You are on page 1of 13

Charts for water hammer in high head pump discharge lines resulting from pump

failure and check valve closure

Department of' Civil Engineering,' British Coltrtnbin. Vut~co~cver,
B.C.. Corrorlo V6T I W5
Received April 25, 1984
Revised manuscript November 12, 1984

Maximum pressure head drops and rises resulting from pump failure and subsequent check valve closurc are calculated and
Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12

plotted for a simple pump discharge line at pump end, midpoint, and thrce-quarter point. Basic paramctcrs such as pipeline
constant, pipe wall friction, complete pump characteristics, and pump incrt~aconstant are accountcd for in the analyses.
Computer studies indicate that pipe friction, pipeline constant, and pump ~nertiahave a major effect on pressure hcad drops
and rises.
Studies indicate further that whereas for large pump inertia thc pressure hcad risc or drop at the midpoint is only moderately
larger than one-half of thc risc or drop at pump end, for small pump inertia this diffcrcnce is much greater. For very small
pump inertia, the pressure head drop or rise at midpoint approaches the values at pump end. This increase in pressure head
drop and rise for very small pump inertia is even morc pronounced at the three-quarter point.

Les hausses et les chutes maximales de pression rCsultant de I'arret subit de la pompe et de la fermeture subskquente du clapet
de retenue sont calculdes et mises en graphiques pour des points de mesurc situCs la sortie de la pompe, au milieu et aux
trois-quarts de la longueur d'une conduite de refoulemcnt alimentCe par une seule pompe. L'analysc tient compte des
parametres fondamentaux comme la constantc de la conduite, le frottement la paroi, I'ensemble dcs caracttristiques,
y-compris la constante d'inertie, de la pompe. Des modkles mathkmatiques montrent que Ic frotternent dans la conduite, la
constante de cette conduite et I'inertie de la pompe ont un effet prononck sur la valeur des hausses et des chutes de pression.
De plus, des Ctudes indiquent que dans le cas ou I'inertie de la pompe est ClevCe la valcur de la hausse ou de la chute de
For personal use only.

pression au point situk au milieu est seulement ldgiirement supCrieure 3. la moitiC de la valeur trouvCe a la sortie de la pornpe,
alors que pour une faible inertie de la pompe, cette valeur est beaucoup plus grande. Lorsque I'inertie de la pompe est trks
faible, la valeur de la hausse ou de la chute de pression au point situC au milieu se raproche de la valeur trouvde i la sortie
de la pompe. Cet accroissement de la valeur des hausses et des chutes de pression lorsque I'inertie de la pompe est trks faible
s'avkre nettement plus prononck au point de mesure situk aux trois-quarts de la longueur de la conduite.
[Traduit par la rcvue]
Can. 1. Civ. Eng. 12, 137-149 (1985)

Introduction influence of pumps on the transients from this instant

Substantial pressure head changes are likely to occur onward.
when pumps fail in a pumping installation that has a In the past, the graphical method of water hammer
long pipeline carrying water at high velocity. A sketch analysis was used to evaluate the transients in individu-
of a typical installation is shown in Fig. 1, where a al pump discharge lines (Parmakian 1963), as well as
pipeline of constant diameter and constant wall thick- for systematic studies (Kinno and Kennedy 1965). The
ness conveys water to a distant reservoir. A check valve graphical analysis is tedious for systematic studies and
is located near the pump, which draws water directly therefore only a limited number of cases could be anal-
from the open sump. ysed this way. The check valve closure results in an
When the power supply to the pump. motor is sud- irregular pattern in pressure head variation and there-
denly interrupted, the pump continues to rotate in the fore requires a large number of data points to describe
forward direction because of the kinetic energy of the the pressure head variation with sufficient accuracy on
pump motor, pump impeller, and the entrained water in the representing graphs. The availability of an electron-
the pump. Since this energy at modern pumping units is ic computer becomes, therefore, particularly useful
relatively small in comparison with the work required since it allows a large number of points to be calculated
for steady pumping, the speed of the pump drops rap- with a reasonable effort.
idly. Soon the discharge reverses if no check valve is In this paper a systematic study is presented for the
installed. In most installations, however, a check valve transient pressure head resulting from the power failure
is provided at the pump. This valve closes automati- to a low specific speed radial pump and the subsequent
cally and suddenly on flow reversal and eliminates the check valve closure. pressure head variations at the
138 CAN. J . CIV. ENG. VOL. 12. 198.5

Maximum upsurge
* pump e n d Maximum u p s u r g e $ p o i n t
(positive os shown 1
Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12

Minimum HGL

FIG. I . Sketch of installation.

three-quarter point have been added to that at the pump

end and the midpoint. This is done because of substan-
For personal use only.

tial nonlinearity found in results obtained for the three-

quarter point and for the midpoint of the pipeline. to be used in the equation of angular deceleration

Main parameters
in which w = unit weight of water, Qr = rated discharge
The main parameters of the pipeline are given below through one pump, WR' = moment of inertia of the
in nondimensional form. The pipeline constant rotating parts of one unit including water in the im-
peller, qr = pump efficiency at the rated condition,
Nr = rated pump speed, Acu = ( N I - N?)/N, is the
nondimensional speed reduction during a small time
in which a = water hammer wave velocity, V , = initial interval At, N, and N2 = transient pump speeds at the
steady state velocity in the pipe, g = acceleration of beginning and at the end of time interval A t , and
gravity, and Hr = rated pumping head. Hr = Ho HI + p, = average relative torque.
+ +H I H , , where H,, = static head, H I = head loss due The nondimensional time
to pipe wall friction and viscosity in the initial steady
state, H I = sum of local losses (check valve, gate valve,
pump header junction, etc.), and H , = velocity head in
the pipeline in the initial steady state. in which t = time in seconds and p = 2L/a is time
The pipe wall friction, local losses, and velocity head constant or characteristic time.
are lumped together and considered in the non- The pump constant KI and the time constant I*. can
dimensional form be conveniently combined into a single dimensionless
HI + HI + H, ratio
PI hl = Hr

in which friction H , = ~ ( L / D ) ( V : / ~ by
~ ) Darcy-
Weisbach, where f = friction factor, L = length of The complete characteristics of a pump (Donsky
pipe, D = pipe diameter, local loss H I = ~ ~ : / 2 g , 1961 ; Wylie and Streeter 1983) with a specific speed
where K = local loss coefficient, velocity head H, N , = 25 (S1 units) or 1275 (USgpm units) are used in
= ~ : / 2 g , h, = total relative head loss, and g = the zone of normal pump operation (see Fig. 2).
acceleration of gravity. T o adapt the pump characteristics for computer use
The pump constant the dimensionless ratios

are employed. In these ratios H, T, Q. and N are the

transient head, torque, discharge, and speed. The sub-
script r indicates these quantities at the rated condition.
With the dimensionless ratios for homologous pumps,
the expressions
Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12

x- = .rr + tan-' 1'

rr - 2 01 I
are obtained for the relative head W H and torque W B . lr $T $r
The data are available in tabular form (Wylie and
Streeter 1983) for equal intervals of x- = .rr/44 rad, and x = lr + tan-' v/a -+
are used to plot the characteristics shown in Fig. 2. FIG. 2. Pump characteristics, normal pump opcration.
Rclative hcad WH and relativc torque WB shown in tcrnis of
Basic assumptions ,r = T + tan ' v l a .
In the analyses, it is assumed that (a) the pipe di-
ameter and wall thickness are constant, (b) the check mum relative pressure head drop and are measured
valve is located near the pump, (c) the closure of check downward from the initial steady state pressure head at
valve occurs suddenly at the instant of flow reversal the pump. The hand-drawn curves shown in the figures
For personal use only.

through the pump and stays closed thereafter, (d) no air are in most cases only slightly smoothed computer
pockets exist in the high points of the pipe and no water plots.
column separation occurs during the transient condi- Using the plotted values of maximum pressure head
tion, (e) the velocity head and local losses are lumped drop at the pump end, midpoint, and three-quarter point
together with pipe wall friction, (f) the suction piping along with the value at the reservoir end, which has zero
is short and can be disregarded, and (g) the pump is pressure head drop, an enveloping curve can be drawn.
working at the rated condition in the initial steady state, This curve will show the maximum pressure head drop
i.e. at the point of best efficiency. along the entire pipeline. This curve. when plotted on
the profile of the pipeline, will enable the designer to
Calculations a n d results see in o n e glance, and with good accuracy, whether or
T w o partial differential equations describing the un- not water column separation is expected at any point
steady flow are transfornied~intofour total differential along the pipeline.
equations by the method of characteristics. A computer Secondly, the program calculates the maximum up-
program is written to solve these differential equations surge at the pump end, midpoint, and the three-quarter
with appropriate boundary conditions for a range of point due to check valve closure upon flow reversal
values of p, 7, and hl normally encountered in pumping through the pump. The results are shown in Figs. 6-8
systems, using the approach of specified time intervals respectively, which give the maximum pressure head
~, and Streeter 1983). In the calculations for the rise in terms of the pumping head H , at these locations.
maximum pressure head drop and rise, the pipe is di- The numbers on the curves indicate the maximum rela-
vided into 40 equal reaches of Ax. A time interval tive pressure head rise and are measured upward from
A t = Axla is used throughout the calculations. the initial steady state pressure head at the pump. The
The program calculates first the maxilnum down- maximum relative pressure head rise at the punip ap-
surge at the pump end, at midpoint, and at the three- proaches I .O for low 7 and high p values in the case of
quarter point near the reservoir. The results are shown frictionless flow.
in Figs. 3-5 respectively, which give the downsurge at In pipelines with large friction, the upsurge may
these locations in terms of the rated pumping head H, on never reach the initial steady state pressure head at the
the 7-p plane for various pipe friction values h,. Sepa- pump. In such cases, the steady state condition repre-
rate curves are drawn in these figures for hl = 0.0, 0. I , sents the ~naxiniumpressure head in the pipeline and
0.2, 0.3, 0.4, and 0.5. Intermediate values of pipe wall consequently a zero pressure head rise is indicated on
friction can be considered by interpolating between the the graphs for the pump end, and a negative pressure
curves. The numbers on the curves indicate the maxi- head rise is indicated for the midpoint and for the three-
Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12
For personal use only.

C A N . J . CIV. E N C . VOL. 12, 198.5

Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12
For personal use only.

Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12
For personal use only.

CAN. J . CIV.ENG. VOL. I?. 1085

Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12
For personal use only.

Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12
For personal use only.

CAN. I . CIV. ENG. VOL. 12. 1985

Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12
For personal use only.

Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12
For personal use only.

CAN. J. CIV. ENG. VOL. 12. 1985

Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12
For personal use only.

Relative velocity * v/vo

FIG. 10. Graphical analysis of downsurge

quarter point. The magnitude of pressure head rise the midpoint and at the three-quarter point for the range
reaches a maximum equal to one-half and three-quarters of T < 1.0. Noticeable irregularity occurs in the max-
of the friction at midpoint and at three-quarter point imum pressure head drop at the midpoint and at the
respectively, indicating the initial steady state three-quarter point in the range T > 1 .O. Here the differ-
condition. ence in head between the computer plot and the maxi-
The curves shown in the figures are hand-drawn en- mum enveloping curves shown on the graphs can go up
veloping curves based on computer plots. A sample of to 20%, the latter ones indicating the conservative val-
the computer plot with corresponding hand-drawn en- ues. Noticeable irregularity also occurs in the maximum
veloping curves is shown in Fig. 9. pressure head rise at the pump end over the entire T
The results were compared at pump end and at mid- range. The maximum pressure head rise at the midpoint
length with those reported by Donsky (1961) and Kinno and at the three-quarter point have even larger irregu-
and Kennedy (1965), obtained using the same pump larities, showing clearly defined ridges and valleys on
characteristics and the graphical method of water ham- the T-p plane (see computer plot, Fig. 9).
mer analysis. Good agreement was observed when The reader may be intrigued by the large irregularity
comparing the downsurge conditions. A large-scale in upsurge at the midpoint and at the three-quarter
analysis, using the data of the same pump and the point, which is shown in Fig. 9. This is caused prima-
graphical method, was carried out by the writers for rily by the check valve closure, which occurs suddenly
the upsurge condition. Again, good agreement was at velocity reversal. The irregularity in upsurge will be
obtained. largely reduced in pumping installations where the
check valve is omitted. On the other hand, these irreg-
Discussion of results ularities become much sharper and occur as localized
Smooth and gradual variation in the maximum pres- pressure head peaks, rather than ridges and valleys,
sure head drop is observed at the pump end on double- whenever water column separation (not covered in this
logarithmic plot in the studied range of variables T , p, paper) occurs in a high point of the pipeline.
and h,.. The same applies for the pressure head drop at The explanation for the irregularity in the plots of the
148 C A N . J . CIV. ENC. VOL. 12. I 9 X 5

pressure head rise at the midpoint and at the three- the downsurge condition.
quarter point is shown graphically in Fig. 10. The figure The maximum pressure head rise depends largely on
shows the graphical analysis for the transients caused the pipe friction. Linear interpolation may be used for
by the failure of a pump with specific speed N, = 25 friction values between the curves. Parabolic inter-
(1275) for a given value of T and for p = 0.75, l .O, and polation may be needed to obtain pressure head rise at
1.3 in a case where pipe friction is ignored. The case (b) a point between the three-quarter point and the reservoir
for p = 1.0 is shown in complete detail in the solid end of the pipeline.
lines, while for the cases (a) and (c), which are shown Figures 3-8 indicate that for large T and low p val-
with dashed lines, only the variation of the pressure ues, the pressure head rise (or drop) at the mid-length
head and velocity at the mid-length is indicated.' is only moderately larger than one-half of the rise (or
Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12

Figure 10 shows that the pressure head at the pump drop) at the pump end whereas for small T and large p
end (P) stays relatively constant for any of the three p values, this difference is much greater. For very small
values during the downsurge condition after the first T values the pressure head rise (and drop) approaches
2L/a s until the check valve closure. There is, how- that at the pump end. This difference in pressure head
ever, large variation in relative pressure head drop and rise at low and high T values is even more pronounced
in the remaining relative pressure head at the mid- at the three-quarter point. The observed nonlinearity
length ( M ) after the first 3L/2a s. Of primary impor- persists to higher T values as friction increases.
tance for the pressure head rise is the preceding pressure
head drop at the instant of zero velocity. For the case Conclusions
(a) this drop is at the minimum at instant 5L/2a s when ( I ) Friction substantially increases the downsurge
V = 0, while for case (c), the velocity reversal and condition.
valve closure occurs a short time after instant 7L/2a s, (2) Friction largely reduces upsurge. In pipelines
when the pressure head drop is nearly at its maximum with large friction, the upsurge may never reach the
value. The positive pressure wave generated by check initial steady state hydraulic grade line.
For personal use only.

valve closure will be superimposed on the value of head (3) In the range of low T values, substantial non-
existing at the midpoint. Thus, there will be a minimum linearity occurs between surges at pump end, midpoint,
pressure head rise for case (a) and a much larger max- and three-quarter point. At very low T values, the down-
imum pressure head rise for the case (c). surges at the pump end, at the midpoint, and at the
From the standpoint of design, it is important to three-quarter point are approxin~atelyequal. This non-
know not only the maximum pressure rise for given linearity decreases toward higher T values.
values of variables based on design data, but also the (4) Pressure head calculations for the three-quarter
maximum probable range of deviation of these vari- point should always be included for installations with
ables and the pressure head rises resulting from the T < -10.
various combinations of these deviations. For example, (5) It is important to consider the probable maximum
Fig. 9 shows that by increasing the value of p = 2.4 by and minimum values of variables T, p, and /I,- when
lo%, the pressure head rise is more than doubled for T analysing maximum surges. Single computer runs can-
equal to 0.25. Therefore a single prediction of pressure not be relied upon, especially at the midpoint and three-
head rise based on design data cannot be relied upon. A quarter point, since small changes in the data can result
knowledge of the relative location of the maximum in a large variation in the pressure head.
pressure head rise in the surrounding area on the T-p
plane is necessary for design purp~ses.Figures 6-8 Acknowledgements
provide such information in the form of enveloping
The authors are grateful to the Natural Sciences
curves. In addition to the pressure head ridges, the
and Engineering Research Council of Canada, who
deviations in the specific speed and design of the pump
supported the research.
must be considered. A of different specific speed
and different design (within the same specific speed)
will have characteristics different from those shown in DONSKY, B. 1961. Complete pump characteristics and the
Fig. 2. Kinno and Kennedy (1965) discuss this problem effects of specific speeds on hydraulic transients. Journal of
in detail for a pump with nearly the same specific speed Basic Engineering. 83, pp. 685-699.
and conclude that the results are applicable even for KINNO,H., and KENNEDY, J. F. 1965. Watcr hammer
substantial variation in specific speed, in particular for charts for centrifugal pump systems. ASCE Journal of the
Hydraulic Division. 91(HY3), pp. 247-270.
PARMAKIAN. J. 1963. Water hammer analysis. Dover Publi-
'For a thorough graphical solution the reader is referred to cations, Inc., New York, NY.
Parmakian (1963), Fig. 55 on p. 84, in which a nearly iden- WYLIE,E. B., and STREETER, V. L. 1983. Fluid transients.
tical case is analyzed in complete detail. FEB Press, Ann Arbor, MI.

List of symbols rated pump discharge (rn'/s)

water hammer wave velocity (rn/s) momentary torque on pump axle (N m) .
pipe diameter (m) rated torque (N .m)
friction factor time (s)
. ,
acceleration of gravity (m/s2) relative time, t' = t / p
momentary head (m) time interval (s)
static head (m) initial steady state velocity in pipe (ni/s)
rated pumping head (m) relative discharge
velocity head in initial steady state (m) B
relative torque, WB(x) = ---
relative head a' + 'v
Can. J. Civ. Eng. Downloaded from by University of Toronto on 10/04/12

relative pressure head change h

relative head, WH(s) = ---
H,- +-H~+ H, +
a' 'v
relative friction, h, = moment of inertia (kg. m')
local loss coefficient unit weight of water (kg/m.')
length of reach (m)
pump inertia constant, K, = relative speed
( 11s)
relative pump speed change during time interval
length of pipeline (m) At, Aa = KIP,At
momentary speed (rpm) relative torque
rated pump speed (rpm) average relative torque during time interval At
pump speed at the beginning of time interval rated pump efficiency
(rpm) time constant, p = 2L/a (s)
pump speed at the end of time interval (rpm) pipeline constant, p = aV,,/2gHr
specific speed (rpm) dimensionless pump inertia parameter, 7 = -
Kl P
For personal use only.

momentary pump discharge (m3/s)