INTRODUCTION will be observed : (1) waves with a short period of a few seconds,
Surge tanks are usually associated with conventional hydro which are elastic shock waves that can bring about elastic
electric power developments incorporating surface power houses deformations of the pipeline wall and of the water column itself,
as represented in Fig. 1. I n recent years many underground and which are known as water hammer; and (2) waves with a
power stations have been built and they, too, require surge tanks, longer periodabout 100 to 500 secondswhich are syn
sometimes giving rise to new and .difficult hydrodynamic chronous with the surges in the surge tank.
problems of their own (Figs. 1 and 2). Evidently, the whole hydraulic system has to be investigated
Considerable changes in the demand for energy may suddenly in regard to both types of pressure wave.
occur in the electrical distribution lines and whole power stations The upstream pressure tunnel is not usually able to with
may suddenly be cut off the grid or thrown on to it. At the stand high pressures, and the surge tank shown in Fig. 1 will
present time no method has been found which would be capable be the best protection against water hammer penetrating the
of dealing with these sudden load changes on the electrical grid tunnel. It is generally built at the downstream end of the pressure
itself, in which there are inherent problems of stability. tunnel, upstream of the steeper steel pipeline or penstock which
Ultimately, the full impact of these variations of energy demand transmits the full available head. Protection given by the surge
is thrown upon the hydraulic system, where disturbances are tank against negative pressure drops may be as important as that
bound to occur resulting in variations of pressure and discharge. against positive pressure rises.
It is possible by measuring the pressure variations at the I n addition, the surge tank makes possible a reduction in the
lower end of the pipeline to follow the effects of unsteady load length and thickness of the costly highpressure steel penstock,
on the hydraulic system. Two different types of pressure wave and also permits stable regulation of the turbine, which may
become impossible with a longer pipeline.
The MS. of this paper was originally received at the Institution on When rejecting or taking on load, the surge tank will either
8th December 1952, and in its revised form, as accepted by the
Council for publication, on 10th February 1953. For the report of the accumulate water or, on the contrary, will deliver previously
meeting in London, on 6th November 1953, at which this paper was accumulated water, thus working as a compensating reservoir.
presented, see p. 124. The level variations in the surge tank have the effect of accelerat
* Consulting Engineer, English Electric Co., Ltd., Rugby. ing or decelerating the flow of the water masses in the pressure
8
92 PRESENT TRENDS I N SURGE TANK DESIGN
STATIC LEVEL
STATIC LEVEL
giving z as a function of t. Z, and T are constants, namely: which is the socalled condition of Thoma (1910) for stability
L f L F of small oscillations in cylindrical surge tanks. Usually a so
2, = uoJ( .
g F
)and T = 27rJ( 
g’f
).
For general treatment called safety factor n with F = nFTh and n> l is to be introduced.
For years the discussion went on to decide the right value to be
of the problem, arithmetical integration or graphical stepby chosen for n. This condition of Thoma is of the greatest im
step methods will have to be used. A detailed exposition of all portance for designers and, in some instances, considerably adds
these methods appears in Jaeger (1949a), and a summarized to the difficulties in designing a proper surge tank. Furthermore,
discussion in Rouse (1951). investigation is necessary as to how this basic formula needs to
be modified so as to deal with large oscillations in surge tanks
STABILITY OF SURGES
of different typesthrottled or differential tanks, systems of
surge tanks. In the more general cases the curve for z as a
The whole theory of surges has hitherto been developed on function of the time t (z = z(t))can be calculated step by step,
the assumption that there is a definite variation of the discharge including the condition of equation (4) as a boundary condition
from Q* to Qo = Q*+dQ, regardless of turbine governing for the arithmetical or graphical integration. The decrement
regimen. It was found that some surge tanks developed unstable d = X i + l / X i (Fig. 1) will then be established (Jaeger 1949a)
oscillations, the causes of which were actually traced to the which is far more important than the value of n. The condition
turbine governor. of Thoma is equivalent to d < 1.
1
ZMIN
I
f' Y€J
Fig. 4. Surge Tank with Variable Section F
Fig. 6. Differential Surge Tank
The surge tank with chambers of variable section (Fig. 4)
consists of a surge shaft, the crosssectional area of which is An analysis of recent projects and designs shows that com
chosen to be as small as possible consistent with stability bined systems of surge chambers with two or more shafts, having
( F > F T h ) . This shaft is provided with an upper expansion restricted orifices, expansion chambers, overflows, and reservoir
chamber and a lower reservoir chamber. As the load on the grid chambers, have been proposed. The aim is to reduce the overall
varies, the water level in the shaft will rise sharply or fall steeply, volume of excavation and to make the design of the chamber
but the extreme surges are limited by the chambers. Thus both suitable for the constructional methods used for steellined
conditions of sharp variation of z at the beginning of the surge pressure shafts which sometimes replace the conventional pipe
and of limited extreme surges are satisfied. In some instances lines. Figs. 8, 9, 10, and 11 represent some recent examples.
the upper expansion chamber is provided with a spillway.
The surge tank with restricted orifice (or throttled surge tank)
yl
STATIC
LEVEL
Fig. 5. Surge Tank with Restricted Orifice Fig. 7. Differential Chamber for Upper Part of Surge Tank
I1
II ,r   i l
' v HIGHEST STATIC W.L. 477
ENLARGED DETAIL
OF PENSTOCK
ORIFICE
SECTION AA
0 10 20 30 40 50 METRES
I
Fig. 11. Combination of Three
Surge Tanks with Restricted
Orifice: Danger of Unstable
Oscillations
STATIC LEVEL
.    
S
TURSINE
fl b
C d e
Fig. 12. The Five Principal Types of Underground Power Station
PRESENT TRENDS I N SURGE TANK DESIGN 97
L O A D I N G CONDITIONS T O BE A D O P T E D over a long period of time. The present endeavour is to disclose
Wise choice of the type of surge tank is only part of the the most recent tendencies in design of surge tanks as the result
designer’s work. It is most probable that definition of the con of this experience gained in the operation of large grids.
ditions under which the turbine is loaded will require con There is no doubt that any surge tank might have to face the
siderable thought and discussion. sudden closing of all the turbines from full gate if a short
In recent publications the fact has been stressed that loading circuit occurred while the power station was running at full load.
conditions are becoming increasingly severe. At the same time occur It is estimated by Electricitk de France that this position will
there is a natural tendency to limit the margin of safety to the surge approximately
tanks in
once a year. T o the author’s knowledge, all
important power stations satisfy this condition.
lowest pemksible value, and to reduce the cost of every item, Alternatively, provision is made
and especially the surge tank as being a feature of considerable above a predetermined level. for spilling when the water rises
magnitude though not actually productive of any energy. Until recently it was accepted that the conditions for taking
A typical example of what may take place owing to changing on load were in the hands of the supervisor of the station. It was
conditions on the electrical network is the accident which considered possible to foresee the magnitude of the maximum
occurred at the Albula power station (Jaeger 1948). This plant load to be carried, owing to demand by big furnaces, large mills,
was built by the town of Zurich between 1907 and 1910, and heavy trains, and so forth.
was equipped with eight Francis turbines of 3,000 h.p. each. Most of the specialists today consider that this reasoning does
The surge tank was correctly designed for rejection of the full not cover the more dangerous cases. When several stations are
load, but it was assumed that the load would be taken on pro connected to a large electrical network, it is always possible,
gressively by opening the eight turbines in succession, one after because of accidents or a short circuit, that one particular station
the other. The load on the network of the town of Zurich was may be isolated automaticallyor may have to be isolated
then developing rapidly and, in 1926, the Waggital power station suddenly to prevent the trouble extending to other parts of the
was built with four generators of 16,500 kVA. at Rempen, and electrical networkand that it may, therefore, have to take on
four similar units of 16,500 kVA. at Siebnen. The Albula station load suddenly. So long as a turbine is not completely shut down
had to meet unexpected load variations when the other more but is only in the act of closing or running idle, it can be re
powerful station was taking load on or off. The Albula surge opened suddenly to full gate. The loading conditions for load
tank was unable to deal with the new conditions and, finally, the increase will, therefore, have to be examined in regard to the
pipeline burst owing to unfortunate circumstances of load possible connexion and disconnexion of entire power stations
shedding in another part of the grid. belonging to the same general network.
There are many other causes of dangerous surges. A station
A I may be losing load, with the turbines closing, and the water
beginning to surge in the tank. If load is put on again, inter
ference of oscillations will occur, and the danger of increased
surges, through superposition of oscillations, is greatest if the
switchingon of the load occurs when the water in the surge tank
is again level with the water in the reservoir.
The reverse conditions may also occur : the, machines may
shut down suddenly soon after taking on load. There is no
means of preventing such an occurrence. The downward surges
may be even more dangerous than the upward ones.
This superposition of oscillations may happen at any level in
the reservoir, and both the positive and negative maximum
surges have to be investigated.
Superposition of openingclosing and closingreopening
movements of the turbine gates were investigated on the model
Fig. 13. Distribution of Load on the Three Power Stations of the Innertkirchen surge tank which was finally designed to
A, Byand C deal with any possible combination of surges.
The Loch Sloy tank (maximum reservoir level 935 feet) has
A Albula. been designed on the assumption that four turbines being on full
B Several low head plants. load may close, and two turbines may reopen to full load when
C Waggital.
the water in the reservoir is at a level of 830 feet or higher. If the
reservoir level falls below this limit, then no more than two
Fig. 13, which reproduces the actual load variation which turbines are allowed to be run at the same time. While the water
provoked the accident, shows that Albula (A) had to take over level in the reservoir is low, water must be saved, and it would
rapidly from Waggital (C) at about midnight; at 5.30 a.m. the therefore be questionable to run the station a t full load. But
load on Albula was thrown off abruptly, causing dangerous provision was made to care for the reopening of two turbines at
positive and negative surges in the Albula surge tank.The surge any time after rejection of the load by two turbines, when the
tank had to be designed anew and enlarged so as to be able to water in the reservoir was down to the lowest level (800 feet).
stand up to the far more severe loading conditions of the expand It must be emphasized that in many hydraulic systems the
ing grid. surge oscillations may dampen out very slowly (Jaeger 1949a).
At the time when this accident occurred, the Innertkirchen This is particularly so if the pressure tunnel is short, its diameter
power station of the Oberhasliwerke was being designed by large, and its walls smooth. Surges may continue oscillating for
Dr. A. Kaech. The lesson learnt at Albula, through hard hours.
experience, was not neglected, and steps were taken to ensure If the period of the surge is comparable with the time required
that the Innertkirchen surge tank would be able to deal with to synchronize a turbine, the final surge, after all the turbines
CXIXP~~OM~ cases of superposition of surges. Furthermore, the have been put into operation one after the other, may not be very
damping of the surge oscillations is extremely efficient, and much less than if all the turbines had opened suddenly together.
suppresses those surge oscillations sustained for an undesirably This persistence of the surges over a long period of time has
long time which are characteristic of many other surge tanks. to be taken into account, especially when power stations are
In a recent paper, Scimemi and Ghetti (1951), analysing equipped with one or two turbines only.
Italian experience, arrive at very similar conclusions. Loading It has been shown that there are a number of instances where
conditions prescribed by the experienced customer are also surges produced by successive load changes can interfere and
increasingly severe, especially when the customer has had a wide accumulate dangerously. Examples from actual practice have
experience of operating a large electrical grid, For definition of been mentioned, and a variety of influencing factorsreservoir
loading conditions, the design engineer relies on the experience levels, number of turbines, length of pressure tunnel, friction
and opinion of engineers who have been operating large grids losses, etc.has been enumerated.
98 PRESENT TRENDS I N SURGE TANK DESIGN
The possibility, pointed out by Scimemi and Ghetti (1951), of Effectof Large Oscillations. Some efforts were made to extend
selecting different ‘statism’* for the different power stations, the theory of Thoma to the case of large oscillations. Let the
thus putting all the weight of this load on one or two specially small oscillations be written in the form of equation (3, which
selected control stations, is of little help in some of the instances can be multiplied by (dx/dt)dtand integrated.
discussed in this paper. There is always the possibility of some The lefthand number of this equation then represents the
stations being isolated accidentally or having to be isolated so as s u m of the kinetic energy and of the potential energy of the
to localize a disturbance and avoid its spreading to other parts of oscillations between the time limits t = 0 and t. If u and b are
the grid. Any large power station, even if not designed specifically positive, then the oscillations are stable and the s u m of the two
to be a control station, may suddenly have to face the load varia energies is always negative.
tions of a whole grid. It can be shown that this is true also for large oscillations and
This survey would be incomplete if mention were not made if the area of Thoma, FTh, is increased by the factor
of another tendency. The French electricity network is charac
terized by its wide extent and numerous interconnexions. The
big industrial centresParis, the north of France, and the east
are fed directly by large steam generating stations which are large oscillations will also be stable. The formula agrees with
located in the vicinity, and the total output of which covers the results of detailed calculations point by point.
about half the yearly demand of the French grid. Measurements carried out by Ghetti (1951) on existing
The large hydroelectric power stations, widely linked with surge tanks indicate that the damping of large oscillations is in
each other and with the thermal stations, feed long transmission some instances more efficient than is indicated by this formula.
lines leading to Paris and the industrial centres. The load The obvious explanation is that large oscillations introduce
variations on the hydroelectric power system have rarely, in additional friction losses when the water is diverted from the
recent years, put a dangerous strain on any individual station. pressure tunnel into the surge tank, which is most frequently at
It has, therefore, been decided by Electricit6 de France that surge right angles to the former. These additional friction losses
tanks shall, in future, be designed for throwing off N 1 load, the depend on the ratio of vF/wf. Treatment of the problem is
negative surges thus produced being analysed as well as the difficult for cases other than that of total closing, when wf = vF,
preceding positive surges. It will be assumed that ahy negative which is of little interest for stability studies. In the more general
surge produced by taking on load will have to be limited to a case, the treatment must follow the lines of the theory developed
reasonable amount consistent with the case of ‘fullload of€‘, thus by Escande (1952) for the surge tank with restricted orifice.
avoiding any considerable additional expense. The problem is, According to circumstances, the damping effect of losses due to
therefore, reversed, the need for greatest economy being put the additional bend may or may not prevail against the additional
first and the safety of any new individual station being based on instability shown by the formula.
the extensive ramifications of a very large network. The rules
of Electricit6 de France leave the matter open, however, by Surge Tank with Restricted Orifice. Additional losses occur
specifying that any new station shall be assessed on its own (Escande 1952) when the water passes through the restricted
merits, while taking all the circumstances into account. Very orifice, in both the upward and the downward directions, so that
important stations will still be designed on assumptions similar the stability conditions in a throttled surge tank cannot be worse
to those advocated by the Italian specialists. than in a simple tank. They must be better. Let it be assumed,
Readers maf very well reach the same conclusions, namely, therefore, that the tank is slightly unstable and that the condition
that any new scheme shall be considered in relation to its of Thoma for small oscillations is not satisfied. In relative values
circumstances, and that the appropriate loading shall be decided the condition of Thomaequation (6)is equivalent to
only after careful study of all aspects of the regulation problem.
2pQhQ>l . . . . . . (7)
where po = Pw2lZ, and = Ho/Z,(Ho being the net head).
SPECIAL P R O B L E M S O F S U R G E S T A B I L I T Y The conclusion of a study by Escande (1952) is that when
2p,,hO<1 (the condition of Thoma not being observed) there is no
The question of surge stability has already been dealt with final damping of the oscillations, but a state of equilibrium is
summarily on the lines of the theory of Thoma and it has been created with stable oscillations of limited constant but by no
pointed out that several restrictive assumptions have been made means negligible amplitude x, which depends on the three
explicitly or implicitly in the mathematical treatment by Thoma factors ho, po, and ro, where ro = relative value of the throttle
(1910). losses.
The surges are supposed to be very small, all terms superior The governor must be adapted to deal with this continuous
to the second order having been neglected; furthermore, the variation of the head without creating unsteadiness in the
governor is assumed to maintain the output N rigidly at a electrical distribution system. Furthermore, there is an increased
constant value. The station efficiency is assumed to be constant. danger of surges becoming superimposed both in the positive
Finally, the station is supposed to be running alone on the grid and negative directions on any occasion of load change. There is
with no possible interference from other stations. In practice, a strong argument in support of observing the condition of
it may be that none of these assumptions will be correct, and Thoma for a throttled surge tank also, because the greater
further investigations may become necessary to ascertain how oscillations x may prove to be objectionable.
surge tanks behave under conditions where this is so, and to
establish why they do sometimes behave differently from the
manner forecast by the theory of Thoma. The whole problem Stability of Differential Surge Tanks. Let rOvZbe the relative
requires more comprehensive treatment. value of the orifice losses for full discharge. The surge in a
A much disputed point is the socalled factor of safety, differential surge tank will certainly depend upon F , , F2, ro,
n = F/FTh, to be introduced into the condition of Thoma and p.
(equation (6)). If the area of the surge tank were equal to the No theory of the stability conditions for differential surge
area of Thoma, then small oscillations would last indefinitely, tanks has yet been established, but it can be shown that the
and a certain factor of safety n must be introduced so that, if theory of the ordinary throttled tank provides the key to part of
the theory for differential surge tanks (Jaeger 1952).
F = nFTh(With n> 1) In some instances, where the orifice at the base of the inner
the surges decrease, the primary surge being followed by others riser is large (ro small), the differential surge tank behaves as an
of decreasing magnitude. ordinary throttled surge tank. There is no spilling, or only a
It has been pointed out already that the decrement A = xi+,./x; negligible quantity, over the inner riser (Fig. 6), the maximum
can be calculated or measured, and that it yields informauon surge being identical with the surge in a throttled surge tank. In
which is more valuable than any discussion of the factor n. such a case, the condition of Thoma refers to the area of the
outer reservoir, F1,and when this condition is not satisfied for
* The curve of ‘statism’ gives the steady speed wo as a function of small oscillations, a state of equilibrium may be reached with
the load N. large oscillations occurring similarly to those in throttled surge
PRESENT TRENDS I N SURGE TANK DESIGN 99
tanks with too small an area ( F < F T h ) . The conditions for the constant. It is possible to show that the varying efficiency of the
downward surge should be identical with those for the upward turbine has a powerful stabilizing effect when dr]/dN>O, but
surge, from which the minimum height of the tank may be adds to the instability on those regions of the curve 7 = $N),
calculated. Instances may occur (when 2p,,ho< 1) where this total where dq/dN<O.
height becomes greater than the first surges for closing and for A long pipeline has the same effect as a negative dT/dN value
opening added together. and, therefore, increases the danger of instability. Both effects
When the area of the restricted orifice is small and the throttling can be examined when analysing a stability curve by a stepby
of the water passing into the outer chamber is effective (togreat), step calculation.
the water will reach the top of the inner riser and will spill over
during the first period of the rising surge. The limit of the upward Stability of Systems of Surge Tanks. A system in which there
surge is then virtually equal to the height of the inner riser. A is more than one surge shaft on a pressure tunnel, or in which
similar limit does not, however, exist for the downward surge. several pressure tunnels run into the same main shaft, will be
The diameter, D2, of the inner riser can no longer be neglected referred to as a 'system of surge tanks'. With the everincreasing
when the damping of the oscillations is being considered, and capacity of modem power stations, systems of surge tanks are
there is an obvious danger of unstable oscillations in the down being used more frequently. The case should also be noted
ward direction. where additional water is introduced into the hydraulic system
Full stability will certainly be obtained when the area F2 of by way of adits on shafts located between the main intake and
the inner riser satisfies the condition of Thoma. Differential the surge shaft or tank.
surge tanks connected to highhead plants are likely to fulfilthis In some instances, the main surge tank has been split into
condition and to be stable. Others, especially those connected two or more shafts for construction reasons. When modernizing
to lowhead plants, may be unstable. and enlarging an existing power station, the designer may have
Differential surge tanks with small orifices, when connected to develop a system of surge tanks by adding new shafts to the
to lowhead power stations, must be specially checked in regard existing one. A new design which has sometimes been adopted in
to their stability. This can be done by a stepbystep calculation, recent years consists of a reaction turbine with pressure tunnels
or, alternatively, with a model to scale. on the upstream and the downstream side, each with its own
surge tank.
Effect of the Velocity Head :Ventun' Eflect. When the surge The following rules summarize roughly the main results
tank is on the upstream side of the turbine, the theory shows (obtained from the general theorem of Hurwitz on stable
(Calame and Gaden 1926 and 1927; Jaeger 1949a) that the oscillations) when two surge tanks are connected to the same
velocity head w02/2g can be added to the friction losses, and the pressure tunnel on the upstream side of the turbine :
If the surge shaft nearer the turbines satisfies the condition
quantity Pwo2 = WOZ
5 +pw02 = ( & + ~ ) w o 2 will appear in the of Thoma ( F 2 > F T h ) the system will be stable, whatever the
denominator in theformula of ThTma &stead of Pwoz. Further dimensions of the second surge tank.
more, this wo2/2g is not lost for the net head, which should If the nearer surge tank is smaller than the area indicated by
include this amount. Thoma, but larger than half this area ( + F T h < F Z < F T h ) then the
When the head is low and the tunnel short, this theoretical s u m of the areas of the two surge tanks (Fl+Fz) will have to
effect of w02/2g may be very considerable, and it has been be larger than F T h . If the second surge tank is close to the nearer
proposed to increase the damping by introducing a Venturi one, then a small margin of the total areas above F T h may be
effect in the form of converging cone and an expanding taper sufficient to secure stability of the system. If the second surge
pipe at the bottom of the surge tank, as illustrated in Fig. 14, tank is nearer the tunnel intake, then its area will have to be very
where this Venturi is shown in association with a throttled surge considerable and the proper solution will be adherence to the
tank. rule of Thoma for the lower surge tank.
The rules giving the minimum permissible area of F I and Fz
1 STAT$ LEVEL ~
as a function of their location are complicated.
When the turbine is a reaction wheel with surge tanks on both
c
I I
Py2
the upstream and the downstream sides, it does not seem possible
to apply the general rule of Hurwitz. Studies yield a somewhat
Merent theory with the result that, owing to mutual interference
of the oscillations in the upstream and the downstream tanks,
the limiting value given by Thoma has to be greatly exceeded
(Jaeger 1949a). These results have been checked by direct
calculations.
When examining modem designs for systems of surge tanks,
it will be found that the simple shafts on which the theorizing
is based are actually replaced by throttled or differential surge
shafts. In some other instances, the shaft is provided with a spill
way. Instability can also occur with arrangements similar to
Fig. 14. Surge Tank with Restricted Orifice and Venturi Fig. 15 where secondary oscillations between the two surge
Effect
r
~
Fadalto :
(Oscillations at the limit of
15th March 1947
25th May 1947
65 1 0.49 H = 110 to 90 metres
L = 2,515 metres
stability) 31.7 I D = 5.8 metres
Crosssectional area of tunnel
f = 317 square metres
Losses P w 2 = 0.66~12
Unloading from 28,000 to 18,000
kW.
Partidor :
(First series of testsdamped
oscillations lasting 2+ hours)
14th April 1947 39.6 60 1
I
0.66
H = 28.6 to 19 metres
L = 914 metres
D = 3.05 metres
(Second series of testsstable 27th to 30th April 1950 Unloading from 500 to 350 kW.
and unstable oscillationswere
produced)
~
Caerano :
(Oscillations dangerously un
stable)
3rd June 1947
I H = 17.0 to 16 metres
L = 1,784 metres
D = 2.80 metres
f = 6.16 square metres
Losses Pw2 = 0.38~2
I Constant load of 690 kW.
Constant load of 720 kW.
Load variations between 900 and
1,070 kW.
I
Heimbach : 1904 H = 70 to 110 metres
(Unsteady oscillations) Qo= 20 cu. metres per sec.
L= 2,700 metres
4.9 f = 6.5 square metres
I
PRESENT TRENDS I N SURGE TANK DESIGN 101
commented on by Ghetti (1951). Considerable hydraulic surges problem of surge stability defined by A and the assumptions
(about 20 per cent of the head H ) were observed both of the concerning the dangerous loading conditions (discussed above
damped and of the undamped types, and they were related to under Loading Conditions to be Adopted), which also depend
frequency variations. I t was possible to have stable and unstable on A . Finally, the correlations between the decrement d of the
frequencies for this same hydraulic system, and for the same surge oscillations and. the frequency variations on the electrical
opening of the turbine (therefore for identical hydraulic con network should be studied closely.
ditions) simply by changing the sensitiveness of the governor.
The conclusions of Ghetti are that the favourable results of the
WATER HAMMER A N D SURGE T A N K D E S I G N
first series of tests were due partly to the type of governors used
in the stations, and that changes in the governor could modify Hitherto it has been assumed that the surge tank reflects
the results of the tests. Evangelisti (1950), when scrutinizing downwards all the pressure waves travelling from the turbine
the first series of tests by Scimemi, came to similar conclusions. along the pipeline in the upstream direction. Water hammer has
He developed a more detailed theory of surge stability assuming already been considered for its reactions on turbine governing.
that the governor also reacts on the surges within certain limits. It will now have to be considered in regard to pressure distribu
This new, rather involved, theory finally yields an equation of tion along the whole hydraulic system, comprising pipeline (or
the type a system of pipelines or penstocks), surge tank, and pressure
+
Cor2+ C l r C2 = 0 tunnel.
It is possible to generalize the water hammer theory of Allievi
where C1 contains terms depending on the governor (r = root (1913) for the case where a surge tank is designed at the upstream
of the accompanying characteristic equation). or at the downstream end of the pipeline. It can be shown
Evangelisti and Ghetti studied the Heimbach, the Pelos, and (Jaeger 1933) that, for a single pressure pipeline, the water
the Caerano power stations on the basis of the full value to be hammer depends on two relative values p = uvo/(2gHo)and
given for C1. It was found that in all three instances the two 0 = 7 / ( 2 L / a )only (where vo = velocity in the pipe and T = clos
first terms of C1 were negative, but in the case of Pelos the ing time of the gates). The hydraulic system, consisting of
corrective terms concerning the governor were positive and pressure tumel, surge tank, and pressure pipeline, can be
would quite likely overbalance the negative terms of C1. This represented by a system of three pipes (1, 2, and 3) intercon
was not so for the two other power stations. The governor of nected at point A (Fig. 16).
Partidor had also to be modified by Ghetti. Inspection of the
curves published by the Italian specialists (Evangelisti, Ghetti,
and Scimemi) proves that a limited amount of surge stability STATIC LEVEL
can be achieved through the action of the governors. __


_ 2
For the time being, only provisional conclusions are possible
on this difficult matter of surge tank stability.
The Italian specialists (Evangelisti, Ghetti, and Scimemi)
essayed to prove by means of mathematical analysis and
field tests the importance of proper turbine governing for
keeping down the surge oscillations. Despite the limitations
theirs is an indisputable achievement. Notwithstanding those
efforts, however, some very important points have not been
cleared up. For further discussion of the Italian results, the
damping effect of throttling at the base of the surge tankand
the mere changes of direction of the flow are equivalent to a
partial throttlingshould also be considered, as it has been
made clear that this provides an additional damping effect. A Fig. 16. Water Hammer in a Pressure Pipe System
second comment questions the finality of the Italian tests. Including a Surge Tank
Several authors have insisted on the stabilizing effect of the 7
value (7 = efficiency factor of the turbine and pressure pipe) in UlVOl a2woa wo3.
P 1 = ; p2= ; P3 =
the domain where dr)/dN>O. Although this point is not men %YO 2gYO 2gYO
tioned in the discussion of the Italian tests, it is presumed that
most of them were carried out for values of dq/dN>O, when it Any wave travelling upwards in the pipe 3 (pipeline) when
was to be expected that the damping effect therefrom would be arriving at point A will be divided. A part of the wave travels up
considerable. pipe 2 (surge tank), another part penetrates into pipe 1 (pressure
It must be stressed that all these factorsadditional surge tunnel), and a third is reflected towards the pipeline in the
stability from governor action, additional throttling losses, downward direction.
efficiency variationmay add their stabilizing effects as they Starting from these fundamental statements, it is possible to
probably did for the tests in Italy, but in certain circumstances calculate the pressure distribution in the whole system. The
they may vanish, or even work in the opposite direction (if problem can be solved by the graphical method of Schnyder
dq /dN< 0). Bergeron (Bergeron 1935), but the analytical method shows
It is of great value to know that there is a certain amount of some superiority in giving more general and quicker results
stability margin in the calculations, but, for the time being, no (Jaeger, 1933, 1949a).
precise figures can be assessed. The practical conclusions to be drawn from detailed calcula
In spite of being disputed, this simple formula of Thoma tions is that in most instances the surge tank conveniently
remains a most important criterion for designers.When allowance protects the pressure tunnel against positive and negative pres
is made for all latterday results, confidence in this formula is sure waves, the percentage of pressure penetrating into the
not greatly impaired. It does not matter so much to know whether tunnel not being dangerous. There are few exceptions. Some
the mathematical instability of the surges begins for an F value will occur when the surge tank is a long inclined pipe with a
slightly smaller than or just equal to FTh. It is known definitely diameter comparable to the penstock or tunnel diameter. I n
that for F = FTh the surge amplitude may reach disquietening such instances an appreciable amount of pressure will penetrate
values. The designer should be more interested in the decrement the tunnel (or pressure pipe). Negative pressure waves will be
d = xj+l/x; of these surges (Jaeger 1949a). A theoretical formula more difficult to deal with and may require that the pressure
which is corroborated by many direct stepbystep calculations tunnel and the connexion to the surge tank should be moved to a
(graphical or analytical), as well as by site measurements, shows lower level.
that d depends primarily on the tunnel characteristicslength, It is essential to check the dynamic pressure drop at the valve
crosssectional area, roughness of the walls, and total friction house, downstream of the surge tank. It is a common error of
losses. For surge tank design, the right estimate of d is of great design to locate the valve house at too high a level, where surges
importance. Furthermore, there is definitely a link between the and water hammer combined may produce pressures below
102 PRESENT TRENDS I N SURGE TANK DESIGN
atmospheric. Relocation of the valve house may entail an in ‘partial working’ surge chamber. Experience gained at
crease in the gradient of the penstock between surge tankand Wettingen power station, Switzerland, has proved such a
valve house. system to be stable.
The problem of waterhammer wave reflection requires
special care when the surge tank is of the restrictedorifice F I N A L REMARKS
(throttled) type. A restricted orifice, like many diaphragms, A fascinating aspect of the design of hydroelectric power
produces concentrated localized losses, to some extent hiding plant is the complexity and the multiple interactions of the
the wave transmission. Formulae for restrictedorifice surge components involved. The surge tank problem is quite typical in
tanks were published by Jaeger (1933). A recent Italian paper this respect. At the starting point of this analysis are found some
(Benini 1950) arrives at the same formulae, which compared simple physical principles and very simple differential equations.
well with the results of model tests. As the pressure rise below The methods of integration are easy but, suddenly, problems
the restricted orifice is very rapid, it is also being transmitted arise. It is discovered that the surge tank pulsates; it reacts to
without reduction upstream, and penetrating deeply into the the turbine governor; and it reacts to the water hammer.
pressure tunnel. It is necessary to insist on the complexity of the reactions of the
With the increasing popularity of throttled surge tanks these surge tank. With the growing dimensions of power stations,
aspects merit full attention. incorporating more and more powerful generating sets with even
It would be a dangerous illusion to think that a surge tank greater discharges, the design ofsurge tanks becomes still more
will stop any type of pressure wave. It is an established fact that complicated. Systems of surge tanks replace the single surge
pressure tunnels have been severely damaged by water hammer tank. If the station is of the underground type instead of the
in spite of ample protection by surge tanks. One type of wave conventional surface power house, the problems to be solved
may pass a surge tank and penetrate into the tunnel (Jaeger may be the more difficult because of conditions peculiar to such
1948, 1949a): there are the upper harmonics, generated by underground stations.
resonance in the pressure system. Resonance of this type m a y
be promoted by a governor, but resonance may also occur with ACKNOWLEDGEMENTS
no fault from the governor whatsoever.
The author’s thanks are due to the Editor of The Engineer
for permission to reproduce Fig. 8, also to the Editor of Water
S U RG E T A N K S C O N N E C T E D T O FEEDER C ANA L S Powm for permission to reproduce Figs. 10 and 11.
Before 1910 many power stations were designed without
proper surge tanks, the pressure tunnel being replaced by an
open canal or a freeflow tunnel. This type of layout has never
completely disappeared, and very recent designs show the same APPENDIX
features. In some instances, the canal is directly connected to
B I B L I O G R A P H Y AND REFERENCES
the turbine intake (Fig. 17u), while in others the canal discharges
ALLIEVI,L. 1913 Atti del Collegio degli Ingegneri ed Archi
Ah tetti, Milan, nos. 15, ‘Tecnici del colpo d’ariete’.
BENINI, G. 1950 Energiu Elettricu, vol. 27, nos. 11 and 12,
‘Sui fenomeni di colpo d’ariete nelle gallerie munite di
pozzo piezometrico con luce strezzata’.
BBRGERON, L. 1935 Revue gtnkrale de l’Hydraulique, nos. 1
and 2, ‘Etude des variations de rtgime dans les conduites
d’eau’.
BINNIE, A. M. 1943 Trans. A.S.M.E., vol. 10, p. A.183,
‘Oscillations in Closed Surge Tanks’.
1945a Proc. I.Mech.E., vol. 153, p. 15, ‘Protective Air
Vessels for Rising Pipe Lines’.
1945b Proc. Cambridge Phil. SOC.,vol. 42, part 2, p. 156,
‘Approximate Methods in Surge Tank Calculations’.
BLAIR, J. S. 1945 Proc. I.Mech.E., vol. 153, p. 1, ‘Controlling
Pipe Line Surges by Means of Air Vessels’.
I
CALM, J., and GADEN,D. 1926 ‘Thbrie des Chambres
d’tquilibre’ (GaudierVillars, Paris).
b 1927 Schweizerische Buuzeitung, July, ‘De la stabilit.6
des installations hvdrauliaues munies de chambres
d’equilibre’.
Fig. 17. Feeding of Power Stations by Head Race Canals : CUBNOD,M., and GARDEL,A. 1952 Bulletin Technique de la
Formation of Translating Waves Suisse Romande. vol. 78. D. 93. ‘Etude des Ondes de
Translation de Faible &phude dans le cas des canam
into a surge tank (Fig. 17b). Turbine regulation will produce d’amenee des usines hydrotlectriques’.
translatory waves in the canal, and in some instances the regula DENNIS,N. G. 1953 Proc. 1.Mech.E. (B), vol. lB, p. 379,
tion may have a character of instability which compares to some ‘Water Turbine Governors’.
extent with the instability of surges (Cuenod and Gardel 1952). ESCANDE, L. 1952 Acadtmie des SciencesComptes Rendus,
This instability may become particularly disturbing if a vol. 234, p. 299, ‘Etude de la stabilite des chambres
number of power stations of similar design are working in series d’equilibre et 6tranglement’. (English translation in
along the same main canal, and even more so, if the canal is used Water Power 1952 vol. 4, p. 221.)
for continuous irrigation. It will then be found that special EVANGELISTI, G. 1947a ‘La regolazione delle turbine
attention has to be given to some alternative means of regulating idrauliche’ (Zanichelli, Bologna).
water discharge and turbine output. The problem has never 1947b Energia Elettrica, vol. 24, ‘Problemi tecnici e
been dealt with completely and undoubtedly it deserves close sperimentali intorno alle vasche d’oscillazione’ and
attention. It is likely that the surge tanks, if there are any, Convegno di Idraulica, Padova, 13th October 1947,
stabilize the system. published 1948, Serie, Milan.
Special mention needs to be made of a particular type of 1950 Energia Elettrica, vol. 27, pp. 253, 353, ‘Pozzi
underground station (Fig. 1 2 4 where a downstream tunnel dis piezometrici e stabilitd di regolazione’.
charges water with a free level, as long as the discharge is steady. GADEN,D. 1945 ‘Contribution Q 1’6tude des rkgulateurs de
For large variations of load, the tunnel will be put under pres vitesse. Considerations sur le probleme de la stabilitt’
sure, the water rising in a chamber which has been called a (La Concorde, Lausanne).
PRESENT TRENDS I N SURGE TANK DESIGN 103
GHETTI, G. 1947a ‘Ricerche sperimentali sulla stabilita di JAEGER,C. 194849 Civil Engineering (London), vol. 43, p. 620,
Regolazione’: Part I. Convegno di Idraulica, 13th and vol. 44, pp. 38, 85, ‘Underground HydroElectric
October 1947, Padova, and Serie, Milan, 1948. Power Stations’.
1947b ‘Sullastabilitd delle oscillazione negli impianti idro 1949a ‘Technische Hydraulik’ (Verlag Birkhauser, Basle) ;
elettrici provisti di un sistema complesso di condotti e French translation 1953 ‘Hydraulique Technique’
pozzi piezometrici’ (Convegno di Idraulica, 13th October (Dunod, Paris).
1947), Padova, and Serie, Milan, 1948, and Energia 1949b Water Power, vol. 1, p. 25, ‘Developments of Intake
Elettrica 1947, vol. 24, p. 542. Works and Surge Tanks’.
1951 ‘Ricerche sperimentali sulla stabilitd di Regolazione’ : 1952 Water Power, vol. 4, p. 334, ‘Surgetank Stability’.
Part 11, Energia Elettrica, vol. 28, no. 11. JOHNSON, R. D. 1908 Trans. A.S.M.E., vol. 30, p. 443, ‘The
GIBSON,A. €1. 1921 ‘Hydroelectric Engineering’ (Blackie Surge Tank in Water Power Plants’.
and Son, London). 1915 Trans. Am. SOC.Civil Eng., vol. 78, p. 760, ‘The
Differential Surge Tank’.
1952 ‘Hydraulics and its Applications’y 5th edition PRASIL,F. 1908 Schweizerische Bauzeitung, vol. 52, p. 271,
(Constable, London); also 1st edition, 1908. ‘Wasserschlossprobleme’.
JAEGER, C* 1933 ‘ T h h i e GhQale COUP de BClier’ ROUSE,H. 1951 ‘Engineering Hydraulics’ ( ~ i l e yand Sons,
(Dunod, Paris). New York).
1938 Technipue Moderne, vol. 30, No. 2, p. 33, ‘L’Agran SCIMEMI, E. 1947 ‘sulla validitd delh regola di Thoma’,
dissement des Usines HydroElectriques’. Convegno di Idraulica, 13th October 1947, Padova, and
1939 Trans. A.S.M.E., vol. 61, p. 109, ‘Theory of 1948, Serie, Milan; and Energia Elettrica 1947, vol. 24,
Resonance in Pressure Conduits’. p. 537.
1943 Schweizerische Bauzeitung, V O ~ . 122, nos. 21, 24, 25, SCIMEMI, E., and GHETTI,G. 1951 Energia Elettrica, vol. 28,
26, ‘De la stabilite des chambres d’equilibre et des no. 9, ‘Sul dimensionamento delle vasche di oscilla
systkmes de chambre d’kquilibre’. zione’.
1948 Civil Engineering (London), vol. 43, pp. 74, 138, THOMA, D. 1910 ‘Zur Theorie des Wasserschlosses bei
244, ‘Water Hammer Effects in Power Conduts :Accidents selbsttatig geregelten Turbinenanlagen’ (Oldenbourg,
Due to Water Hammer’. Munich).
Discussion
Mr. HENRYHEADLAND, M.Sc., M.I.C.E., M.I.E.E. (London), effect of sudden shortcircuit loading, and to the advent of auto
who opened the discussion, said that, in those days of rigid matic reclosing and possibly singlephase switching on high
economy of expenditure, the authorities for whom consulting voltage transmissionsystems where the circuit breakers could
engineers were called upon to design and construct hydraulic reclose automatically a few cycles after opening.
schemes concerned themselves more with detailed design, and The author had mentioned model tests in a number of places,
surgetank dimensions were frequently challenged. For that without stressing their limitations in regard to surge phenomena,
reason alone the paper was valuable in emphasizing the risks or mentioning the application of differentia1 analysers or com
which might be run if engineers were to be led into accepting putors or the recent work of the Massachusetts Institute of
meagre designs and operating restrictions which might be Technology dealing with electrical analogues in a comprehensive
forgotten or become inapplicable with changing system con study of the whole system, namely, the network, alternator,
ditions. turbine, governor, penstock, surge chamber, and tunnel.
In the section of the paper which had described the simple The paper was, perhaps, a little vague on a number of points
restricted orifice and differential tanks, the author might have which the author had been at some pains to emphasize, namely,
indicated the conditions under which those various types were the decrement and limits of surge amplitude. He had yet to find
used and why the practice of engineers on the continent of a surgechamber designer or turbine designer who would give
Europe seemed to be to prefer the restricted orifice type, while a specific answer as to the acceptable magnitude of those charac
their American contemporaries, until recently, had shown a teristics for surge chambers upstream or downstream of the
distinct preference for the differential type about which the turbine. With underground stations utilizing tailrace tunnels
author had levelled some pertinent criticism from the analytical under pressure, the need for surge chambers was a matter on
point of view in regard to stability and other considerations. which the less experienced engineers had sometimes to be con
He was a little surprised that the author had made no reference vinced. As far as he was aware, the conditions under which
to the behaviour of surge chambers during starting and stopping downstream surge chambers became necessary, had never been
operations on pump storage schemes, including such stations as defined, and perhaps the author could say whether the criterion
Etzel, where the turbine was called upon to assist in dealing with relating to the penstock and machinetime constants for stability
surges following an electricity supply failure, or where the of governing in an installation with a free flowing tailrace could
recently developed combined turbine and pump could be used. be correctly extended to include the time constant of the tailrace
The author had rightly emphasized the loading conditions tunnel.
and surgetank design. That was probably the most unsatis I n his opinion the paper was a valuable one, apart from its
factory aspect of the whole problem and was a field in which a technical interest, because it definitely pointed to the need for
coordinated review by the hydraulicdesign engineer and field tests and to those aspects of design where investigations
systemoperating engineer would be valuable, He felt that the might profitably be pursued by those who were mathematically
author might have made a short reference to the work done in or experimentally inclined.
Italy, and more recently in Germany, on the effects of repeated
loading and unloading of the turbine. Mr. H. D. MORGAN, M.I.C.E. (London), said that he con
Two pointsand perhaps they were unimportanthad sidered the paper to be very useful for the reason that it dealt
occurred to him in considering this loadon condition, and he with all the various special types of surge chamber which had
mentioned them merely to illustrate the need for an appreciation been devised from time to time. He did not remember ever having
of the wider aspects of design. They related to faulty synchroniz seen them collected together before in the same way as the
ing, particularly in a singlemachine station, simulating the author had done. He wished, however, to draw the author’s,
104 DISCUSSION ON PRESENT TRENDS I N SURGE TANK DESIGN
attention to one case with which he had not dealt, namely, that the square of the velocity was obviously wrong for low
in which the main tunnel had a large number of intakes at velocities.
intervals. The method had also been used to design an enlargement at
In mentioning that, he was thinking of the Lochaber tunnel, the top of the surge chamber, because for some reason which he
where a large number of shafts brought water in at intervals, had never quite understood (he had not participated in the work
which made the mathematical treatment very complicated, from its inception) there had been an overflow tunnel and that
because there was a set of equations from shaft to shaft right tunnel had been set with its invert at approximately the same
through, and the work became very heavy. Of course, as the level as the top waterlevel of the reservoir. The idea had been
quantities entering those different intakes were variable, it was that, in order to avoid extending the surge chamber above
only possible to consider. the worst cases; but at Lochaber ground, it would stop there and overflow would take place in
where the tunnel was 15 miles long, when there was heavy rain the very unlikely event of a complete shutdown. But that had
on the flank of the hill and a lot of water coming into the intakes, rather left out of account the instance when all the side streams
that flow could supply pretty nearly the full load on the station; would be contributing a great deal of water, and it had been
and, of course, the friction terms in the equation were affected, so feared that although a normal discharge through the tunnel
that very much more severe effects were obtained at those times. would not do very much damage, an unusually large discharge
Fig. 18 illustrated that. It would be noted that finite difference would probably wash out the works railway. It had therefore
equations had to be used. Those equations were formed for been decided to attempt to limit the quantity of overflow, and
performing a stepbystep calculation. he had been given the task of investigating that.
The whole of that analysis had been dealt with in considerable The basic equation of surge in a tunnel with side stream intakes
detail in a paper by Cole (1927)*. At that time there had also was
been a model with which experiments had been done so as to
check the reasoning in the paper. He noticed that that paper had
not been mentioned in the bibliography of the present paper;
the author was probably familiar with it, and in his own opinion, where the terms were as in Fig. 18. The maximum value of y
it did contain some points of particular interest. was limited by the overflow tunnel but, with the original arrange
When the Lochaber scheme had first gone into operation, +
ment, the value of ( y R ) was so small compared with the other
certain tests had been made to check that method of calculation. terms that u fell asymptotically to zero along a line such as cy in
It was very difficult to achieve a textbook rejection of load; it Fig. 19.
had been necessary to recalculate as nearly as possible following A bellmouth orifice was therefore put into the overflow tunnel
what was done, which had not been very easy. I n the power to restrict the discharge to a safe amount and allow y to increase
house they had made preliminary adjustments of some of the to the greatest extent possible, the surplus water passing over a
spears and that had put a small surge on to the system which it weir into a reservoir excavated beyond. An emergency weir was
had been very difficult to distinguish afterwards. For the first provided to return excess water to the overflow tunnel in the
maximum surge, agreement to within about 1 ft. 6 in. had been unlikely event of the reservoir being wholly filled. The complete
obtained, the agreement being not so close for the first down arrangement is shown in Fig. 19.
surge. I n his opinion, the biggest difference had been in the Because of the higher water level in the surge shaft, the curve
time of oscillation, which had not conformed and had been as of u1 was now ca’x falling to zero in a finite time. The discharge
much as 10 per cent out. The only possible reason he could think through the overflow tunnel is represented by aa’, and the
of for that was that the assumption that the friction varied as total amount of water going into the reservoir by the area caa’.
This arrangement worked quite well in practice, except that
* COLE,R. S. 1927 I.C.E. Selected Engineering Paper No. 55. the annoyance caused by occasional small surges passing down
DISCUSSION ON PRESENT TRENDS I N SURGE TANK DESIGN 105
I I
TIME BY POWER
HOUSE CLOCK
f
It
FLOW IN
TUNNEL FROM
/ENTUN METERS
OPERATION 
CARRIED OUT
the tunnel was zero. That was the time to shut down the rest of The author’s opinion might be amplified by saying that, from
the plant. the civil engineering point of view, they were sometimes more
The amounts of plant shut down had been calculated by trial costly to deal with. For example, the tunnel had always to be
and error, and Fig. 21 showed what had taken place in practice. kept below the lowest downsurge level and also below the
About 65 per cent of the load had been shut off initially and the corresponding hydraulic gradient. I t was necessary, therefore,
remainder of the load had been rejected when the surge crossed to ensure in certain instances that the down surge never went
the line XY. below a certain level, because otherwise the tunnel gradient
would become too steep for convenient or economical driving.
Mr. OSCARELSDEN,M.Sc., M.I.C.E. (London), said that Satisfactory conditions could usually be produced with a simple
he thanked the Institution for inviting him to be present at the chamber, if it was big enough, but various attempts to economize
presentation and discussion of that interesting paper. had produced the chambers illustrated in Figs. 47, and also
The first problem encountered in the design of a surge chamber the spilling type, which was not illustrated in the paper. Calame
was to determine the worst flow variations to which.the system and Gaden (1926) in their wellknown textbook had discussed
would be subjected, and it was refreshing to find that the author those types, and the determination of their dimensions. Their
had devoted a fair amount of space to a subject which was methods had been aimed at producing a chamber of minimum
virtually ignored in the majority of standard textbooks from volume on the assumption that in an excavated chamber the
which the ordinary civil engineer, such as himself, had to start. cost was proportional to the excavated volume of rock. In times
In his own experience (which was limited mainly to the of scarce materials, shortage of skilled labour, and shortage of
Scottish stations) surge chambers had always been designed for skilled designers, it was not a foregone conclusion that reduction
fullload rejection at maximum waterlevell and in some of excavated volume would produce a proportionate saving either
instances it had also been necessary to determine the down surge in time or in cost, and a considerable saving in paper estimates
caused by fullload rejection at minimum waterlevel. was necessary to justify any complication in design. I n recent
Occasionally, it had been necessary to consider overlapping or years, he had given consideration in special cases to the use
compound surges, such as had been mentioned by the author on of spilling chambers or chambers of variable section, but they
p. 97 and as had also been described by Calame (1934)*. had not been found to have any economic justification. The works
It was necessary to make various assumptions for load thrown necessary to carry the water from a surge chamber spillway to the
on, sometimes according to the wishes of the customer, but he nearest watercourse could be quite expensive.
suggested that the loadthrownon assumption was of importance He disagreed with the author’s description of Fig. 5 as repre
with a differential chamber, which showed to its best advantage senting a surge chamber with a restricted orifice : the area of the
when designed for only partload thrown on. The maximum port or orifice as shown appeared to be appreciably greater than
loadchange was in practice never complete and never instan the crosssection of the tunnel. That arrangement had been
taneous. The machines were idling under about 10 per cent of adopted in all the surge chambers with which he had been
the discharge before they were loaded, and would return to this associated; in each instance, they had been designed, and had
condition on abrupt rejection of load. Occasionally the idling functioned, as simple surge chambers. Every surge chamber had
flow was allowed for in the design assumptions, but in calcula an inherent throttleeffect caused by the unavoidable change of
tions the changes were usually taken as instantaneous. direction and change of section at the junction between the
In the lowerhead stations, the area of the surge chamber was tunnel and the surge chamber. When the port or ports had an
usually governed by the Thoma equation, and common practice aggregate area equal to or greater than that of the tunnel, the
was to adopt a factor of safety of about 2. For the higherhead throttle effect was usually small, even for relatively short tunnels.
installations, the surge amplitudePwo2+z in Fig. 1was often His own rather loose definition of a throttled surge chamber was
limited to about 15 per cent of the net head, and the chamber was that it was one in which a really severe throttle was introduced, so
often considerably greater than was necessary to secure stability that it had a pronounced effect upon the oscillations within the
alone. The author had mentioned on p. 91 that negative pres chamber and upon the ultimate dimensions of the chamber itself.
sure effects were as important as the positive pressure effects. He was particularly interested in the author’s remarks on the
* CALAME, J. 1934 Bulletin Technique de la Suisse Romande, 26th differential surge chamber, and he would like to hear the
May, 9th June. author’s reasons for the apparent popularity of that type of
DISCUSSION ON PRESENT TRENDS I N SURGE TANK DESIGN 107
chamber in the United States, having regard to the fact that it in touch with reality, whereas when analytical methods were
was seldom, if ever, used in Europe. used it was necessary to return from time to time to reality to
In regard to the detailed technique of design in Britain, he check the reasonableness of conclusions. The graphical method
was glad that Mr. Morgan had mentioned the paper by Cole dealt perfectly easily with, for instance, the partition and partial
(192?), which he considered might be said to have been the reflection of the elastic waves when they came to a pipe junction,
starting point of surge chambers in many British offices. In more or with throtties or any other of the circumstances which arose.
recent years they had tended to adopt the shortened integration He wished, therefore, to put in a plea for the stepbystep
method quoted by Schoklitsch (1937)* which, though less treatment of both the elastic waves and the disturbances which
precise than that of Cole, was more rapid and was sufficiently neglected elasticity.
accurate for practical office use. A graphical method had been
described by Calarne and Gaden (1926)t and also by Escande
(1950)$. The second graphical method, which produced a curve Mr. S . P. HUTTON, M.Eng. (Associate Member), A.M.I.C.E.,
in polar form, presented an attractive method of assessing the said that he had found the author’s interesting survey of that
stability of a surge chamber. specialized subject to be very stimulating.
For the sake of the amenities and for other reasons, surge Most of the knowledge gained so far on that subject had been
chambers were usually formed by excavation. The excavation from the practical experience involved in making stepbystep
of the chamber was usually carried out from a tunnel beneath by calculations and comparing the results with those actually
raising a stope upwards to the surface. Stoping was a slow and obtained in the field. Relatively little work had been done in
somewhat hazardous operation, but that was to some extent the laboratory, although it was known that there was a growing
true of shaftsinking, and the stope had the advantage of being use of laboratory work. As Mr. Headland had mentioned, there
selfdraining and more convenient for muck removal. The stope had to be limitations in that field, and he would like to discover
also took advantage of equipment already installed for spoil the author’s present opinions as to how far smallscale experi
disposal from the tunnel. The stope generally formed a pilot ments under controlled conditions in the laboratory might be of
shaft down which the material was shot when the chamber was service to the design engineer.
fmally excavated to its full size. The surge chamber was usually There were, obviously, many difficulties associated with the
lined with concrete, which was often placed within continuously appropriate scale laws to apply between model and prototype.
climbing shutters. That was a rapid and, on the whole, satis For instance, one of them, which the author had stated, was that
factory method of lining the chamber, when it was of simple increased pipefriction losses tended to increase the stability
cylindrical form. of the system. Such losses increased relatively with decreasing
pipe size or model size correspondingto the decrease in Reynolds
Mr. H. R. LUPTON, O.B.E., M.C., M.A. (Member), said that number. He wondered whether that meant that a model was
the author had mentioned that there were two different kinds likely to be more stable than its prototype, or whether it was
possible to make suitable adjustments so as to keep all important
of pressure disturbance. The first were the shortperiod features to scale. That was one illustration. There must be many
disturbances which were dependent on the compressibility of other complications in that work, and it would be useful to know
the water and the elasticity of the containing surfaces, and the how far laboratory experiments were able to contribute in
other the longperiod disturbances (which had been the main predicting fullscale performance.
subject of the first part of the paper). The latter would take
place even if the water were incompressible and the containing He asked whether, with the restrictedorifice type of surge
surfaces rigid. The actual effect in the end was a combination of tank, when subject to oscillations, the flow losses in both direc
the two. tions in and out of the tank were assumed to be equal. The fact
To work those out separately and then to add them together that the water turned a 90 deg. contracting or expanding bend,
dependent on whether it was entering or leaving the chamber,
might not be accurate. Pressures and velocities might simply be would obviously give different Iosses. Additional differences
added arithmetically, but errors were introduced when, for would arise because mass was either abstracted from or added to
instance, there was friction or throttle, or anything of that kind, the main flow in the penstock. For that type of restrictedorifice
because (a+b)2 was greater than a2+bZ. tank, and for some of the others perhaps, the losses might be
Error was incurred if the surges were worked out in unequal for various reasons, and he would like to know whether
dependently, and then combined, wherever the relation that was likely to have an important bearing on the surge
between heads, accelerations, and velocities was not linear. characteristics of the complete system.
Therefore, he made a plea (with which he thought the author The author had mentioned the importance of the inter
sympathized, for he had already mentioned it) that the stepby relation between the turbine governor and the surgetank
step method, which had to be used where the layout was very characteristics. He asked whether the author had ever en
complicated, should be used to include both the elastic waves and countered any resonance failures where backlash in valves had
tne longperiod disturbances which were the main subject of tended to set up oscillations in the system in conjunction with
the paper. It seemed to him that only a combined investigation of the other factors already mentioned. Certain valve designers on
that kind could be perfectly safe. the European continent seemed to have gone to great lengths to
He knew that that was laborious, but he had worked out many operate valves directly by hydraulic means using toroidal pistons
instances of surge in pumping mains though not in hydro and cylinders without any linkage gearing or counterweights, and
electric schemes and the process was, of course, similar to the he wondered whether the reason for that rather expensive con
SchnyderBergeron diagram (Bergeron 1935) which the author struction was merely to provide as compact a unit as possible, or
had shown. The boundary conditions were the conditions whether it was to avoid the possibility of any oscillations that
imposed, in the one case, by pump pressures and flows and, in might be caused by backlash in operation.
the case of the hydroelectric station, by turbine pressures and
flows, and by the levels in tanks or pressures in air vessels as the
case might be. The effects of throttled pipes and of pipe junc Dr. 0. C . ZIENKIEWICZ, A.M.I.C.E. (Edinburgh), said
tions were quite amenable to treatment by the Schnyder that the discussion illustrated one of the points made in the
Bergeron analysis referred to. paper, namely, the need for cooperation between ,the various
Of course, distributed friction had to be regarded as con sides of the profession engaged in the design of surge tanks. It
centrated at various pointsthe more, the more accurateas was the civil engheer who was ultimately responsible for the
throttles and he, personally, had found that the graphical method surge tank, and it was necessary not only for him to be given
was much the easiest, since it was possible to remain constantly data from the designers of the turbines and generators, but also
that he should understand the problems involved in the regula
* SCHOKLITSCH,
A. 1937 ‘Hydraulic Structures’, vol. 2. tion of those and, possibly, be able to influence the design af
t CALAME,J., and GADEN,D. 1926 ‘Thkorie des chambres the machinery to effect overall economy.
d’bquilibre’.
$ ESCANDE, L. 1950 ‘Methodes nouvelles pour le calcul des chambres I n the introduction to the paper, the author had stated the
d‘kquilibre’. essential reasons for the existence of surge tanks. The first of
9
108 DISCUSSION ON PRESENT TRENDS I h SURGE TANK DESIGN
those (and to his own mind the most important) was the suppres Mr. PETER LINTON, B.Sc. (Associate Member), A.M.I.E.E.,
sion of waterhammer waves in the socalled lowpressure said that the author had done a great service in presenting his
sections of the aqueducts and the occasional reduction of those excellent review of recent developments in the theory and con
pressure waves in the highpressure sections. It seemed to him struction of surge tanks, a field of development to which he (the
that that aspect was frequently overshadowed by the necessary author) had contributed in no small measure. He himself
and elaborate calculations of the stability and amplitudes of mass believed that nobody would be misled by the few equations
oscillations which, after all, were only a consequence of the quoted in the paper into thinking that problems of that type
introduction of the surge tank itself. might be solved by a few lines of arithmetical calculations. On
Recently, at Edinburgh University, he had had the opportunity the contrary, the types of surge tank proposed recently, so as
of being in charge of an investigation into the fundamental to reduce excavation costs, required extensive stepbystep
aspects of pressurewave transmission through various types of calculations even after their basic theory had been established.
surge tank. That had been an experiment with a model. Previous He wished to reinforce some of the remarks which had already
speakers had referred to models, and he considered that it was been made during the discussion in connexion with tests on
impossible to simulate all the actions of a surge tank in one models. It seemed to him that one of the characteristic features
model j but, for the particular purpose of studying transmission of the average British engineer was his mistrust of mathematical
of waterhammer waves, it was possible to obtain the necessary methods. That might be due to his lack of experience in handling
similarity. In the investigation, orifice and differential surge anything but the simplest differential equations, but he thought
tanks had been the main subject of research, and as he hoped it was more probably due to a conviction that many of the
that that work would be published shortly, he would confine his assumptions from which mathematicians cheerfully constructed
remarks to a brief statement. their theories did not always hold in practice. There was, of
It had been found, as the author had stated, that by the course, always the possibility of errors, either in the abstractions
graphical methods of Bergeron (1935)and by other methods of required to make the problem amenable to calculation or in the
calculation it was possible to calculate to a very good degree of extensive and often tedious numerical work involved, unless the
accuracy the transmission of pressure waves in all types of surge task was entrusted to a specialist who could generally tell
tank. An accuracy of 23 per cent had been obtained even for whether the answer was reasonable or not. However, even the
orifice or differential surgetanks. Broadly speaking, it had been specialist would often have difficulty in convincing the engineer
found that even with relatively small sizes of orifice the trans that the result might be accepted as correct without further
mitted pressures were not excessive in all instances except those socalled safety factors j for, after all, heavy damage and even
of pressures due to instantaneous stoppages of flow. The case loss of life had to be expected if a waterpower scheme were
of instant stoppage of flow was obviously not a practical danger. incorrectly designed.
Though sometimes assumed in design, the justification of that There appeared to be a very strong case for the extensive
assumption lay in the fact that as long as the total time of closure employment of tests on models in the design of surge tanks.
did not exceed the period of the simple conduit the pressures The role of experiment in engineering had been the subject of
produced would be identical. As far as the magnitude of the correspondence in the technical press and there seemed to be
transmitted pressures was concerned, it could be shown that, general agreement that results obtained from tests on models
with closures taking a finite though small time, those might be were valuable even when they did no more than confirm and
reduced to a small fraction of their ‘instantaneous’values. Thus, illustrate the mathematical calculations. In the instance under
for checking the efficiency of a surget+ installation in respect discussion, however, such tests could provide information over
of the transmission of pressure waves, it was imperative to know a much wider field, When applied to a particular installation,
the maximum rates of increase and decrease of flow. they would generally supply numerical results in a shorter time
He would be greatly interested if the author would add to than the stepbystep calculations, especially when a large
the list of operating conditions assumed by continental designers number of loading conditions had to be investigated.
some figures relating to that problem. It had to be realized that Models might be used on new types of surge tank for which
those maximum rates were not necessarily related to the no theory had so far been developed. Even more important,
governing constants. Accidental gatemovements might occur however, was their application to the determination of the
at much greater speeds, but even so would never reach the experimental constant in the surgetank equations. Figs. 47 in
impossible ‘instantaneous’ case. the paper showed that modern surgetanks consisted of orifices,
The second reason for surge tanks, as stated in the paper, bends, tees, and other hydraulic elements placed so closely
was the instability of governing which might occur with very together that they were bound to interact with each other. The
long conduits. That was a question of balance. Even in relatively calculation of the appropriate losscoefficients from purely
short conduits it might be possible to introduce instability by theoretical considerations were out of the question, as was their
too sensitive governing. There must be many practical borderline measurement on completed surgetanks owing to the difficulty
instances where the quite considerable cost of provision of a in metering the large flowrates involved. Accurate values of
surge tank might be avoided by the installation of less sensitive the loss coefficient which played an essential role in the deter
governing and by larger inertia of the machines. The economics mination of the stability of the system could only be obtained
of such projects ought to be investigated carefully by both the from a carefullymade model based on the wellestablished
civil and mechanical engineers concerned. similarity laws of fluid dynamics. Finally, if the model was kept
Finally, in regard to the statement in the paper: ‘It will be in working order during the actual construction of the scheme,
observed that the slow surges in the surge tank cannot be it was possible to determine immediately the effect of any
explained by elastic deformations similar to those producing the modifications which might be made necessary by geological and
water hammer of short periodicity’, he was in agreement with other factors arising during excavation.
the remarks already made by Mr. Lupton. Possibly that state The author had brought out very clearly the double function
ment had been wrongly worded. The basic equations of un of a surge tank, as had several contributors to the discussion.
steady motion involving compressibility ought to be capable It had to prevent the transmission into the pressure gallery of
of explaining all the transient phenomena more accurately than elastic pressure waves, the socalled water hammer, and also to
the approximate relations used in the conventional mass provide a balance reservoir for the water absorbed or rejected
oscillation equation. In the treatise by Bergeron (1935) there by the turbines when their load was altered. Practical surge
were a few examples of surge tanks being treated in regard to tanks were imperfect in both respects. A certain small part of
water hammer by linking the problem of the elastic equations the waterhammer pressure generally found its way into the
with the mass surges. The author himself in his earlier works gallery, while the kinetic energy of the water column produced
had shown that that was possible. The only reason in practice mass oscillations whose danger lay not so much in the pressure
why that was not generally done was that the periods of the two corresponding to the maximum level reached, as rather in the
oscillations differed so widely, and that provided an excuse for possibility of overflowing and air entrainment at the two
the separating of the two phenomena. There were some instances extrembs of the oscillation and instability of the governor during
where it might be necessary to consider the two in combina load changes. The actual period of those two types of surges
tion. differed very considerably. Waterhammer periods were of the
DISCUSSION ON PRESENT TRENDS I N SURGE TANK DESIGN 109
order of several seconds and became very short indeed on any so that a considerable reduction in size was possible on the
model of reasonable size, thus requiring special equipment to model without causing difficulty in recording the level fluctua
measure the pressure fluctuations accurately. Dr. Zienkiewicz, tions.
who had referred to models, would have had considerable Work in that field appeared to be in progress at a number of
I
experience of the difficulty of evolving reasonable models which continental laboratories, and it would greatly assist research in
gave reproducible results. The mass oscillations in a surge tank Britain if the author were to comment briefly on the methods
had periods of several minutes, and that time was approximately employed and on the value of the results obtained in such tests
proportional to the square root of the linear size of the system, on models.
Communications
Mr. T. M. CHARLTON, B.Sc. (Eng.) (Associate Member), wrote As they had suggested (CuCnod and Gardel 1950t, 1952,
that he welcomed the author’s comprehensive review of surge 1952a$), the cause of that instability disappeared if, after a rise
tank design. It was probably true to say that the occurrence of in the water level of the surge chamber, the power output of
waterhammer had a primary influence upon the choice of the the station rose in order to reduce, or else to suppress, the closing
regulation or governing characteristics of a water turbine. While of the turbine, and wice wersa in the case of a fall in the water
relief valves and deflecting equipment operated by the governor level. That variation of the electrical output proportional to the
could be used to limit the severity of waterhammer when load hydraulic head could be realized :
was rejected, there was no similar expedient available for limiting
the (negative) waterhammer due to the acceptance of load. The (I) by modifying the voltage of the generator according to
latter was, therefore, often of first importance in the design of a the head or water level, if the load had a resistive character,
hydroelectric installation. The provision of a suitable system (2) by modifying the frequency in the same way, if the net
of surge tanks reduced the pressure variations due to water work included motors.
hammer as well as, for practical purposes, relieving an appreciable The stability condition was then :
portion of the hydraulic system from such pressure variations.
The question whether to employ surge tanks to limit the water F > ( 1 K)F,
hammer, or whether to design the whole of the conduit (including where F = crosssectional area of surge chamber ; FT = limit
its profile) so that the pressure variations due to waterhammer as of crosssectional area as given by Thoma’s stability criterion;
a result of the slowest possible action of the turbine governor a n d K =  AN”o = control coefficient, with AN = variation of
could be withstood safely, had to be settled with due regard to AHIHO
considerations of economy. power proportional to the hydraulic head, AH = water level
He was surprised that the author seemed to express a pre variations in the surge chamber, and Ho, No = value, respec
ference for the analytical approach to waterhammer investiga tively, of the water level and load in steady state conditions.
tions. While a lack of generality (in the sense that each problem On first approximation, they could admit that the speed
had to be solved ab initio) was admitted, the graphical method control acted at once compared with the natural period of
of SchnyderBergeron (Bergeron 1935) was usually considered oscillations. However, in some particularly unfavourable
to be very easy to handle. instancesif the periods of oscillation of the speed control
It was worth recalling that Dr. F. Vogt’s ‘Berechnung und (governor) had approximately the same value as those of the
Konstruktion des Wasserschlosses’(Vogt 1923)* contained some surge chamberthat hypothesis was not fulfilled. The stability
useful curves for the determination for a simple cylindrical surgeof the whole installation, which included the station, its surge
tank (with or without a throttle) of (1) the maximum upsurge chamber, its pressure gallery, its speed control, and the network
when the full load was rejected instantaneously and (2) the maxi which it supplied, had to be considered (CuCnod and Gardel
mum downsurge when the load was increased instantaneously 1952a). In those instances the control of electrical load from the
from any part load to full load. Although load changes required a hydraulic head had to be completed with a control of electrical
finite time to occur in practice, the period of oscillation of a surge
load from the frequency.
tank was usually long compared with the action of the governor or The efficiency of that kind of stabilization had been confirmed
relief device, and the assumption of instantaneous action did not through experiments carried out by the authors under the
cause appreciable errors which, in any event, were on the side direction of Professor Bonnard and with the collaboration of the
of greater safety for purposes of design. Hydraulic Laboratory of the &ole Polytechnique de Lausanne
(Cubnod and Gardel 1953)s.
Mr. MICHELC ~ N O (Geneva)D and Mr. AND^ GARDEL Those experiments had taken place in the Oelberg Power
(Lausanne) wrote that they had investigated, by both theoretical Station, Fribourg, Switzerland. That station had been chosen
study and experimental results, new means of ensuring stable because it had a differential surge chamber (Johnson type). By
operation of hydroelectric stations with surge chambers. In closing the gates connecting the well to the chamber itself, the
1910 already, D. Thoma had shown the repercussions of the surge chamber area (300 sq. metres) had been reduced to that
stability conditions of waterlevel oscillations in a surge of the well area only (7 sq. metres) which was very much less
chamber on the choice of the dimensions required for that than the required area of Thoma (40 sq. metres).
chamber. The characteristics of the station used for the tests were:
That ‘condition of Thoma’ had been modified afterwards to nominal power, 7,500 h.p.; nominal rate of flow, 32 cusec.;
take account of some more or less secondary effects. Those
f CU~NOD,
different calculations had acknowledged the hypothesis that the Suisse M., and GARDEL, A.‘ 1950 Bulletin Technique de la
Romande, vol. 76, p. 209, Stabilisation des oscillations du
load produced from the station under consideration was constant. plan d’eau des chambres d’kqmlibre’.
Owing to the action of the turbinespeed control, a rise in the $ CU~NOD, M., and GARDEL, A. 1952a Bulletin Technique de la
water level of the surge chamber led in compensation to a closing Suisse Romande, vol. 78, p. 201, ‘Stabilitd de la marche d’une centrale
of the turbine; that stressed that rise and led to unstable con hydrokIectrique avec chambre d‘kquilibre, compte tenu des carac
ditions, if the surgechamber area was insufficient. tdristiques dynamiques du rkglage de vitesse’.
§ CU~NOD, M., and GARDEL,A. 1953 Bulletin Technique de la
* VOGT, F. 1923 ‘Berechnung und Konstruktion des Wasser Suisse Romande, vol. 79, p. 365, ‘Essai de stabilisation du rkglage
schlosses’ (F. Enke, Stuttgart). d’un groupe hydroklectrique avec chambre d’kquilibre’.
110 C O M M U N I C A T I O N S O N P R E S E N T T R E N D S I N SURGE T A N K D E S I G N
maximum head, 20.8 metres; length of pressure gallery, 270 restricted orifice concerned would have to be such as to ensure
metres; crosssectional area of pressure gallery, 18.3 sq. metres; that the upsurge whichoccurredat the timeof totalinstantaneous
length of penstock, 47 metres; crosssectional area of penstock, closure of the discharge Qo would be greater than that engendered
10.2 sq. metres; Kaplan’s turbine, speed, 250 r.p.m.; moment by any operation of partial closure.
of inertia of the rotating parts, PD2 = 170 tons metres*. They had found (Escande 1950)* that in that case the extent
Preliminary experiments had been carried out in December of the upsurge due to instantaneous closure might attain 125per
1951. The results had been negative, for it had not been possible cent and that the extent of the maximum downward surge due
to obtain stable conditions in spite of a control of the electrical to instantaneous opening might be 625 per cent.
load by the water leveI. Those tests had shown that if the surge
chamber was reduced to the well, its oscillations were nearly the Calculation of the Pressure Waves Transmitted to the Delivery
same as those of the speed control (1820 seconds). The problem Tunnel by an Ins@ciently Throttled Surge Chamber. The author
of the stability of the system had to be considered as a whole had drawn attention, quite rightly, to the high pressures
(surge chamber and speed control). engendered in the delivery tunnel by the increase in pressure due
The main experiments had taken place in June 1952. The to loss of load which caused the transmission of the discharge
control coefficient of the load by the water level and the frequency through the restricted orifice following a rapid stoppage of the
had been chosen according to the values found in the theoretical discharge from the turbines.
study. When, on the other hand, they had approached the matter
During those experiments, the turbine opening had been solely from the point of view of the transmission of highpressure
submitted to a voluntary perturbation, and then, the system had waves proceeding from the pressure tunnel, they had, first of
acted on its own, the load being a hydraulic resistance. all, tested a method of calculation, by means of the graphical
In spite of a surgechamber area which had been reduced in method of Bergeron (1935) in the case of a connexion of a certain
the proportion of 40 to 1, and in spite of particularly unfavourable length between an ordinary surge chamber and the delivery
conditions due to the very short period of the surgechamber tunnel (Escande 1946a)t.
oscillations, steadystate conditions were obtained. Furthermore, when considering the restricted orifice as a
connecting pipe of very short length, they had, in order to effect
Theoretical considerations confirmed by experiments had a toandfro path along the pressure tunnel, applied that method
therefore shown that, owing to judicious control of the electric to the study of the high pressures transmitted by the action of
load by the surgechamber water level and the frequency, a abrupt closure of the turbine discharge, in other words, by the
hydroelectric station with a surge chamber could operate disturbances of short duration at time 8 set up by a wave.
steadily no matter what the dimensions of that surge chamber They had in the fmt place carried out that investigation while
might be. Consequently, it was not necessary to take Thoma’s paying due attention to those same characteristics for the delivery
stability criterion into consideration in fucing the dimensions of and pressure tunnels and while neglecting the loss of load due to
a surge chamber. traversal of the restricted orifice as well as the pressure variations
That stabilization was obtained by holding less rigorously to resulting from the change in water level in the surge chamber.
the voltage or the frequency. However, they had stated that the They had been able to deduce from that investigation an expres
voltage or frequency variations resulting from that kind of sion which would immediately give, to a very approximate
stabilization had very little importance (CuCnod and Gardel degree, the maximum high pressure transmitted : the error
1950). Those variations only occurred exceptionally when a corresponding to the employment of that formula was always
hydroelectric station operated separately from others., in the direction of safety and became practically negligible for all
It was known that when a station with a surge chamber periods of cessation of discharge occurring in practice, even in
operated in parallel with other stations, and when the power of the case of stoppage due to a serious accident.
that station was less than a third of the total power in the network, The investigations had shown them that, on the borderline
that station operated under stable conditions, no matter what between abrupt and slow closures, the high pressure transmitted
might be the area of the surge chamber. Owing to the steadily corresponded only to an inconsiderable fraction of the high
growing interconnexion of generating stations, that condition pressure at the governor. They had come to the conclusion that
was nearly always realized. However, any station might, at any fur all normal behaviour of the governor, putting into operation
time, operate independently. In that exceptional case, it did not slow closures, of duration several times greater than the limit
seem economical to obtain its stability through extending the of abrupt closures, the wave transmitted to the delivery tunnel
dimensions of the surge chamber, when the same result was was practically negligible (Escande 1946b)$.
obtainable by means of a relatively simple and cheap regulating They had completed the work by taking into account such
device. factors as the loss of load due to passage through the restricted
orifice, as well as those of different characteristics of the delivery
Professor L. ESCANDE (Toulouse) wrote that the paper con and pressure tunnels and, finally, that of the simultaneous
stituted an outstanding review of the design of surge tanks and presence of a restricted orifice and a connecting pipe of relatively
of their development. inadequate crosssection.
He was personally in agreement with the author as to the
general contents of his paper and the present communication Conditions far Stability of Oscillations in Two Surge Tanks
was merely by way of indicating very briefly certain special Situated Respectively in the Upstream and Dmnstream Portions
points which had come to light during some work of himself and of a Plant. I n collaboration with Huron, they had investigated
his colleagues and which had seemed to them to be of some the stability of oscillations in a system comprising two surge
interest for practical application. chambers which had been introduced one at the extreme down
stream end of the delivery tunnel and the other at the extreme
Influence of Time of Closure in Surge Tanks with Restricted upstream end of the tailrace, and which operated under pressure
OriJce. He and his colleagues had shoyn that, contrary to (Escande and Huron 1953)s.
what occurred in an ordinary surge chamber, the actions of They had carried out that investigation by making use in the
instantaneous opening and closure were not those which general equations of the same hypothesis of linearity as that on
provoked the strongest oscillations in restrictedorifice surge which the equation of Thoma (1910) was based. On applying to
tanks. A uniform rate of closure taking place in a limited period the differential system already obtained by the author the classic
of time could increase to a notable extent the rise or fall in the
water level. * ESCANDE, L. 1950 Acadkmie des SciencesComptes Rendus,
They had investigated, in particular, the case of surge shafts vol. 230, D. 1932.
t ES&E, L. 1946a Acadkmie des SciencesComptes Rendus,
of constant crosssection, furnished with a restricted orifice vol. 223, p. 130.
which, while reducing to the minimum the upsurge resulting $ ESCANDE, L. 1946b Acadimie des SciencesComptes Rendus,
from an instantaneous closure, would prevent the maximum vol. 223, p. 499.
pressure in the delivery tunnel from exceeding the value which ~ESCANDE, L., and HURON, P. 1953 Acadhie des Sciences
it attained at the end of the first upsurge. Furthermore, the Comptes Rendus, vol. 236, pp. 1329, 1405.
COMMUNICATIONS ON PRESENT TRENDS I N SURGE TANK DESIGN 111
formulae for working calculations, after introducing the Laplace account a sudden acceptance of the full and not only of a part
conversions, they had obtained an equation of the fourth degree of the installed power. Another important question raised by
and they had derived therefrom five expressions, constituting the author concerned the successive movements of closing and
the conditions of stability, of the form : opening of the turbine gate. The worst possible case would be
a l > o ; Q > 0 ; a3>0; a4>0;
a long series of load acceptances and rejections, which followed
one another with a rhythm which coincided with the natural
T = a32a1(~~2~(3~l~~4)<0 period of the conduittank system. But that occurrence was so
the values of q, m2, a3, a4 being calculated from the data for improbable, and so difficult to deal with, that it was normally
the installation by the means which they had worked out. disregarded. Consequently, a rather perplexing problem arose
as to what succession of movements was to be taken as safe
Condition of Stability of Surge Tanks with Venturi Effect. enough, in the sense that it attained a fair compromise between
They had demonstrated (Escande 1954a)* that, when Po‘ cost and risk of inconvenience. In that connexion it was to be
denoted the loss properly socalled in the delivery tunnel, and remarked that the effect of repeated movements could become
Eo the height due to the velocity at the neck of the Venturi quite dangerous even in the simplest cases, quoted by the
restriction, and when Po = Eo+P/, then, for a velocity, wo, author, of openingclosing or closingreopening movements.
corresponding to a steady rate of discharge through the turbines, For instance, if a load rejection followed a load acceptance with
Qo, concerning which the stability was being investigated, the a delay of half a period, the conduit discharge to be discontinued
formula usually considered (Jaeger 1949)t for the crosssectional became far larger than the steady state value (in the case of a
limit of the chamber, Fo,below which the oscillations increased cylindrical tank and negligible friction losses, the discharge
instead of dying out, might be modified and retained in the form : was exactly doubled).
Such questions of successive surges were even more delicate
for the damped tanks (with or without an internal riser). In the
simple restrictedorifice chamber the consequence of the inflow
where Ho, L,andf had their usual notation. exceeding the one assumed in the calculations was a far larger
loss of head through the orifices, which was likely to become
Conditions of Stability of Oscillations in Two Surge Tanks with dangerous not only for the upstream conduit, but for the down
Venturi Effect, Situated Respectively Upstream and Downstream stream penstock as well (owing to the reflections of negative
of the Turbines of a Plant. They had investigated the stability pressure waves being only partial). The differential tanks were
of a system of two surge chambers with Venturi effect, one in danger for a closingreopeningclosing succession, because
placed at the extreme downstream end of the delivery tunnel the emptying of the chamber was normally slow, and the second
and the other at the extreme upstream end of the tailrace of a closing was likely to find the chamber still partly filled. All that
plant supplying power at a constant rate on an isolated network. confirmed and emphasized the observations of the author, who
Although a Venturi effect would not be of any interest in had suggested that the riser and ports ought to be studied and
connexion with the tank situated in the tailrace, they had proportioned with great care.
assumed its existence for the great majority of results obtained : A good part of the paper was devoted to the problems of operat
those were immediately applicable to normal surge chambers, ing stability. He would like to dwell a little longer on such topics
namely, the particular case when the Venturi effect became nil. which had been the object of much research by himself.
They had established a method of calculation of the first In his opinion, the classic treatment by Thoma arrived at a
order of approximation and had derived five conditions of clearer physical evidence, and a greater suitability for wider
stability, as in the case of the ordinary surge tanks already development, if the geometrical and kinematic quantities
discussed above (Escande 1954b)*. describing a given plant at a given steady state were combined
In collaboration with Guevel, they had carried out a calcula in fewer and more significant parameters. After that course, the
tion of the second order of approximation and had shown conduit was represented by a time T, and a dimensionless
what was the new form then taken by the five expressions which parameter hb which were defined (using the author’s notation)
constituted the conditions of stability (Escande and Guevel by the formulae
1954)s.
The examination of a certain number of examples had made it
possible for them to record that the employment of those two
methods had led to variations which did not exceed 8 per cent Another parameter, the time characteristic T,, given by
in the cases concerned (Escande 1954c)[[.
In conclusion, he wished to stress the fragmentary character (9)
ofhis remarks and to emphasize once again the most interesting
nature of the author’s paper. described the tank.
The physical meaning of the new quantities was readily
Professor GIUSEPPE EVANGELISTI disclosed. The parameter hg was the ratio between the head loss
(Bologna, Italy) wrote that in the conduit
the author’s paper, dealing with the wide question of surge tanks was measured by and the net head, so that the conduit efficiency
in hydroelectric plants, was a most valuable one, and would be be called the ‘time l/(l+hg). The time T g w h i c h might properly
read with interest and profit by all kinds of technicians: both conduitwas twice the characteristic’ or ‘specific inertia’ of the
those little acquainted with the subject as well as the specialists. the conduit divided by kinetic energy of the water column in
the power delivered to the turbines;
As the author had observed, surge tanks had to meet two T,, the time characteristic of the tank, had the same meaning
different kinds of requirements, i.e. they had (1) to be able to with respect to the potential energy of position accumulated
deal safely with the largest possible load rejection and acceptance, inside the tank (its height supposed to be equal to the net head).
and (2) to provide stability during normal operation. In other words, at the given steady state each kg. per sec. of
In regard to the first problem, he agreed with the author’s output power required an amount of 8 . T, kg. of kinetic energy
views, namely, that the load variations to be taken into account and 3 . T, kg. of potential energy to be stored in the conduit
ought to be more rigorous than had been thought sufficient some and the tank, respectively.
years ago: for instance, it seemed indispensable to take into It was easily shown that by making use of the parameters,
* ESCANDE, L. 1954a Acadtmie des SciencesComptes Rendus, h,, T,, T,, the coefficients of the characteristic equation of the
vol. 238, p. 183. oscillations
f JAEGER, C. 1949 ‘Technische.Hydraulik’, p. 223.
f ESCANDE. L. 1954b Academe des SciencesComntes Rendus,
Cor2+ClrjC2 = 0 . . . . (10)
voi. 238, p. 537. were defmed as
$ ESCANDE, L., and GUEVEL,
Comptes Rendus, vol. 238, p. 756.
P. 1954 Academie des Sciences Co = TvTg C1 = 2hgTvTg C2 = 12hg . (11)
11 ESCANDE, L. 1954c Acadtmie des SciencesComptes Rendus, It was well known that the necessary and sufficient condition
vol. 238, p. 865. for the stability of the characteristic exponents (i.e. of the roots
112 COMMUNICATIONS ON PRESENT TRENDS I N SURGE T A N K DESIGN
of equation (10))were that both C,and C2 were positive, and near as possible to the power house held good also for stability
such condition reproduced, in a different form, the results purposes.
obtained by Thoma. Such a rule could be specified in a numerical evaluation.
The condition C2>0 (which was independent of the dimen Considered as a whole, the pressure system had a specific inertia
sions of the tank) meant that hg had to be less than 0.5,i.e. that (Tg+ Tc)and a loss of head (h,+hc). If a tank could be inserted
the conduit efficiency had to exceed 213 : it was easily checked at the downstream end of such system, the lowest limit of T,,
that such condition was actually indispensable, since otherwise still compatible with stability would be, according to equation
the derivative dQ&No would be negative, and an increase of (131,
the rate of flow would cause a decrease in the power output. Tg+ Tc
The second and far more important condition 2h,T,,Tg>0
(which reproduced equation (6) of the paper) showed that the
T, = ___
2(h,+hc) . .  * . 
requirement of a minimum tank area emerged from a condition The actual insertion of a tank at an intermediate point,
to be satisfied by the energy contents. In fact, any change of leaving Tgand hg upstream, and T, and h, downstream, required
power output required a change of discharge which, in turn, a larger minimum value of T,, which was to be calculated from
meant a variation of the energies contained in the hydraulic equation (13) :
system. More precisely, an increase of powerand of discharge T g + Tc
too, if C2>Orequired an increase of kinetic energy in the
conduit, and a decrease of potential energy in the tank (owing to
T 
 2hg(l2h,) *  
the friction losses in the tunnel having increased and the water The ratio T,, between values (14) and ( 1 5 ) of T,, coincided,
level in the tank having descended). It was easily seen that the according to definition (9), with the ratio between the areas of
stability condition 2hgT,, Tg>O required that the tank energy the respective tanks, and could be assumed to be an ‘insertion
variation prevailed, so that, in total, the variation of the power efficiency’ of intermediate tank. Such efficiency was given by
delivered by the hydraulic system and that of the energy stored hg
therein were in opposite senses. Such a condition was not
surprising; nothing else was required other than that the pres
70 = (l2hc)h,+la,  * ‘ 
sure system had to possess a ‘hydraulic statism’ of its own, not The insertion efficiency depended only on the distribution of
dissimilar from a prime mover with a direct governor, which head losses between the sections upstream and downstream of
was stable if there was a speed statism causing the total energy the tank and became rapidly lower as soon as the head loss in
(stored as kinetic energy of rotating masses) to decrease if the the downstream section reached the same order of magnitude
power increased. as the upstream one (for example, if hg = ho the efficiency was
lower than 50 per cent).
The above remarks explained the seemingly strange fact that
the stability area, F T h , did not depend on the conduit length. A He had shown (Evangelisti 1949)* that equation (16) held
longer conduit meant a higher value of Tg;.but the increased with a fair approximation in the case of systems of surge tanks
friction losses intensified the stabilizing action of the tank by (the inaccuracy being due only to the degrees of freedom being
the same amount. For the same reason, Thoma’s condition more than one, thus bringing in stability conditions of a higher
imposed tank areas out of proportion if the conduit was short and order). In such a way a simple criterion of equivalence was
the operating head was low, thus justifying the investigations available which reduced any system of conduits and tanks to a
intended to ascertain whether and how far a reduction was simple conduittankpenstock system by successively shifting
possible without running into serious inconveniencies. the tanks along the conduits.
The author had recalled the simplifying assumptions of
Thoma’s treatment. Many of such assumptions could be over Interconnexion of Power Stations. The intense stabilizing
come, with no appreciable complication in the final results, if action of the interconnexions had been referred to by the author
the time constants of the downstream components were small in the classical case of a hydroelectric plant being inter
with respect to the period of the conduittank system (as usually connected with another station free from effects of water inertia.
happened in practice). He considered that a brief r6sumC of It seemed of some interest to point out that, in the more general
such results might prove of some interest. case of both connected stations having a surge tank, the same
action remained, and reached an intensity which perhaps was
unexpected. Omitting the mathematics (though the final
Effect of a Penstock Downstreamfrom the Tank. Let the pen formulae were not at all complicated) the main result was that
stock be of length Lc, with an area fc and a steadystate friction each station acted as a stabilizer to the other regardless of its
loss of head given by Pv2 = P(Qo2/fc2). Introducing the penstock own stability: to such a point that two plants, both widely
inertia and friction loss parameters, Tc and h, : unstable in themselves, might produce a strongly stable system
when interconnected. The effectiveness of the mutual stabiliza
tion was zero if the two hydraulic systems were identical, and
rapidly increased with their diversity.
the coefficients of the characteristic equation (10) take the form It seemed to him that such a property ought to obviate any
Co = T,,{(12hc)Tg2hgTc) concern about the stability of surge tanks inserted in a large
network, since the characteristics of structure and size of the
C1 = 2hg(l2hc)Tw(Tg+Tc) different pressure systems were scattered over a large range of
C, = 12(hp+hc) . . . . . . a (13) values. In such conditions the introduction of a new tank into a
big network was quite likely to improve, rather than endanger,
The discussion of those formulae was easy. Leaving aside the the overall stability.
third coefficient, which merely repeated that the efficiency of
the whole hydraulic system had to exceed 213, the second coeffi The conclusion was obvious, and conformed with the author’s
cient showed that the influence of the penstock was adverse in remarks: the only practical case when a problem of stability
every way :the time Tcsimply added up to p, (showing that the arose was the isolated operation. Such an operation would take
inertia effects in the penstock were the same as those in the place only in special cases of emergency. Notwithstanding that
conduit) but the loss of head in the penstock tended to render or, rather, just because of thathe agreed with the author
inoperative, instead of favouring, the stabilizing action of the that the isolated service had to be taken into account. But an
tank. In general, T , and h, were small, so that the penstock effect isolated service might be admitted to be somewhat imperfect in
was also small. But sometimes Tc and h, were of the same order regard to the constancy of voltage and frequency (and, conse
of magnitude as T, and hg, and in such cases Thoma’s original quently, that of the output power). Therefore, the question arose
formula became inadequate. of how, and within what limits, such imperfections could
improve stability.
Since the loss of head acted in favour or to the detriment of
stability according to whether it was located upstream or down * EVANGELISTI, G. 1949 ‘Adduzioni in pressione e stabilitl di
stream of the tank, the recommendation to install the tank as regolazione’ (Societl Italiana Progress0 Scienze, Rome).
C O M M U N I C A T I O N S ON PRESENT TRENDS I N SURGE TANK DESIGN 113
Regulation with Variable Hydraulic Power. All the factors It was well known that the governor of a turbine contained a
(interconnexions, variations of efficiency of turbines and genera stability reserve, by which it could master the reaction by the
tors, load characteristics of the network, etc.) which caused the penstock (which was nothing more than a conduit with no surge
power to deviate from perfect constancy did not alter the degree tank at all). A general theoretical study, carried out by him since
of the characteristic equation, which therefore kept the same 1947, had resulted in the conclusion that, as soon as the rapid
form (10). It could be shown that the coefficients of the equation governing disturbances had died away, the governor was again
then took the form : ready to act as a stabilizer on the slow oscillations of the conduit
Co = T,{(1 2n,hc)T,2n,h,Tc}; tank system.
+
C1 = 2h,( 1 2n,h,) Te,ny( Tg Tc); The influence of the governor on such oscillations emerged
from three factors, which were [in order of increasing im
C2 = 12nY(h,+h,) . .
. . . . . . (17) portance) : ( 1 ) the permanent statism, ( 2 ) the governor stabilizer
where all (and only) the negative terms were multiplied by a (temporary statism or accelerometer), and (3) the slowness of the
factor n,. gate movements.
In regard to the constant power, the stability became worse Though there were large differences in different cases, the
or better according to whether n, was greater or smaller than average value of the stabilizing action by the governor might be
unity. When n,. was equal to, or smaller than, zero, any tendency taken as being equal to some seconds :that meant that a positive
to instability disappeared. term of some seconds was added to the term 2hgT, in the second
He gave a general procedure for determining the expression coefficient of the characteristic equation to counteract the nega
of n,. For instance, in the case, dealt with by the author, of an tive terms due to the water inertia. Such a contribution could be
interconnected hydroelectric plant taking over a fraction K of rather inconspicuous for the long conduits, but become signifi
the total load, the value of n, was given by cant for the short ones.
3k 1 The theoretical results were being controlled by a new pro
n, =  gramme of experiments being carried out on the same plants
2 .
of Societa Adriatica di Elettricita quoted by the author. The
In regard to the isolated service, it could be shown that the results of those experiments were awaited, and would be used to
most important factors affecting the variations of the output check the accuracy of the theoretical investigations.
power were the turbine efficiency ?the efficiency of the pen
stock being already taken into account by equations (13)and Mr. JOSEP FRANK (Erlangen, Germany) wrote that modern
the statism of the regulation. In such a case the expression of stability research had in essence confirmed the conclusions of
n, becomes Thoma (1910), in so far as they dealt with the hypothetical case
1 +s,
1 + 1[ H arl of the perfect governor. It had to be considered as a rare piece
of good fortune when a theory of more than forty years old was
n, = ?1
I r 27
1
. . . (18) proved to be correct after much doubt and critical examination.
However, it had been shown that Thoma’s theory applied only
to a certain limit. Modern hydraulics were still based on Thoma
Such a formula contained’the two partial derivatives of the owing to the fact that the majority of actual plants lay within
efficiency with respect to head and discharge, both to be that limit. In Fig. 22 were compared the results of Thoma and
calculated at the steadystate pointhe did not agree as to the
0 3 0 H
convenience of considering the efficiency, for the sake of
simplicity, as a function only of the power, because such a posi
tion was incorrect and the results obtained were largely in
accurate. The symbol s represented the statism of the electric
network, and had a value depending upon the character of the
network itself (including the tension regulation). Usually, the 0 20 
value of s varied between 1 and 3 ;6, represented the mechanical
statism of the turbine governor.
Considering the efficiency effects alone, formula (18) showed P 
that the value of n, became larger than unity if the turbine
efficiency improved with increasing head (a?/aHo>0) and
decreasing rate of flow (a?/aho<O). Therefore, when the steady 0.10
state head and rate of flow were, respectively, lower and greater
than those of maximum efficiency, the stability was less good

than at a constant power.
That fact was worth pointing out, because such steady states
had the maximum values of the specific inertias Tg and T,, and
showed the greatest tendency to instability even at a constant
power.
On the other hand there came into consideration the stabilizing
action of the statism, which depended on the product sS,.
Numerical calculations showed that the greatest values of s6,
still permissible in an isolated service (for instance, s6, = 0.20)
could always eliminate the adverse influence of the turbine
efficiency. The surplus, however, was normally small, so that it
could be concluded that the original assumption of constant
power (ny = 1 ) resulted correctly in practice, being the result of a
balance between the worst possible operating conditions and the the stable and unstable regions were reproduced as given by
full use of the possibilities allowed by an isolated service. Thoma [1910), Frank [1942), Jaeger (1943), and Evangelisti
(1951)$. It had to be noted that Jaeger’s formula, in the afore
Effect of the Turbine Governor. It was not possible to mentioned notation, took the form
consider the effect of the governor on the conduit surge system B
in detail without excessively prolonging his communication,
1
p
 
which had perhaps exceeded the due limits already. A very brief
outline would be enough. (Detailed information was to be found t FRANK,J. 1942 Bauingenieur, vol. 23, No. 21/22, p. 149, ‘Zur
in Evangelisti 1953*.) Stabilitat der Schwingungen in Schachtwasserschlassern bei un
veranderlicher Leistungsentnahme’.
* EVANGELISTI, G. 1953 Energia Elettrica, vol. 30, p. 137, ‘Pozzi $ EVANGELISTI, G. 1951 Energia Elettrica, vol. 28, No. 12, p. 637,
piezometrici e stabilitii di regolazioneSeconda Nora’. ‘Sopra la stabiliti delle grandi oscillazioni nei pozzi piezometrid’.
114 COMMUNICATIONS O N PRESENT TRENDS I N SURGE TANK DESIGN
and that Evangelisti’s formula, for the case of load enlargement which lay between FTh and FE,endeavoured to reach a standing
from zero to full load (n’ = Q*/Q = 0), was expressed in that oscillation with reduced amplitude as was shown in Fig. 23.
notation However, the analytic solution of that problem was difficult
E = 1 P
~{l/(4+12j9115/32)+3P2} and was only possible with approximations. The results had
on that account to be scrutinized with the aid of the stepby
313 step method for which a process of second approximation was
Evangelisti’s theory was also easily applicable to the case of load to be recommended asthe usual methods of first approximation
enlargement from half to full load (n’ = 03), giving were too inexact. He himself had examined a particular case
E = $& d(64 488 +
15P2) 913 8 )
using such a process of second approximation and thereby
obtained the result shown in Fig. 24, and notwithin the
examined interval of timethe standing oscillation according to
The Thoma condition was given by Escande. Irrespective of the final solution of that problem which
E = 
2 * (1PI still required research, in every instance stress had to be laid on
B the author’s remark that the Thoma crosssectional area was to
be recommended also for throttled surge tanks, if oscillations
, li
+ 10M.
only from a physical viewpoint but probably also for certain could give some exact information on that subject concerning
limiting cases even if, according to Evangelisti, the charac certain calculations where they had obtained different results
teristics of governing mechanisms and of the supply network from those given by him. An interesting exchange of views might
had especial significance for stability. be made on that question. Meanwhile, they appended a certain
In the clarification of the stability question with throttled number of the results they had obtained, and, in part, the sup
surge tanks, Escande (1951a and 1951b)* had recently met with porting calculations which corresponded to them.
success, He had given a limit of crosssectional area of As would be seen, their remarks did not in any way constitute
a criticism of the author’s paper; and they wished to emphasize
that their comments referred only to a few welldefined points
in that paper.
Their results, which related to the stability of an installation
and had ascertained that the movement in a surge tank of area F,
t RAMPONI,F. 1946 Energia Elettricu, vol. 23, No. 3, ‘Sulle
* ESCANDE,
L. 1951aLe Gknie Civil, vol. 128, No. 12, p. 227, ‘8tude oscillazioni nei pozzi piezometrici per manovre alteme ripetute’.
de la stabilitk des chambres d’tquilibre A ttranglement’. $ FRANK, J. 1953 Die Wusserwirtschuft, vol. 44, No. 1, p. 18,
1951b Le Gknie Civil, vol. 128, No. i3,.p. 254, ‘8tude com ‘Rhythmische Belastung und Entlastung in der Wasserschloss
pltmentaire de la stabilitt des chambres d’tquilibre A ktranglement’. berechnung’.
C O M M U N I C A T I O N S ON PRESENT TRENDS I N SURGE TANK DESIGN 115
with two surge tanks situated upstream of the turbines, would The hypotheses were the same as those given above. The
be set out under the three heads of Hypothesis, Results, and conditions of stability were finally found to be
Supporting Calculations. F1> R iFThl
Hypothesis. (1) It was supposed that efficiency of the tur F2 >R ~ F T ~ ~
bines was constant in the area of adjustment around each point where FlyF2 were the sections of upstream and downstre?m
of operation. surge tanks, respectively; FTh,, F T were~ ~ the correspondmg
(2) It was supposed that the governor maintained an absolutely Thoma sections calculated with the net head and considered
constant power in the period. otherwise as if there had been separate installations; and R1,Rz
(3) Equations of the oscillating phenomena were linearized, were coefficients which were a function of a certain number of
i.e. it was supposed that the oscillations had fairly low amplitude parameters but which, in practice, never exceeded 1.2.
I I I

=
 SI 4.HO


 s,
so that the governor operated in the linear part of its charac With reference to the section of the paper (p. 100) concerning
teristic and so that the curvatures of the head loss and power 'Interaction of Surges, Water Hammer, and Turbine Govern
characteristics would be negligible. ing', they wished to bring to the author's attention a paper by
(4) I t was supposed that the head losses under permanent Meyer and others (1952)f in which they had proved that for the
regime were low in comparison with the net head. characteristic equation
Results, Stability of installation having two surge tanks, one Curn+Cnlrn*+ ... +Clr+Co = 0
of them, surge tank B, being just upstream of the turbines and the conditions Co>O : C1>0 : C2>0 and, more usually, Cp>O
the other, surge tank A, farther upstream. were always necessary but were never sufficient to ensure stability.
(1) When calculating the Thoma section of surge tank B They had also studied a practical case for which the group of
taking into account the pipe situated between the two surge conditions C,>O : C1>O : Co>O gave an absolutely Merent
tanks, the condition result from that of the more exact calculation.
F2 >F T ~
was found to be insufficient, where F2 = section of surge tank Supporting Calculations. In the case of two surge tanks
B, and FTh = Thoma section. situated upstream of the turbines (Fig. 25), the following were
In addition, it was necessary that the installation should the equations of the phenomenon :
satisfy the condition
72Ph20 = 440 . . . . . . (23) If it were supposed that the section of the surge tank situated
Pso = 430440 . . . . . (24) just upstream of the turbines satisfied the Thoma condition for
+
h20 450 = 0 . . . . . . . (25)
the section of the pipe just above the turbines, then :
The characteristic equation in ‘p’ was sought. q,? was extracted 8272 = 92
from equation (19) and q20 from equation (20) giving equation The characteristic equation then became :
(21) as:
1
71T2e1e2P4+P3P1 71 7282 +p2[72e2 + +
7181 82)]
7181( 1
sin aT
i
cos cos p sin
T” a y
27rL 2nL
in material readily available to the British engineering profession. C O S ~
aT“ +pz sin 
a T
Mr. GEORGE R. RICH(Boston, Massachusetts) wrote that the in which h = the relative head, (HHo)/Ho;p = aVo/2gHo;
author had presented a valuable summary of the present state of a = waterhammer wave velocity, ft. per sec. ; qomax = peak
development of surge tank design. His endorsement of the value of relative gate opening change, (QQO)/Qo, or ampli
Thoma formula confirmed the experience of himself that when tude of assumed sinusoidal relative gate opening oscillation
that criterion was increased by the proper allowance to provide (dimensionless); x = distance along pressure conduit taking the
suitable damping, it still remained a most useful device for origin at the reservoir x = 0 (at the turbine x = L ) ;T“ = period
gauging stability. In the United States, the general practice was of oscillation of the governor, seconds ;c = generaltimevariable,
to increase the Thoma tank diameter by 25 per cent for differen seconds.
tial or restrictedorifice tanks and by 50 per cent for simple If the surge tank were located over a node in the standing
tanks. wave pattern of water hammer established along the conduit by
The effect of the various departures from the conditions con an oscillating governor, the tank would be of no benefit.
templated by the original Thoma derivation could be very The fundamental equation showed that under the assumed
conveniently evaluated by the formula given in Stucky (1951)* : forced oscillation of the turbine gates, nodes would occur at
values of x (measured from the reservoir as origin) equal to 0,
a T / 2 , a T , 3aT”/2, 2 a T , 5 a T ” / 2 . . . naT/2. Maxima and
minima might be expected at x equal to a T /.4 ., 3 a T / 4 , 5 a T ” / 4 ,
7 a T / 4 . . ,(2nl)UT”/4.
in which, as noted by the author, If T were taken as 5 seconds and a equalled 3,000 ft. per sec.,
dN N the first node would occur at a distance x = aTn12 = 7,500 feet
h== from the reservoir. Consequently, if the penstock had an
dNs Ns additional length of, say, 500 feet, a surge tank located 7,500 feet
measured the effect of station interconnexion. from the reservoir or 500 feet from the turbine would be of no
p might for all practical purposes be taken as the relative benefit in reducing water hammer caused by sustained oscilla
decrease of each positive surge amplitude as compared with its tion of the governor or in preventing such waterhammer
immediate predecessor. /3, in other words, measured the rapidity pressure from entering the tunnel. That numerical example
of damping : illustrated the manner in which that equation might be used to
EfiEn+ 1 detect unfavourable locations for the surge tank. The resulting
8 = En pressures resulting from assumed periods of forced oscillation
p was defined as followed : The turbine performance curves of the governor and their accompanying assumed maximum
were plotted using relative efficiencies as ordinates and relative amplitudes of percentage gate opening might readily be cal
gate openings as abscissae. The point on the turbine curve at culated for any location on the pressure conduit by means of
which tank stability was being investigated was taken as par or the same formula.
unity in calculating the relative values. A tangent to the per In connexion with systems of surge tanks he recalled with
formance curve was drawn at the point under investigation. interest the initial arithmetical integration studies for the Fort
p was the angle which that tangent made with the horizontal or Peck surge tank in which a controlgate shaft of quite appreciabl
axis of relative gate opening. capacity was located at about onethird the length of tunnel from
In addition, g was the acceleration of gravity, ft. per sec. per the reservoir. The differential surge tank was designed to have
sec. ; c was the coefficient in the expression for tunnel friction 125 per cent of the Thoma value for the entire conduit neglecting
head loss, head loss (feet) = cVo2; c did not include conduit the controltower cylinder, but fear had been entertained that
velocity or Venturi velocity head; Ho was the steadystate head the controltower oscillations would penalize operation of the
on the turbine not including penstock friction head loss (feet); main differential tank. The analysis showed in that particular
A was the tunnel area, sq. ft.; L was the tunnel length, feet; instance that the effect on tank levels was not appreciable,
cwwas the coefficient of velocity head loss in both the tunnel and although the capacity of the control tower was sufficient to cause
the Venturi contraction (if any), velocity head loss (feet) a definite difference in tunnel velocities upstream and down
= cw(Voz);Co was the friction head loss in the penstock, feet; stream of the tower. First the downstream velocity would be
F was the surge tank area in sq. ft. greater and then the upstream velocity would be greater resulting
As noted by the author, the benefit resulting from inter from pendulation of the controltower water level.
connexion might be very great. The influence of p was to give In connexion with a large differential surge located at the end
improved stability at the lower gate openings and decreased of a long tunnel which supplied water for both electric power
stability on the drooping portion of the turbine performance generation and municipal water supply, the question arose as to
curve in the vicinity of full gate. The inclusion of velocity head the necessity of plugging two construction adits located at about
loss was to improve stability, particularly in short tunnels ;but, the third points of the tunnel. The diameter of the adits was
as the author had wisely cautioned, too much reliance ought not about equal to the diameter of the pressure tunnel. Since the
to be placed on very substantial Venturi contractions. As would differential tank was given 125 per cent of the Thoma value, its
be apparent from the formula, the effect of penstock friction stability was assured and the construction adits were left open.
was to decrease stability, although that penalty would not usually Because of the limited capacity of the adits, it was estimated that
be great in economically designed penstocks. An allowance of their effect in producing velocity differentials in the three
25 per cent over the boundaryline Thoma value of diameter sections of tunnel would not be important, and that their
for the differential or throttled tank would usually be found general effect would be slightly beneficial but certainly not
ample to cover all the elements in the expanded Thoma criterion. detrimental to the performance of the main tank.
The author had mentioned the interesting possibility that, in
certain cases of resonance, the surge tank became ineffective as Mr. P. 0. WOLF,A.M.I.C.E. (London), wrote that there was
a means of preventing water hammer from entering the tunnel. a fundamental difference between the function of a surge tank
Using the basic theory of Gaden (1945), he himself had developed (or air vessel) in a pumping main and that in a tunnel or pipe
line feeding a turbine. In the pumping main, the chief purpose
* STUCKY,A. M. 1951 ‘Chambres d‘Equilibre’, p. 78, equation (59)
(Ecole Polytechnique de 1’Universiti de Lausanne, LausanneMulti t RICH,G. R. 1951 ‘Hydraulic Transients’,p. 65, equations (3)(21)
Office Machtzum, Lausanne). (McGrawHill, New York and London).
118 C O M M U N I C A T I O N S O N PRESENT TRENDS I N SURGE TANK DESIGN
was to avoid serious disturbances due to shutting down of the fulfilled, very considerable errors might occur when the head
pump, whereas in a waterpower plant a surge tank was neces race was short.
sary as a reservoir to permit a sudden increase in flow to deal As far as concerned stability, he considered that it would be
with a sudden increase in load. Following a sudden reduction in of great interest to be able to set down the limit amplitude of the
load, it was not essential to cut off the flow abruptly. In impulse residual oscillations whenever those latter existed. For that
wheels it was usual to provide jet deflectors or dispersers which purpose he suggested a formula deduced from considerations
allowed the power input to the turbine to be cut in a very short which, although schematic, were quite accurate to a first approxi
interval of time; thereafter, the flow in the pipeline could be mation. That formula gave good orders of magnitude when
throttled as slowly as desired, by the valveclosing mechanism applied to Partidor’s Italian tests. It ought to be noted that, in
which ought to be designed for that purpose. In the case of that case, another similar formula, set forth elsewhere, would
reaction turbines, bypass or pressurerelief valves were often have given infinitely large oscillations.
provided to permit the turbine gates to shut rapidly without any In point of fact, he considered that that problem was too
sudden reduction of velocity of flow in the pipeline ;again, the complex to be dealt with by one single formula. The latter would
rate of closure of the pressurerelief valve could be arranged so enable it to be only roughed out to start with and the final
as not to excite heavy excess pressures. The waste of water due solution would have to be found by stepbystep integration. It
to a slowing down of the speed of valve closure was negligible. ought to be mentioned that that might be done on a scale model
The author in Fig. 3 had illustrated very clearly the advantage with Durand similitude provided that the boundary conditions
of a large surgetank area in the maintenance of the stability of were correctly reproduced. The penstocks, the turbine efficiency
operation of a turbine. On increasing the flow to the turbine it curves, etc., might thus be taken into account.
was evidently desirable to avoid a great reduction in net head. That type of investigation was of particular interest in the
Conversely, on decreasing the flow to the turbine, excessive case of large power plants operating under low and medium
upward swings would be avoided if some of the flow was again heads, for it was with such plants that the old Thoma formula
bypassed to the tailrace. might give immense crosssections.
The author had stated (Jaeger 1949a) that a purely hydraulic Some authors intimated that if the governor parameters were
surge system in which the flow was proportional to some positive correctly modified, stability conditions could be considerably
power of the net head (usually H*) must always be stable. improved. Since that necessarily resulted in a decrease in govern
Such conditions might be approximated, if water were fed ing quality, he was of opinion that, whenever highquality
from the reservoir in the surge tank into the pipeline on the governing was required, it was safer not to rely entirely on the
downward swing (possibly through a system of large flaps or governor for the stability of the hydraulic layout, and that, as far
other nonreturn valves), and if the bypass valve were opened as possible, that hydraulic layout ought to be stable in itself.
on the upward swing. That arrangement amounted, in effect, to As to the most unfavourable manoeuvres, he did not think
a superposition of two opposing oscillations. That such a super it necessary to mention further the case of throttled surge tanks
position could be successfully achieved in a single period had but would merely emphasize the fact that, at Grenoble, use was
been shown by Mr. Morgan in his diagram of the oscillations made of a simple graphical method which gave the most
in the Lochaber surge system (Fig. 19a) following shutting of dangerous manoeuvre law and the consequent oscillation for a
the valves in two carefully calculated stages. given discharge variation.
The conclusions of Ghetti and Evangelisti that the surge
stability of a system might be affected by the sensitivity of the
governor had again been stressed in the paper. That aspect was AUTHOR’S R E P L Y
doubtless being studied by governor designers. No reference had Dr. C. JAEGER wrote in reply that the discussion and com
been made, however, to the possibility of providing an electrical munications had produced a considerable amount of new
load to assist in maintaining hydraulic stability. That load material, some of which was of great importance. A few days
b h i c h might be provided by a resistance tank) would be the after the presentation of the paper to the Institution, the French
electrical equivalent to the bypass valve or jet deflector, and technical journal La Houille Blanche had published an issuet
would be brought into the circuit during the upward swing of which had been devoted nearly entirely to the problem of
the mass oscillation in the surge tank. Again, there was no surges. Several of the authors who had contributed to that issue
solution there to the problem of a sudden increase of electrical had also participated in the written discussion of his paper,
demand which could not, at the time, be met out of electrical developing new theories and producing new results.
storage, but only by drawing water from the reservoir in the He (the author) had been informed that important research
surge tank. After the first downward swing, however, the use of work was proceeding at the Laboratoire d‘Hydraulique de
an electrical resistance could in one period lead to steady l’Ecole Polytechnique, University of Lausanne, Switzerland,
hydraulic conditions. and he had paid a visit there. The findings of that research
Mr. B. ZICMAN(Grenoble) wrote that the paper was of laboratory had been summarized in a letter addressed to him
definite interest to all who dealt with such calculations, for, with dated 29th March 1954, and he had been given permission to
characteristic clarity, the author had set forth present trends and mention it (see p. 120).
given many bibliographical references on the subject. Discussion of the paper had centred on three main points,
He and his colleagues had, themselves, had occasion to deal namely : the basic assumptions for the design and calculation of
surge tanks ;the problem of turbine regulation; and the reflection
with such calculations and he had published a paper (Zicman of waterhammer waves in pressure systems provided with a
1953)* in which a certain number of conclusions, drawn from surge tank.
experience, had been given. Some of those conclusions con A considerable amount of agreement had been reached on
cerned subjects dealt with by the author and he would like to many points under discussion*.
mention them, especially since the article concerned had been He proposed to treat his reply under five main headings,
published only a few days after reception of the author’s paper. namely: Basic Assumptions for the Design and Calculation of
The dynamic equation of the volume of water in the headrace Surge Tanks ;Problem of Stability ;Water Hammer in Hydraulic
tunnel led to two terms usually Iaown as P and R. In his Systems Provided with Surge Tank; Surges in Underground
opinion, it was extremely important that the gathering of Power Stations; and Model Tests on Surge Tanks.
velocity

W2
Basic Assumptions for the Design and Calculation of Surge
2g Tanks. Contributors such as Mr. Headland, Mr. Morgan,
should always be found in those two terms instead of solely in Mr. Elsden, and Professor Evangelisti had agreed on the im
the endeavour to stabilize the installation with a Venturi tube. portance of a thorough examination and discussion of the worst
Experience had indeed shown that if that condition was not
t L a Houille Blanche, vol. 8, No. 5, October 1953.
* ZICMAN,B. 1953 L a Houille Blanche, vol. 5 , p. 579, ‘Mkthodes $ For more detailed mathematical treatment of surges and water
ncuvelles Four le calcul des cheminkes d’kquilibre’. hammer see Jaeger (1949a).
AUTHOR’S REPLY ON PRESENT TRENDS I N SURGE TANK DESIGN 119
possible flow variations to which the system would be subjected. Rich had written that according to American practice that safety
Some ideas which a few years previously had been considered as factor ought to be n = 1.5 for ordinary tanks and 1.25 for
being merely of academic interest were currently being discussed restrictedorifice or differential tanks. European practice mostly
and approved by engineers concerned with practical design. took n = 1.71.8. Mr. Elsden had proposed n = 2. He himself
Mr. Headland and others wanted repeated loading and suggested that much depended on the friction (losses in tunnel)
unloading of the turbines to be considered. That proposal was which was being assumed in the estimate of the limit value FTh,
in line with the conclusions of some Italian experts he himself since FTh was inversely proportiond to the assumed losses and
had mentioned in the paper. When the Innertkirchen surge tank was, therefore, proportional to the square of the frictionloss
had been tested on model, alternative loading and unloading factor. Furthermore, he proposed to refer to the actual decrement
had been included in the test programme. The problem of d = ( Z ~ + ~ ) / Zof
; the oscillations (see Jaeger (1949a) (French
repeated loading and unloading was also the subject of a recent translation (1953), pp. 2535)).
article by Frank (1953)* who had developed some earlier Professor Escande had mentioned a theory he had developed
calculations by Calame and Ghizzetti (1950)t. The case of faulty (Escande and Huron 1953)g on the stability of two surge tanks
synchronizing, so dangerous for a single machine station, came built respectively into the headrace and tailrace tunnels of a
into that more general analysis. hydroelectric power station provided with reaction turbines.
Professor Evangelisti had pointed out the danger of alterna The same problemwhich was typical of problems which
tive closing and opening movements for restrictedorifice t a n k s , occurred with underground stationshad been dealt with by
since the pressure rise or drop at the orifice varied with the Mr. Meyer (1953)s.
square of velocity through the orifice. Mr. Headland had men The method used by both scientists was the same and
tioned the case of pump storage and the case of Etzel power consisted in linearizing the differential equations and discussing
station (Switzerland) where the turbine and the pump were the roots of the joint characteristic equation. Their conclusions
combined to work together, and he might be interested to know were very similar and to a wide extent confirmed the results he
that some years before Etzel, similar cases had been con (the author) had arrived at several years previously using a
sidered of rapid switching from turbine to pump, and vice versa, somewhat different method. It had been well proved (Jaeger
and had been exhaustively analysed for the pump storage station 1943, 1949a (French translation (1953), p. 245)) that two surge
of LacBlanc LacNoir in the French Vosges. tanks located on both sides of a reaction turbine reacted one
Professor Escande and Mr. Zicman had mentioned the case on the other and that the areas given by the formula of
of the throttled surge tanks, where slow gate movements could Thoma had to be increased. I t might be stressed that all the
be more dangerous than rapid gate movements. contributors had agreed that surging and damping out in such
At the very moment when so many were agreed on more intricate systems had finally to be checked by a calculation
liberal assumptions for basic surgetank calculations an opposite point by point. Fig. 26a reproduced from Jaeger (1943) showed
tendency was being revealed. He had mentioned in the paper the results of such a calculation. In introducing the paper he had
some new proposals concerning basic assumptions for surge mentioned different types of unstable oscillation and Fig. 27
calculations as put forward in an internal memorandum by was given, which supplemented Fig. 15.
Electricit6 de France. In his comments, Mr. Meyer had discussed the somewhat
A similar tendency appeared among engineers in charge of the similar problem of the stability of two surge tanks built in series
design of extremely large power stations concentrating about on the headrace pressure tunnel of the hydroelectric power
1,000,000 h.p. or more on one surge tank with anything from system. The method of approach he used was the same as the
six to twelve turbines, if not more. When examining the trans one used for two surge tanks located one on either side of the
mission system connecting those stations to the existing and to reaction turbines. He had subsequently learnt from corre
the future grid, it was hardly thinkable for the whole station spondence with Mr. Meyer that he proposed to compare his
to be reconnected rapidly to the grid after having been thrown results closely with those of Ghetti (1947b) and Straubel(1943)Il
out. and it is to be hoped he will allow his findings to be published
T o adhere to the design assumptions as described in the soon.
previous paragraphs would lead to surge tank dimensions which The problem of the effect of the velocity head, w02/2g,on the
it would be hardly possible to construct. In such exceptional stability had been dealt with by Mr. Zicman, Mr. Rich, and
cases the basic assumptions for the calculations might have to Professor Escande. The importance of that factor as a method of
take into account : adding to the stability of a surge tank by purely hydraulic means
(I) The closing down of the whole station after a short had not been overlooked by him :he agreed that the formula of
circuit and the dangerous negative surge which followed the Calame and Gaden (1927) was too optimistic.
first positive surge. The formula produced by Professor Escande was more reliable
(2) The sudden automatic opening of all the turbines from, and confirmed those of Stucky (1951)q mentioned by Mr. Rich.
say, 70 or 75 per cent of load to full load. That might happen He awaited with interest further publications (not yet available
when the water level in the tank was still surging up and down, but announced in the comments) by Professor Escande on the
owing to previous movements of the gates, thereby producing stability of surge tanks provided with a Venturi which increased
surge interference. the effect of w02/2g.He had been informed by Messrs.D. Bonnard
(3) The conditions created in the station when the next and A. Gardel (Lausanne) that they would concentrate on that
most powerful station on the grid was accidentally dis very important problem too, and expected to solve it by means of
connected (short circuit) after having been successfully model tests. The case of the value of w02/2g when the surge tank
synchronized and connected. was on the downstream side of the turbines had not been dealt
with by those authors and still remained to be solved.
The contribution by Professor Evangelisti who for years had
Problem of Stability. The contributions by Professor Escande, specialized in the stability problem was of great value. The pen
Professor Evangelisti, Mr. Meyer, Mr. Zicman, Mr. Rich, and stock inertia and friction losses which he (the author) had
Messrs. Cuenod and Gardel had been mainly concerned with previously mentioned as being detrimental (Jaeger 1949a
the stability of surges in surge tanks. The importance of those (French translation 1953, pp. 2479); Stucky 1951; Rich 1951)
contributions was a measure of the importance attached to the were being considered by Professor Evangelisti, and their
subject. It was also well worth noting that most of those con effect was shown in a simple formula. The wellknown formula
tributions had a direct or indirect bearing on the theory of
underground stations, a fact which showed the importance of $ ESCANDE, L., and HURON,R. 1953 La Houille Blanche, vol. 8,
that subject. The problem of safety factor n = F / F T where
~ n No. 5.
was the ratio of two areas, had been mentioned by several. Mr. MEYER,R. 1953 La Houille Blanche, vol. 8, No. 5.
// STIZAUBEL,1943 Wmserkraft und Wmserm’rtschaft,vol. 38, No.
* FRANK,J. 1953 Wasserwirtschaft,vol. 44, October, ‘Rythmische 10, ‘Zur Theorie gekuppelter Wassersch6sser’.
Belastung and Entlastung in der Wasserschlossberechnung’. s[ STUCKY, A. 1951 ‘Cours d’amknagement des chutes d’eau;
f GHiZZETTi, A. 1950 ‘Oscillazioni dell’acqua in impianti idraulici Chambres d’equilibre’, p. 78, formula of Gardel (Ecole Polytechnique,
dotati di pozzi piezometrici’ (Rome). Lausanne).
120 AUTHOR'S REPLY ON PRESENT TRENDS I N SURGE TANK DESIGN
Fig. 26a. Oscillations in the System of Two Surge Chambers Shown in Fig. 29b
Case I I1 I11 IV

L1, metres . 6,330 6,330 6,330 6,060
&,metres . 111 111 111 270
f1, s q . metres . 7.07 7.07 7.07 7.07
f2, s q . metres . 10.75 9.52 9.52 8.55
F1, sq. metres . 15.95 19.64 19.64 44
Fz, sq. metres . 28.15 35.25 35.25 19.64
If, metres . 261.47 273.25 253.25 261
Qo,cu. metres per sec. . 11 16.75 24 21
dn12 1.67 2.12  2.12
n h * 0.83 1.25 1.12 2.56
I
Downstream surge tank.  Upstream surge tank.  Variations in gross head.
of Calame and Gaden (1927) on the effect of the turbine According to Professor D. Bonnard the following results had
efficiency was completed by including the differential d.1/dQo, been obtained :
in addition to dq/dNo. Professor Evangelisti had also pointed
out and emphasized the stabilizing effect of the connexion to a (1) It had been demonstrated that on a model it was
large grid and had referred to his work (Evangelisti 1953) on the possible to study with great accuracy the conditions of stability
effect of the turbine governor on surge stability. He had an of the surges in a surge tank, on the supposition that the
nounced further tests which would be carried out on the Italian turbines were governing for constant output and the power
hydroelectric power stations of the Societh Adriatica di Elet station was isolated and not connected to other power stations.
tricith.
6o SECONDS 9o I20 I so
30
30
7 R
I
10
10
9
E
L 0
t?
I
640
PRESSURE SHAFT
HORIZONTAL SECTION C C
.~
Fig. 30*. Surges and Water Hammer for Total Closure in 3 seconds
f =.31.1 sq. metres.
Q = 100 cu. metres per sec.
a = 900 metres per sec.
T = 3 seconds.
A t = 6 seconds.
_
 Graph of surges. Water level in secondary surge tank.
~ Pressure below restricted orifice. __ Curve of capacity of surge tank.
 Water level in shaft.  Capacity of secondary surge tank.
The same problem had been dealt with by Professor Escande flow downstream tunnel, large enough to cope with any trans
who had mentioned several previous publications and by Dr. latory wave with no downstream surge tank was the right
Zienkiewicz in connexion with model tests carried out at solution. Whenever possible, the solution with freeflow tunnel
Edinburgh University. and no surge tank ought to be adopted.
Mr. Rich had mentioned the case of resonance of water The decision might be more difficult when a highhead station
hammer waves and had given a formula he had published in 1951. was provided with a very long tailrace tunnel.
He himself had devoted some of his efforts to investigating the In such cases a downstream pressure tunnel with a down
different types of resonance of pressure waves (Jaeger 1936t, stream surge tank might be less expensive than a long freeflow
1939) and had published (Jaeger 1948,1949a (French translation tunnel large enough for translatory waves to pass freely. In
(1953), p. 322)) particulars of one case where resonance of that estimating the cost of excavation for the two alternatives it was
type (eleventh harmonic) had been produced causing very severe necessary to consider that excavating a unit volume of rock in a
damage to an existing tunnel. In several instances he had found surge tank was more costly than excavation for a long tunnel.
that resonance was due to an upper harmonic of the penstock The cost of the concrete lining had to be included. The type of
of the pressure tunnel. turbine (if impulse or reaction wheel) might be of importance for
the decision, especially if governing and stability conditions had
Surges in Underground Power Stations. Several contributors to be considered. The stability conditions might determine the
had mentioned the problem of surges in underground stations horizontal area of the surge tank.
and his own views had been requested as to when a surge tank In recent times several projects had been developed showing
on the downstream side of the turbines was indispensable, and an upstream and a downstream surge tank on both sides of a
when it was not. reaction wheel. Escande (1953)$, Meyer (1953), and he himself
In cases like that of Innertkirchen with highhead impulse (1943) had dealt with the difficult problem of the stability of
wheels, a relatively short downstream tunnel of about 1,400 such a system when oscillations in both tanks were likely to
metres and no major floods in the river at the downstream end interfere.
of the power system, there was no problem whatsoever; a free With a lower head and higher discharges the cost of a tailrace
tunnel able to face large translatory waves became prohibitive.
* Figs. 28, 29, and 30 are reproduced by courtesy of La Houille In many instances a stable surgetank chamber was found to
Blanche from BOWARD,M., and MOLBERT, J. 1953, p. 260, Figs. 1 be not less expensive, and the socalled ‘partialacting surge
and 11, and Diagram 1.
t JAEGER, C. 1936 La Houille Blawhe, MayAug., vol. 35, pp. 71,
97.
* ESCANDE, L. 1953 ‘Methodes nouvelles pour le calcul des chambres
d’equilibre’ (Dunod, Paris) ;1953 La Houille Blanche, vol. 28, No. 51.
10
124 AUTHOR’S REPLY ON PRESENT TRENDS I N SURGE TANK DESIGN
tank’ might be the right answer. That type of surge tank had output. In Grenoble a system with sliding scales had been used
first been proposed by himself for the Wettingen power station successfully. The method used at Lausanne had already been
and extensively tested on a model before finally being adopted dealt with at length above. There had been great expectations
(MeyerPeter and Favre 1932)*. In all those instances the final of electronic computation machines and electrical analogy
decision would probably depend to a large extent on costs and (Paynter 1953)t. Those might provide a useful aid and relief to
comparative economics, the cheapest solution being favoured. designers but did not solve all their main problems. The
In recent years, difficult designs had had to be considered, hydraulic designers were most interested in the effect of turbu
where the main object of the designer had not been to achieve lent flow, in additional losses in bends and in the losses at the
economy but stability of the surges. With heads of about 100 junction of the surge tank with the pressure tunnel, in the
metres or less and the large discharges now usual in big projects, additional effect of the mass of water and the friction losses in
a stable downstream tank might be too costly or even impossible the surge tank, and in orifice losses for restrictedorifice tanks.
to design. Systems entirely under pressure with no surge tank at All those questions could be solved on hydraulic models only,
all would then be considered as possible alternatives. But there sometimes with the help of aerodynamic model tests, but hardly
were two warning signals to the designer: the governing con with the help of electrical analogy.
ditions and the danger of cavitation for steady and for unsteady Extremely valuable information had been forwarded by Dr.
flow. Zienkiewicz concerning tests on water hammer in systems with
That short review of the main types of underground power surge tanks, a problem to whose theoretical solution he himself
station in relation to the surge tank problem showed that there had devoted much of his time.
was no rule of thumb for deciding when a surge tank was in There was an obvious misinterpretation of his statement
dispensable and when it was not. There were many contradictory concerning slow surges which were not to be interpreted by
conditions to be satisfied and in difficult cases many alternative the waterhammer theory. That statement should refer to
solutions might have to be investigated before a decision the different periodicities of the two types of surge and to the
became possible. length of detailed calculation of surges by the method of elastic
waves. That that could be done had been demonstrated theore
Model Tests on Surge T a n k . Several contributors, Mr. tically by himself (Jaeger 1933, p. 188). More recently, the
Headland, Mr. Hutton, Dr. Zienkiewicz, Mr. Linton, and Mr. surge tank of IskreArc had been analysed in regard to water
Paynter, had referred to model tests. Much information could hammer by two engineers of Electricit6 de France (Bouvard and
be gained by directly comparing their comments. Molbert 1953)#.
A very accurate technique for testing surges had been evolved. The request had been made for more details to be published
In many instances, the pressure tunnel could not be represented on the technique and results of model tests for actual surge tanks.
at full length and the model would be distorted, the scale for the Numbers of surge tanks had been tested. Surges had been
lengths being different from the scale for the horizontal areas recorded, the efficiency of new designs, new arrangements,
and for the vertical dimensions. It was essential to test before orifices, Venturi arrangements, stability, and so on, had been
hand the friction losses in the pressure tunnel, since the final checked. But very little had been published. Bibliographical
choice of the model scale would depend on those losses. Very details were available in a few textbooks (Jaeger 1949a (French
smooth pipes might have to be used, and in some instances it translation 1953); Escande 1953).
had been found that turbulence did not develop fully in those Summarizing, it could be said that model tests for checking
pipes. surges andwaterhammer pressure rises had frequently been used
For the Innertkirchen surgetank tests, the model scale had succes’sfully and that the techniques for those tests were well
been adjusted to the average value of the friction losses. The known. In many instances, the problems to be investigated had
scale ratio was given by a generalization of the similarity law of nevertheless to be simplified to make tests possible. The great
Froude. Great importance ought to be given to stability for importance of tests carried out on existing prototypes should
lowhead schemes. That required special devices (electronics) to again be stressed.
be used to reproduce on the model the condition of constant
* MEYERPETER,
E., and FAVRE,
vol. 100, Nos. 4, 5
H. 1932 Schweizerische Bauzeitung, *
t PAYNTER,
H. M. 1953 Trans. Am. SOC.CivilEng.,vol. 118,p. 961.
BOUVARD, M., and MOLBERT, J. 1953 La Houille Blanche, vol. 28,
p. 260, ‘Calcul de la cheminCe a etranglement de la chute IskreArc’.