Waterhammer Potential in Pumps and Systems
Waterhammer Potential in Pumps and Systems
Waterhammer Potential in Pumps and Systems
Systems
by Dr. C. Samuel Martin,
Waterhammer Consultant and Professor Emeritus, Georgia Tech, Atlanta,
GA
By definition waterhammer is a pressure (acoustic) wave phenomenon created by relatively sudden changes in the
liquid velocity. Although the name waterhammer may appear to be a misnomer in that it implies only water and the
connotation of a hammering noise, it has become a generic term for pressure wave effects in liquids. Strictly
speaking, waterhammer can be directly related to the compressibility of the liquid. For relatively slow changes in
flow for which pressure waves have little to no effect the unsteady flow phenomenon is called surging.
In pipelines, sudden changes in the flow (velocity) can occur as a result of 1) pump and valve operation in pipelines,
2) vapor pocket collapse, or 3) even the impact of water following the rapid expulsion of air out of a vent or a
partially open valve.
Potentially, waterhammer can create serious consequences for pipeline designers if not properly recognized and
addressed by analysis and design modifications. There have been numerous pipeline failures of varying degrees and
resulting repercussions of loss of property and life. Three principal tactics for mitigation of waterhammer are 1)
alteration of pipeline properties such as lowering of pipe profile to increase local pressure or increasing pipe
diameter to reduce velocity, 2) implementation of improved valve and pump control procedures, and 3) design and
installation of surge control devices.
Waterhammer Basics
For wave propagation in liquid-filled pipes the acoustic (sonic) velocity is modified by the pipe wall elasticity by
varying degrees, depending upon the elastic properties of the wall material and the relative wall thickness. The
expression for the wave speed a is
(equation 1)
where is the liquid mass density of the liquid, K is the bulk modulus of the liquid, E is the elastic modulus of the
pipe wall, D is the inside diameter of the pipe, and e is the wall thickness. The pressure change associated with the
rapid velocity change across a waterhammer (pressure) wave is the well-known Joukowsky equation
(equation 2 )
Experience shows that waterhammer is usually not a problem with pump start-up if the pipeline has no voids
(neither gas nor vapor), but is filled with liquid. Likewise, pump power failure will typically not be a problem if the
minimum pressure in the system due to this depressurization does not fall to vapor pressure.
Pump Start-up
In many installations pumps are periodically shut down and restarted, generating pressure excursions of various
degrees. The normal start-up of centrifugal pumps against a closed pump discharge control valve or seated check
valve does not typically yield consequential waterhammer pressures provided the pipeline system is free of voids.
The torque and power requirements for higher specific speed mixed flow pumps and axial flow pumps is an issue for
pump start-up against a closed valve, but waterhammer is normally not a concern for full pipe systems.
On the other hand, pipes containing voids can lead to waterhammer subsequent to pump start, depending upon the
rate of filling. The most serious waterhammer potential exists for systems with vapor or steam pockets that will be
collapsed suddenly as the void is extinguished. It is possible that the flow rate at the moment of void collapse can
exceed the design flow if the pump runs out on its curve. Entrapped gases such as air can result in significant
pressure rise during pump start because of the so-called spring effect of the gas. The peak pressure, which is not
necessarily waterhammer, can attain values several times the driving pressure, depending upon the rate of
pressurization.
Pump Shutdown
Normal shutdown of pumps is often accomplished by initial closure of pump discharge valve, followed by stopping
of the pump driver. If the control valve time of closure is properly chosen to a) preclude vapor formation from too
rapid closure, or b) to prevent high reverse flow from too slow closure, waterhammer can be avoided.
For relatively short systems with large static lift, check valve slamming is a possibility if the liquid column has a
quicker response than the valve. Swing check valves or other designs are frequently employed in pump discharge
lines, often in conjunction with slow acting control valves. A check valve should open easily, have a low head loss
for normal positive flow, and create no undesirable transients by its own action. For short systems, a slow
responding check valve can lead to waterhammer because of the high reverse flow generated before closure. A
spring or counterweight loaded valve with a dashpot can 1) give the initial fast response followed by 2) slow closure
to alleviate the unwanted transient. The proper selection of the load and the degree of damping is important,
however, for proper performance. Check valve slam is also a possibility from stoppage or failure of one pump of
several in a parallel system, or resulting from the action of an air chamber close to a pump undergoing power failure.
Check valve slam can be reduced by the proper selection of a dashpot. Mitigation of waterhammer is possible by a)
special check valve designs that allow for faster response, b) counterweights or springs to accelerate the initial valve
motion, and c) dashpots to cushion the final valve motion to reduce mechanical shock and valve slamming, which
can lead to waterhammer.
(equation3)
where TM is the driver (motor) torque, T is the resisting fluid torque, WR2 is the total moment of inertia of all
rotating parts (motor, pump and entrained liquid, shaft, and possibly flywheel), g is gravitational acceleration, w and
N are the rotational speed in rad/sec and rpm, respectively, and t is time. For loss of power to an electric motor TM =
0 in Eq. (3), yielding the rate of pump deceleration in terms of the unbalanced fluid torque T
(equation4)
The significance of the moment of inertia WR2 is apparent from Eq. (4). Moreover, the rate of rotor deceleration
dN/dt is also influenced by the torque delivered by the pump at the corresponding speed N and flow Q. For
simplicity the coupling of the fluid system can be represented by linear momentum for a liquid column of length L.
For an uniform pipe of diameter D, cross-sectional area A, and instantaneous flow Q
(equation5)
where H is the level of the hydraulic grade line (HGL), x is the distance along the pipe, and f is the Darcy-Weisbach
resistance coefficient. Equation (5) is valid for both compressible flow (waterhammer) and incompressible (rigid
column). To demonstrate the effect of liquid column deceleration, we consider a short pipe of uniform diameter and
insignificant friction for which the effect of waterhammer is minimal, reducing Eq. (5) to
(equation6 )
where H is the static head on the system. For analysis of pump power failure problems Eq. (4) is required to assess
the change in pump speed, while the change in flow can only be determined from Eq. (5) alone for relatively short
pipe systems for which elastic effects are not important. Most analyses require consideration of the liquid and pipe
elasticity by the method of characteristics (MOC).
Figure 1
A centrifugal pump is delivering liquid at a constant rate when there is a sudden loss of power from the prime
moverin this case an electric motor. For the postulated case of no discharge valves, or other means of controlling
the flow, the loss of driving torque leads to an immediate deceleration of the shaft speed, and in turn the flow. The
four curves are head (H), flow (Q), speed (N), and torque (T), expressed in percentage of the rated values. With no
additional means of controlling the flow, the higher head at the final delivery point will eventually cause the flow to
reverse (Q < 0) while the inertia WR2 of the rotating parts maintains positive rotation (N > 0). Up until the time of
flow reversal the pump has been operating in the normal zone, albeit at reduced flow. In order to predict system
performance in regions of negative rotation and/or negative flow the analyst requires characteristics in these regions
for the machine in question. Indeed, any peculiar characteristic of the pump in these regions could be expected to
have an influence on the hydraulic transients. It is important to stress that the results of such analyses are critically
governed by the following three factors: 1) availability of complete pump characteristics in zones the pump will
operate, 2) complete reliance upon dynamic similitude (homologous) laws during transients, and 3) assumption that
steady-flow derived pump characteristics are valid for transient analysis.
Eight possible zones of operation, four normal and four abnormal, will be
Figure_2
Figure_3
discussed here with reference to Figure 2, developed by Martin (Ref. 1).
Fig 2 & 3 - Click image for full-size
In Figure 2 the head H is shown as the difference in the two reservoir
image
elevations to simplify the illustration. The effect of pipe friction may be
ignored for this discussion by assuming that the pipe is short and of relatively large diameter. The regions referred to
on Figure 2 are termed Zones and Quadrants, the latter definition originating from plots of lines of constant head and
constant torque on a flow-speed plane (Q - N axes), as shown in Figure 3. Quadrants I (Q > 0, N > 0) and III (Q < 0,
N < 0) are defined in general as regions of pump or turbine operation, respectively. It will be seen, however, that
abnormal operation (neither pump nor turbine mode) may occur in either of these two quadrants. A very detailed
description of each of the eight zones of operation is in order. It should be noted that all of the conditions shown
schematically in Figure 2 can be contrived in a laboratory test loop using an additional pump (or two) as the master
and the test pump as a slave. Most, if not all, of the zones shown can also be experienced by a pump during a
transient under the appropriate set of circumstances.
Quadrant I. Zone A (normal pumping) in Figures 2 and 3 depicts a pump under normal operation for which all four
quantitiesQ, N, H, and T are regarded as positive. In this case Q > 0, indicating useful application of energy. Zone
B (energy dissipation) is a condition of positive flow, positive rotation, and positive torque, but negative head
quite an abnormal condition. A machine could operate in Zone B by 1) being overpowered by another pump or by a
reservoir during steady operation, or 2) by a sudden drop in head during a transient caused by power failure. It is
possible, but not desirable, for a pump to generate power with both the flow and rotation in the normal positive
direction for a pump, Zone C (reverse turbine), caused by a negative head, and resulting in a positive efficiency
because of the negative torque. The maximum efficiency would be quite low due to the bad entrance flow condition
and unusual exit velocity triangle.
Quadrant IV. Zone H, labeled energy dissipation, is often encountered shortly after a tripout or power failure of a
pump, as illustrated in Figure 1. In this instance the combined inertia of all the rotating elementsmotor, pump and
its entrained liquid, and shafthas maintained pump rotation positive but at a reduced value at the time of flow
reversal caused by the positive head on the machine. This purely dissipative mode results in a negative or zero
efficiency. It is important to note that both the head and fluid torque are positive in Zone H, the only zone in
Quadrant IV.
Quadrant III. A machine that passes through Zone H during a pump power failure will then enter Zone G (normal
turbining) provided that reverse shaft rotation is not precluded by a mechanical ratchet. Although a runaway
machine rotating freely is not generating power, Zone G is the precise mode of operation for a hydraulic turbine.
Note that the head and torque are positive, as for a pump but that the flow and speed are negative, opposite to that
for a pump under normal operation (Zone A). Subsequent to the tripout or load rejection of a hydraulic turbine or the
continual operation of a machine that failed earlier as a pump, Zone F (energy dissipation) can be encountered. The
difference between Zones F and G is that the torque has changed sign for Zone F, resulting in a braking effect,
which tends to slow the free wheeling machine down. In fact the real runaway condition is attained at the boundary
of the two zones, for which torque T = 0.
Quadrant II. The two remaining ZonesD and Eare very unusual and infrequently encountered in operation, with
the exception of pump turbines entering Zone E during transient operation. Again it should be emphasized that both
zones can be experienced by a pump in a test loop, or in practice in the event a machine is inadvertently rotated in
the wrong direction by improper wiring of an electric motor. Zone D is a purely dissipative mode that normally
would not occur in practice unless a pump, which was designed to increase the flow from a higher to lower
reservoir, was rotated in reverse, but did not have the capacity to reverse the flow (Zone E of Quadrant III mixed
or axial flow), resulting in Q > 0, N < 0, T < 0, for H < 0. Zone E, for which the pump efficiency > 0, could occur in
practice under steady flow if the preferred rotation as a pump was reversed. There is always the question regarding
the eventual direction of the flow. A radial-flow machine will produce positive flow at a much reduced capacity and
efficiency for N < 0 (Zone E of Quadrant IIreverse pumping) compared to N > 0 (Zone A of Quadrant Inormal
pumping), yielding of course H > 0. On the other hand, mixed and axial-flow machines create flow in the opposite
direction (Zone E of Quadrant III), and H < 0, which corresponds still to an increase in head across the machine in
the direction of flow.
The transient scenario illustrated in Figure 1 can be related to Eqs. (4) and (5) by the slopes dN/dt and dQ/dt,
respectively. The slope of the pump speed curve, whether from analysis or measurement, can reveal the sign of the
pump torqueeither positive or negative. For this example the pump torque remains positive (dN/dt < 0) until the
machine enters Quadrant III at the runaway pointboundary between Zones G (turbining) and F (braking).
Air-Vacuum Breaker
The introduction of air into a pipe subsequent to pump power failure is one of many possible Fig 6 - Click image
solutions to combat the possibility of liquid-column separation and collapse and the potential for full-size image
dire consequences. There are questions regarding the modeling of the vacuum breaker or
other air admittance device as well as the handling of the air mass during the subsequent transient. Figure 6 depicts a
supply system delivering water to safety related cooling elements of a nuclear plant. Analysis indicated that liquid
column separation and serious waterhammer pressures would occur for pump power failure of two or more of the
five pumps. Following an evaluation of various protective measures air chamber, flywheel, one way surge tank
the efficacy of the more economical vacuum breaker scheme was assessed. In order to be comfortable with the
calculations based upon standard MOC calculations, site measurements of transient pressures were made.
Figure 7 is a graph showing plots of the 1) recorded and 2) calculated pressures at pump
discharge for a 200 mm check-valve type vacuum breaker shown in Figure 6, along with
Fig 7 - Click image
analytical traces for 3) no surge protection (column separation), and 4) a 2000 ft3 air
for full-size image
chamber. In summary, the vacuum breaker was deemed a practical engineering solution even
though the peak pressure was still approximately 50% of that with column separation. Because of economics, the air
chamber was not chosen notwithstanding its effectiveness in a significant further reduction of peak pressures.
Initially, there was reluctance regarding air admittance via air vacuum valves because of concern regarding
problems with entrapped air during subsequent restarting of the pumps.
protection. It was shown by analyses that vacuum breakers alone at the two main high points along the piping can
protect the force mains from the low pressure problem, but due to the return of the water column from the reservoir,
the pressure rise is quite severe. This necessitates the installation of a surge suppressor at the pump station, which in
conjunction with the vacuum breakers, can protect the line from column-separation type water hammer.
The results of PNET analysis are shown on Figure 8 in terms of minimum HGL envelope,
indicating that the air chamber under proper operation protects the pipeline from negative
Fig 9 - Click image
pressure. Figure 9 compares the measured values of pump manifold pressure with predicted for full-size image
ones from computer code PNET for a three pump trip.
References
1.Martin, C. S., Representation of Pump Characteristics for Transient Analysis, ASME Symposium on
Performance Characteristics of Hydraulic Turbines and Pumps, Winter Annual Meeting, Boston, November 13-18,
1983, pp. 1-13.
2.Martin, C. S. and Cobb, L., Experience with Surge Protection Devices, BHr Group International Conference on
Pipelines , Manchester, England, March 24-26, 1992, pp. 171-178.
3.Water Hammer Prevention, Mitigation, and Accommodation, EPRI NP-6766, Final Report, Research Project
2856-3, July 1992, Volumes 1-6.
Dr. C. Samuel Martin is a retired Professor in the School of Civil and Environmental Engineering at the Georgia
Institute of Technology. He has been an active researches in the field of fluid transients, two-phase flow, cavitation
and hydraulic machinery. Professor Martin has developed hydraulic transient analysis techniques for pump turbines
and has been published extensively. Professor Martin has been a consultant to numerous industries in the general
subject of hydraulics, but in particular in the specialty field of waterhammer and abnormal characteristics of pumps
and turbines. Dr. Martin received his Ph.D. in Civil Engineering from the Georgia Institute of Technology.
Figure 2
Figure 3
Figure 5
Figure 6
Figure 7
Figure 10
Figure 11