipe, of areas a^, a^, a^,

extremity of the length

and suppose the valve to be at the l^. Imagine the pipes rigid and the
Following sudden closure of
the

water incompressible.
valve, a

wave

of zero velocity
is

and

of pressure (63-7i lbs.

per

sq.

inch above normal)

then transmitted to the juncmaintained during the

is

tion of pipes 1

and

2.

Here the pressure changes suddenly

This
is

to 63'7j above normal.

passage of the wave through the second pipe, and

followed

by a change of pressure to GS'lv, at the junction of 2 and 3, and so on to the end of the pipe line. But immediately the pressure at the junction of 1 and 2 attains its value 63-72,
the wave in pipe 1
of pressure 63-72
flected
is

reflected

back to the valve, as a wave

and

of velocity

i Wj, being again

zero
velocity

re-

from the valve as a wave of

63'7{)2

and

pressure

(tii Wj)}

above normal.
1,

This wave then travels to and fro along the pipe

making
and
3

a complete journey in h-r-Fp seconds, until such time as

the wave in pipe
3, reflected

from the junction

of 2

with pressure 63'73 above normal and with velocity v^v,,

again reaches the junction of
1

and

2.

This occurs at an

instant ?2-rf^ seconds after its reflection.

it

At

this instant

takes up a velocity and pressure corresponding to the

velocity

and pressure at the junction end

of pipe

1,

and as

this pressure

and velocity may be either positive or nega-

tive

depending on the ratio of the lengths of the branches

1 and 2, it is evident that after the first passage of the wave the pressure conditions at any particular instant are The greater the number of variapractically indeterminate.

46

WATER HAMMER IN HYDRAULIC

PIPE LINES

tions in area

and the more involved does the phenomenon

is

hecome.

Where a

pipe

very short, the period of the oscillations

of

of pressure at

any point becomes so small that the pencil

is

the ordinary indicator

records the

unable to record them, and simply

mean pressure in the pipe. Thus where a short branch of comparatively small
is

diameter
bore,

used as the outlet from a long pipe of larger

the pressure as recorded

by an

indicator will be

sensibly the

same at any

instant, as in the large pipe at the

point of attachment of the outlet branch.

This point, as well as the

effect of

an enlargement of the

pipe section on the magnitude of the

well brought out

hammer

pressure,

is

by the following

results of experiments

by E. B. Weston.*
In each experiment of this series the outlet valve was
closed suddenly, but the exact time of closing
f
is

not given.

111

ft.

of 6" pipe.

Series I.

Pipe line consisting of

58

EXPEKIMENTAL RESULTS
'111
ft.

47

of 6" pipe.

58
Series II.

Pipe line consisting of

48

WATEE HAMMER IN HYDRAULIC

5.

PIPE LINES

Sudden opening of a Valve.

falls

If

the valve at the lower end of a pipe line be suddenly

opened, the pressure behind the valve

by an amount
towards the

lbs.

per

sq.

inch,

and a wave
(approx.)
[,
'

of velocity v

valve \'n=pf,/
I

^ K'w
of

and

of

pressure

below

statical, is

propagated towards the pipe

inlet.

The magnitude

depends on the speed and amount

if

of

opening of the valve, and

the latter could be thrown wide

fall to

open instantaneously the pressure would

ing on the discharge
side.

that obtain-

In the author's experiments,

'5

with the valve thrown open through

the

of a complete turn
lbs.

maximum drop

in pressure
lbs.

was 40

per

sq. inch,

the

statical pressure

being 45

per

sq. inch.

With the

valve opened through '10 of a complete turn the

lbs.

maximum drop was 20

case the time of opening

per

sq. inch,
lbs.

and with
sq. inch.

of a

complete turn the drop was 11

per

In each

was

less

than '13 seconds (l-^Vp).

is

In the case of a horizontal pipe, or one which

that the absolute statical pressure
is

so situated

everywhere greater

than

p,

this

pressure wave reaches the pipe inlet with

its original
is

approximately

amplitude, and at this instant the

velocity v

whole column

moving towards the valve with

and pressure p below normal.

The pressure at the inlet is, however, maintained normal, wave returns from this end with normal pressure and with velocity 2v. At the valve this wave is reflected
so that the

with a velocity which

is

the difference between 2v and the

velocity of efflux at this instant, and with a corresponding

pressure.

As the velocity of efflux will now be greater than V, the wave velocity will be less than v, and the rise in This wave is reflected pressure less than p above normal.

SUDDEN OPENING OF VALVE

from the
inlet to the valve,

49
is

and here the cycle

repeated,

the amplitude of the pressure wave diminishing rapidly until

steady flow ensues.

Fig. 11 shows a diagram obtained

by the author from the

its inlet, so that,

experimental pipe line under these conditions.

Where
pressure

the pipe slopes upwards towards

beyond a certain point

is less

in its length the absolute statical

than

the drop in pressure

caused at the valve

by sudden opening,
then on the passage
of the first

wave

of

^''*- ^^

negative pressure the wave motion becomes partially discon-

tinuous after this point

is

reached,* and the wave travels on

Vru/c. opeNtij.

Fig. 12.

to the inlet with a gradually diminishing amplitude.

The

amplitude with which

probably 2 or 3

it

reaches the inlet, and which will be

lbs. per.sq. in. less

than the absolute

statical

pressure at inlet, decides the state of velocity of the reflected

* Actually before this, since

water gives up

its

dissolved air rapidly

when the pressure

falls to

within 2 or 3

lbs. of

a complete vacuum.

50

WATER HAMMER IN HYDRAULIC

PIPE LINES

wave.
case,

This will evidently be less than in the' preceding

and under such circumstances the wave motion dies out very quickly. As the valve opening becomes greater, the efficiency of the valve as a reflecting surface becomes
less,

so that with a moderate opening the pressure

may

never even attain the pressure due to the statical head.

This
is

shown

in Fig. 12,

which

is

a diagram obtained by

the author from the experimental pipe line

when

the outlet

valve was opened suddenly (time <-13 sees.) through half a

complete turn.
6.

Sudden Opening neglecting Elasticity of

Water Column.
Neglecting the
effect of elasticity of

the water column, the

pressure behind the valve and the velocity of efflux accompany-

ing a sudden valve opening

may be

calculated as follows

Imagine the

effective
a

valve opening to instantaneously

of
this

assume a value

and to remain
(6), (p. 9),

value,

so

that

^=^ a
at

^. at

Equation

now becomes

w^2g^
dvt

gj.df

2gmJ,

'

while equation (9) becomes

"'-di .^=l{^('+^(^)")-^*}

<>

or

k'f2=c'-<at

where

SUDDEN OPENING OF VALVE

so that

51

j^^ + r=ijit
we
get
(11")

From

this

i-log?^ + Z? = i-,
while since
Vi,

=
;;

when i=0, we have I)=0.

^
2*
.

\it.

+ e-i

per sec,

(12")

' J

giving the velocity of efflux at an instant

valve opens.

seconds after the

As

increases, this approximates to the value c or

'2gh

J
By
substitution in (9") the value of
-j-i

at

any instant and

this

therefore of

-j:

may be

obtained,

and knowing

the

pressure behind the valve

may be

readily obtained.

series of experiments recently carried out at the Uni-

versity of Wisconsin * on a hydraulic ram, in which the

velocity of flow

down

the drive-pipe, corresponding to a

given interval of valve opening was measured, enable the

results obtained

by using formula

(12"),

or

its

modification,

to be verified.

Values of
85 '35
ft.

Va,

as obtained

by experiment on a

drive-pipe

long and 2'05 inches diameter, are plotted in

In this

Fig. 13 against the curve representing this formula.

case

/= '09

(by experiment); a-7-a|,=:3-97.

No. 205, 1908,
p. 143.

* Bvlletm of the University of Wisconsin,

52

WATEK HAMMEK IN HYDRAULIC

its seat

PIPE LINES

The valve used

consisted of a flat disc 4*5 ins. in diameter,

the inside diameter of

being 2'75

ins.

Its

lift

was

Fig. 13.

5 ins.,

and in every case the time

of

opening was too short

to be measured.

The experimenter

states that in these experiments the

recorded velocities are somewhat low owing to the effect of

the friction of the rod carrying the disc used to measure the
velocity.

The agreement between observed and

calculated values

is,

considering the nature of the apparatus, very close indeed.

CHAPTER IV
1.

Further Phenomena connected with Pipe flow.

carrying out the experiments already described,
.

While
further

phenomena were observed by the author which would appear to be of some importance

when

considering the

question of the fluctuations of

pressure likely to be obtained
in a given pipe line.

To
it

follow these

more

clearly

should be understood that the

water after passing the main

regulating valve

(Fig.

14),

passes through a second valve

B, before being measured.

the

In
de-

experiments

already

scribed the valve

was kept
differs

wide open. from

This valve

Fig. 14.

in only one particular.

its seat,

Instead of bedding metal to metal on

washer,
Fig. 15,
seat.

a thin leather
as

^Y

thick, is fixed to the valve

is

body

shown

in
its

and

thus interposed between the valve and

In one or two experiments, the valve

being wide open,

was opened

slightly.

Whenever
pipe,

noticed that pressure waves were set

from end to end of the

but

it

it was up and were reflected was found that these,

this

was done

d2

54

WATEE HAMMER IN HYDEAULIC

maximum,
after

PIPE LINES
up
to a

instead of dying out actually increased in amplitude

certain

which this amplitude was maintained

seemingly indefinitely.
It

having been noted within what range of opening of

this occurred,

was closed and

set to

some point within

this

range.

was

then

opened very slowly so as not

h-':fi:
-

to set

up any
of

perceptible dis-

turbance
o-

the

indicator

lever.

Flow would then take

steadily
for

place

few

seconds or even minutes, but

always,
Fig. 15.

after

short

time

the
slightly,

pencil

would

tremble

and the
as

oscillations

would grow
sq. inch).*

until the amplitude

approached that obtained by a sudden opening of the valve

(as

much

28

lbs.

per

In fact experiment

showed that over a certain range

impossible to maintain steady flow.

of valve

opening

it

was

On

entirely closing valve B, the oscillations died out, how-

ever, in about

30 seconds.

In every case the period of a

complete oscillation of the pencil lever was '26 seconds.

When

it

is

remembered that the pressure of the water

its seat, this

tends to keep the valve off

phenomenon seems,
fact of the

at first sight, very paradoxical.

Its explanation

may, however, be found in the

leather washer fixed to the valve body, for while the pressure
in the annular space between this washer and the valve
is

body
c.

the same as that on the outlet side of the valve

Fig. 15

that between the washer and the valve seat

at a or
will

be

* Data obtained from a number of the experiments are given in the Appendix, Table II,

FUKTHEE PHENOMENA
less

55

than this (by Bernoulli's theorem), since the velocity

c is

of flow at

less

than at any point between

and

6.

Also,

the greater the volume passing the valve per

second at
c,

any

instant, the greater will

be the pressure at a and at

and, for a given valve

difference of pressure

opening, the greater will be the

of the washer,

between the two sides

its seat.

tending to force the latter to

Now

a sudden opening

is

followed by a reduction of
is

pressure behind the valve and flow

attains a certain value
v,

initiated.

When

this

depending on the valve opening,

the difference of pressure on the two sides of the washer

becomes

sufficient to cause

it

to flutter to its seat.

The
end

pressure rises suddenly due to the stoppage of motion and a

wave

of pressure

above normal
it is is

is reflected

to the open

of the pipe.

Here
This

reflected with

normal pressure and

velocity v.

reflected

from the valve with pressure

p below normal and

zero velocity, and again from the open

end with normal pressure and velocity v towards the valve.

But following the drop

its

in pressure, the washer will leave

seat
is

and flow

is

set up, so that this state of velocity v

is

which

propagated towards the valve

v'.

superposed on a
its seat

velocity of flow

As

the valve flutters to

under

the

influence

of

the increasing pressure,

the

maximum
v.

pressure rise will

now be

proportional to v+v' instead of to

is

The

cycle as outlined above

thus repeated, but with a

pressure variation which becomes greater at each succeeding

cycle until such time as a balance
is

obtained between the

energy entering the pipe during each cycle and that expended
in stretching its walls

and in giving kinetic energy to the

escaping water.

The
flow
is

initiation of this

phenomenon from a

state of steady

in all probability

due to pressure fluctuations caused

56

"WATER

HAMMER

IN HYDRAULIC PIPE LINES

side of the

by uneven eddy formation on the discharge

valve.

The above explanation

with valve

is

borne out by the fact that

oscillations

it

was

found impossible to obtain these

with valve A, or

when the

leather washer
affairs

was removed. any main

fitted

Evidently this state of

may

exist in

with an outlet valve of a type similar to that illustrated, and

it

would appear very inadvisable to use any such valve on a

branch pipe communicating with a pipe line used for supplying

a prime mover, or in any other case where (as

when coupled
acci-

up

to a

water meter) such a fluctuation of pressure,

dentally set up, would prove detrimental.

CD

P
Ph
P-i

II

g s ^

II

8|8

C3

APPENDIX

59

TABLE

II.

Experimmtal Results on water-ram caused by sudden or gradual

opening of valve B.

60

WATEK HAMMEE IN HYDRAULIC

PIPE LINES

BIBLIOGRAPHY
Water Hammee

/
^

Carpenter, E. C. Trans. Am. Soc. Mechamccd Engineers. Vol. 15. Church, I. P. Jowrnal of Franklin Institute. April and May
1890.

Gibson, A. H.

Boston Journal of Commerce. 1896. Trans. Ass. O.E. of Cornell University, 1898. Hydraulics. Constable & Co., London. 1908.
P. 213. Bulletin University of Wisconsin.

Harza,

L. F.

JoDKOWSKT, N.

^
^

No. 205. 1908. Pp. 152, 157. Journal of Imperial Academy of Sciences of St.
Petersburg.

1900.

Merriman. Hydraulics. Wiley & Sons, New York. 1903. P. 390. Unwin, W. C. Hydraulics. A. & C. Black, London. 1907. P. 196. Walckenaer, M. Ann. des Ponts et ChaussMs. Trimestre. 1899. 'Weston, E. B. Trans. Am. Soc. C.E. June 1885. P. 308.

Printed by T. and A. Cokstaele, Printers to His Majesty at the Edinburgh UnivBrsity Press