You are on page 1of 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/286902410

Water hammer and other pipe transient flow problems

Article  in  Appita Journal · October 2014

CITATIONS READS

0 1,274

2 authors, including:

Geoff Covey
Covey Consulting
71 PUBLICATIONS   287 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

chemicals from biomass View project

All content following this page was uploaded by Geoff Covey on 26 January 2016.

The user has requested enhancement of the downloaded file.


WATER HAMMER AND OTHER Water Hammer

PIPE TRANSIENT FLOW Water hammer is caused by shockwaves moving


through the liquid at the local speed of sound (also
PROBLEMS referred to as ‘celerity’). It can occur with
incompressible fluids (i.e. liquids) when flow is
ROSS PATTERSON1 AND GEOFF suddenly stopped downstream (or started upstream)
COVEY1 and a shock wave is created as flow velocity is
suddenly altered.
1
Director, Covey Consulting Pty Ltd, 832 High
It is generally considered that water hammer will not
Street, Kew East, Vic 3102
occur if the time to close the valve is less than:
KEYWORDS: Fluid Flow, Water Hammer, Mass Tmin = 2L/C (1)
Oscillation.
Where:
ABSTRACT: Various types of transient flow can
L is the length of pipe
cause minor or major damage to pipework and
associated equipment. Water hammer, caused by too C is the celerity (note this speed depends on
rapid closing of a downstream valve is the most liquid compressibility, the elasticity of the pipe
commonly encountered problem, but is not eh only wall and the presence of gas bubbles in the
cause. In the present case empty pipelines acting as liquid)
surge chambers were able to cause massive Tmin corresponds to the time for the shock wave to be
oscillation of the fluid column. The problem could be reflected and to travel the full length of the pipe in
resolved by comparatively simple measures both directions.

INTRODUCTION In a pipe with elastic walls, the celerity is given by 1

Hammering and unsteady flow in piping systems is a


common occurrence and often results in minor or
C = √𝐾⁄𝜌 / √(1 + ψ𝐾𝐷⁄
𝐸𝑒 (2)
major damage. Normal piping design techniques will
often fail to predict situation which will lead to the Ψ = 1-ν2 (3)
problem, and its proper diagnosis and cure in an Where:
operating plant can sometimes be difficult.
D is the pipe diameter
Although in common usage any form of ‘banging’ or E is the Young’s modulus of the pipe material
‘jumping’ in pipework during transient conditions is
referred to as ‘water hammer’ this term should e is the thickness of the pipe wall
strictly only be applied to problems related to the too ν is the Poison ratio of the pipe material
rapid closing of a valve downstream of a flowing
column of liquid, or pump starting or stopping ρ is the liquid density
upstream. There are at least five phenomena that can This means an closing time greater than.0.1 sec is
give rise to the problems: required for 100m of pipe, or 1.3 secs for 1000m.

 Water hammer The problem obviously gets more significant for long
pipe runs.
 Flashing.
Water hammer is most likely to occur when the flow
 Mass Oscillation. is stopped (i.e. valve closing) because then there is
 Column separation the momentum of the moving column of liquid to be
arrested. Hammer on opening of a valve into a full
 Bore flow into an empty pipe. pipe is less likely as there is no equivalent sudden
change in momentum.

Appita 2014 124


For simple piping systems, the likelihood of water If a pressure wave travels down a pipe, then the
hammer can be predicted simple calculations, and for negative part of the wave can cause problems, even
more complex systems by use of numerical if the positive swings are not of concern. If the local
programs. pressure falls below the vapour pressure of the water
then cavitation will occur and a slug of vapour and/or
In this paper, water hammer will be used only for previously dissolved gas will form. As flow
this phenomenon, simply to distinguish it from the stabilises (or direction reverses again) the pressure
other possible causes. will rise to above the vapour pressure, the slug will
collapse and the two halves of the column will
accelerate towards each other. When the columns
Flashing moving in opposite directions meet, there will be a
sudden loss of momentum and a pressure shock
If liquid is vented to a line or vessel in which at some
wave will be generated in a similar to that seen in
point the system pressure is lower than the vapour
classical water hammer.
pressure of the liquid, flashing is likely to occur. In
some systems the formation and subsequent collapse As with water hammer this is most commonly
of vapour bubbles can result in banging in the low associated with the stopping of flow, but in this case
pressure part of the system (commonly observed in usually by closure of an upstream valve. However,
tanks heated with direct steam but without adequate it can also be induced by oscillating flow. It is also
mixing, also downstream of restricting valves more likely if there are local high points in the
handling hot liquid). system.
Pumped condensate piping can be a real problem. In Column separation is difficult to analyse
fact any flow of a liquid which is near saturation quantitatively, and expensive software is required to
needs care. It is always a good idea to ensure that predict the pressure transients that it can create.
throughout the pipe run, and under transient Other software is available that does not perform the
scenarios, the liquid remains a liquid. Obviously full calculations, but it does predict when such
always place the control valve at the end of the pipe separation might occur.
run.
As noted above, column separation will occur when
Gravitational effects can be a real problem. A classic the local pressure falls below the vapour pressure.
problem (which we don’t see a lot of in Australia) is Therefore it is more likely in liquids that are hot or
rattling pipes in a pumped hot water radiator system. contain dissolved gasses.
In the upper floors it is quite easy to generate
cavitating flow conditions.
Rapidly filling pipe (bore flow, rushing flow)
This occurs when a valve opens (or a pump starts)
Mass oscillation
into an initially empty pipe – particularly if there is a
If there is some form of surge tank in the system (or constriction at the discharge of the pipe, but not a
some element that can act as a surge tank) then mass fully closed discharge end.
oscillation may occur. This is because as the surge
Initially the liquid enters the pipe at high speed and
device empties and refills it can induce (or prolong)
so a substantial momentum is established. Then
pressure swings in the system.
when the liquid approaches the constriction there is a
Compared with water hammer, the pressure swings sudden deceleration, similar to that in classical water
associated with mass oscillation are generally much hammer as it is retarded by a closing valve. This
less extreme and of a longer period. The oscillations gives rise to shock waves and rapid pressure
are gradually damped out as a result of fluid friction, fluctuations in the pipe.
Mass Oscillation can be predicted from the mass and If the discharge end of the pipe is sealed (or nearly
effective spring rate, just as with any oscillating so) then air trapped in the pipe will cushion the
mechanical system. deceleration and the shock will be attenuated.
Conversely, if there is only a minor constriction at
Column Separation
the discharge end, the flow will not be significantly
retarded and again there will not be large shocks.

Appita 2014 125


The open end of the pipe can also act as a point of DIGESTER PIPING - ANOTHER CASE STUDY
reflection.
A pulp mill had three batch digesters with side relief
This behaviour cannot be modelled in a predictive lines connected to a common pipe discharging into
manner at present. Part of the difficulty is that there an atmospheric pressure tank. The nature of the
will be mixing of gas and liquid in the pipe so that operation was that only one digester could relieve at
the leading part of the flow is bubbly. However, it is a time, and so for most of the time there was no flow
rarely possible to predict the degree of entrainment from the digesters to the receiving tank. The liquid
of gas. The velocity of sound and effective relieved was hot, but well below its boiling point.
compressibility of the liquid is a very strong function
After some years’ operation it was found that
of the fraction of bubbles in it, and this in turn affects
hammering was developing at the start of the side
the time and severity of the shock waves.
relief process – particularly when relieving from the
Another problem that can arise from bore flow is digester with the longest pipe run to the receiving
when the fast moving column hits a sudden change tank. At times there would just be loud noise, at
of direction (e.g. an elbow or short-radius bend). In other times pipes would be displaced from their
this case there is a sudden change of momentum and supports and on some occasion more severe damage
the shock can cause local damage to the piping, or resulted. Although the problem was initially loosely
the pipe to jump off its supports. described as ‘water hammer’, the fact that it was
induced by the opening of a valve at the upstream
Of course there is always a net reaction at any pipe
end of the line indicated that it was not water
bend which needs appropriate bracing caused by the
hammer as defined above but some other transient
direction change. But this is exacerbated when there
flow behaviour.
is a traveling slug of fluid.
As already noted, this behaviour cannot be modelled The design of the pipe work and the comparatively
accurately, but slowing the opening of the side relief long periods between uses meant that the lines could
valves and the use of large-radius bends should drain off liquid between relief periods. This would
reduce the severity of the problem. mean that the falling legs from the digesters not
relieving would be full of air and sealed at the top by
closed valves. Therefore, they could act as surge
A SIMPLE CASE STUDY tanks. In stopping flow (i.e. valve 2 closing) the
addition of surge tanks would tend to dampen
A long 350 dia stainless pipe was installed passing
pressure waves. However, in this scenario they could
effluent over pipe racks for about one kilometre. The
promote oscillating flow which could present
effluent pumps had variable speed drives and were
problems itself, and also lead to column separation.
run up to speed quite slowly, but ongoing transient
flow problems occurred. The pipe shook itself clear Modelling
of brackets, and moved around on the bridge in a
manner that was quite un-nerving. WHAMO is a MS-DOS program written by US
Army Corps of Engineers in the 1990’s 2 3. It used to
The pipe discharged into a tank at the far end with no
be available as a free download, but is no longer
end restriction.
supported. It was primarily intended for use on
The very simple solution was suggested by a local reservoirs and especially for hydro-electric systems.
EWS engineer. This consisted of a 50NS non return Therefore the examples given are primarily for long
valve fitted to the top of the pipe near the far end. runs of large diameter pipes. However, the
The idea was to allow a reverse flow of air back into algorithms work satisfactorily for smaller pipes and
the pipe to kill the negative pressure pulses. This was the program has even been used successfully to
an instantaneous and totally successful solution, for analyse domestic water piping4.
almost nil cost. If only all engineering fixes were that
The manual for the program includes the use of a
easy.
Windows graphical interface, but on the website it is
noted that this is not available (for public use at
least). The package also includes a graphical output
module, but difficulty was experienced in using this
and instead the calculated results were exported to an

Appita 2014 126


Excel spreadsheet for graphical modelling. There are Despite these constraints, the program worked well
also a number of constraints on the program that are for the required purpose.
not documented in the manual and which are not
tracked by the built-in error reporting system.

Figure 1 the piping network analysed (numbers at junctions and bends are ‘nodes’

Sirvole 4 showed that the WHAMO program gave


very similar results to those from other transient flow
Analysis
programs.
A model of the system was programmed into
The program will identify water hammer and mass
WHAMO and various scenarios analysed.
oscillation problems, and will also identify when
column separation is likely, but it does not analyse The first case considered is one which was not
the consequences of column separation*. It does not actually encountered, but is included to show what
attempt to analyse bore flow. true water hammer would look like in this system. It
is the case where flow has been established and then
the downstream valve (V2 in Fig 1) was closed
*
WHAMO tends to list frequent possible column
separations in the first 0.1 s of operation. These are algorithms and they are not included in the analysis
almost certainly merely transients in the calculation presented below.

Appita 2014 127


rapidly (over 0.6 sec). For this simulation the two Fig 3 shows the pressures at various points in the
inactive drop legs (nodes 650 t0 900 and 1000 to 950 system*. As with the previous case, there are
in Fig 1) have been treated as blanked off. The fluctuations in pressure at various points, but the
results are shown in Figure 2. nature of the fluctuations are quite different from
those with true water hammer:
 There is a delay before the peaks begin.
This displays classical water hammer with the
pressure rising rapidly as the valve approaches the  The pressure fluctuations are less extreme
fully closed position and then displaying high and of much longer period.
frequency (ca 4Hz) oscillations that are gradually
 The peaks are at slightly different times at
damped. (Note – WHAMO does not include a
different places in the system.
correction for impossible vacuums, hence pressures
much lower than -101kPa g can be displayed).
Responses of this type are known to cause damage to The program calculated that column separation
piping and equipment. occurred at 3.2 seconds.
The next scenario considered is the actual operation The separation of the peaks in Fig 3 suggests that
of the system at the time of damage, with the mass oscillation may have occurred. This was
upstream valve opening in 0.6 sec. confirmed by considering flow rates at various points
as shown in Fig 4.

2500
n250
2000 n1350
n350
1500 n550
n500
1000
n900
Pressure kPa g

500 n920

0
0 0.5 1 1.5 2 2.5 3
-500
Time (sec)
-1000

-1500

-2000

Figure 2 Pressure response to rapid closure of down-stream valve.

Appita 2014 128


120
100
80 n950
60 n1350
Pressure kPa g

40
n350
20
n550
0
n500
-20 0 1 2 3 4 5 6 7
n900
-40
n920
-60
-80
Time sec

Figure 3 Pressures in line after opening upstream valve in 0.6 sec.

over 5 seconds.

50

40
Node 1350
30 Node 550
Node 920
Flow L/s

20 node 880
Node 1220
10 node 900
0
0 1 2 3 4 5 6 7
-10

-20
Time sec

Figure 4 Flow rates at various points in the system after opening upstream valve in 0.6 sec

This figure clearly shows the oscillating flow in the 5 and 6 show pressure and flow responses for a valve
system. In particular, the flows to the two empty opening.
drop legs (nodes 900 and 1220) are seen to be out of
phase and contributing to or causing the oscillations. These clearly show that slowing the valve opening
rate would substantially reduce the magnitude of the
The behaviour illustrated in Figures 3 and 4 is quite pressure peaks and of the mass oscillations.
consistent with that which was observed in the plant.
As a temporary measure the valve opening rate was
The simplest way to avoid the problem would be to reduced, but operationally this presented some
slow down the rate at which the valve opens. Figures problems, so other solutions were considered.

Appita 2014 129


100

80

60 n950

40 n1350
Pressure kPa

n350
20
n550
0 n500
0 2 4 6 8 10
-20 n900

-40 n920

-60
Time sec

Figure 5 Pressures in line after opening upstream valve in 5 sec.

fall was after the branches from the three


digesters.
The problem was recognised as being primarily due  Fit non-return valves in the falling lines just
to the empty legs acting as surge tanks which could ahead of where they joined the header.
interact and produce oscillating flow. This could be
cured by one of two means which were hydraulically Hydraulically the two solutions were equivalent, but
equivalent: it was cheaper and simpler to use the approach with
non-return valves. The pressure and flow responses
 Raise the discharge header to about the with the valve opening in 0.6 seconds are shown in
level of the discharge valves so that all the Figures 7 and 8.

50

40
Node 1350

30 Node 550

Node 920
Flow L/s

20 node 880

Node 1220
10
node 900

0
0 2 4 6 8 10
-10
Time sec

Figure 6 Flow rates at various points in the system after opening upstream valve in 5 sec

Appita 2014 130


120
n250
100
n1350
80
n350
60
n550
Pressure kPa

40
n500
20 n900
0 n920
0 2 4 6 8 10
-20

-40
Time sec
-60

Figure 7 Pressures in lines with non-return valves with valve opening in 0.6 sec.

50
45
40 Node 1350
35
Node 550
30
Flow L/s

Node 920
25
node 880
20
15 Node 1220
10 node 900
5
0
0 2 4 6 8 10
Time s

Figure 8 Flows in lines with non-return valves with valve opening in 0.6 sec.

It is apparent that the addition of the non-return radius bends with long radius bends, and if necessary
valves eliminates all of the flow instability and most by some reduction in the valve opening rate.
of the pressure fluctuations. This was confirmed in
practice by the elimination of ‘hammer’ from the
system.
1
Elimination of bore flow problems Henclik, H Mathematical model and numerical
computations of
As already discussed, WHAMO does not attempt to transient pipe flows with fluid-structure interaction
model the effect of bore flow. In the case described Transactions Of The Institute Of Fluid-Flow
above it was apparent that bore flow could Machinery No. 122, 2010, 77–94
2
WHAMO ref
potentially present problems, but in practice, with the 3
WHAMO manual
elimination of the mass oscillations due to the 4
Sirvole, K. : Transient Analysis In Pipe Networks
interacting surge tanks this was not a problem. M.Sc Thesis, Virginia Polytechnic Institute
& State University, 2007
Had it been necessary, bore flow problems could
have been eliminated by replacing elbows and short

Appita 2014 131

View publication stats

You might also like