Content uploaded by Arris S Tijsseling
Author content
All content in this area was uploaded by Arris S Tijsseling on Aug 03, 2015
Content may be subject to copyright.
1 Copyright © 2010 by ASME
Proceedings of the ASME 2010 Pressure Vessels & Piping Division / K-PVP Conference
PVP2010
July 18-22, 2010, Bellevue, Washington, USA
PVP2010-26131
WATER HAMMER AND COLUMN SEPARATION DUE TO ACCIDENTAL
SIMULTANEOUS CLOSURE OF CONTROL VALVES IN A LARGE SCALE TWO-
PHASE FLOW EXPERIMENTAL TEST RIG
Anton Bergant
Litostroj Power d.o.o.
Ljubljana, Slovenia
Jos M.C. van ’t Westende
Deltares
Delft, The Netherlands
Tiit Koppel
Tallinn University of Technology
Tallinn, Estonia
Janez Gale
Litostroj Power d.o.o.
Ljubljana, Slovenia
Qingzhi Hou
TU Eindhoven
Eindhoven, The Netherlands
Zoltan Pandula
TU Budapest
Budapest, Hungary
Arris S. Tijsseling
TU Eindhoven
Eindhoven, The Netherlands
ABSTRACT
A large-scale pipeline test rig at Deltares, Delft, The
Netherlands has been used for filling and emptying
experiments. Tests have been conducted in a horizontal 250 mm
diameter PVC pipe of 258 m length with control valves at the
downstream and upstream ends. This paper investigates the
accidental simultaneous closure of two automatic control valves
during initial testing of the test rig. The simultaneous closure of
both valves has induced upsurge and downsurge at the same
time. Large water hammer and column separation have caused
failure of pipe supports and leakage at pipe joints. The incident
was caused by a fault in an electronic conversion box due to
power failure. Afterwards the downstream end automatic valve
has been modified to a manually operated valve to avoid the
accidental simultaneous closure of the valves. The accidental
transient event has been fully recorded with pressures, flow
rates and water levels. The measurements of the accident are
presented, analyzed and discussed in detail. Photographs show
the damages to the system.
Keywords: pipeline test rig, water hammer, column
separation, fluid-structure interaction, accident
INTRODUCTION
Filling and emptying of pipelines is common in many
hydraulic applications, such as water distribution, hydropower,
sewage and cooling, conveyance of storm-water flows during
intense rainfall, fire fighting systems and oil transport through
long pipelines. Rapid filling and emptying of the piping system
may be considered as a specific case in which both vaporous
and gaseous cavities may be present. During filling or drainage
part of the pipeline is filled with liquid and part of it is filled
with gas (or vapour), while in between the pipeline is filled with
a mixture of both. Undesired transients may occur and lead to
pipe rupture due to overpressures, pipe collapse due to vacuum
conditions, and entrapped gas pockets preventing complete
filling or drainage. Design engineers should be able to predict
fluid transient events in systems and take measures to keep
transient loads within the prescribed limits. The modelling of
filling and emptying of piping systems is a relatively complex
task and different pipe configurations with many types of initial
and boundary conditions may lead to a wide variety of solutions
[1-6]. Developers and users of computational codes need full-
scale data with which to compare their theoretical models.
Unfortunately such data are limited and available only for
special cases [3-4, 7].
2 Copyright © 2010 by ASME
Here, a large-scale pipeline test rig at Deltares, Delft, The
Netherlands has been used for filling and emptying experiments
[8]. Previously the test rig had been designed and used for
classical two-phase flow experiments. Pipeline filling and
emptying experiments have been conducted in a modified
horizontal 250 mm diameter PVC pipe of 258 m length with
control valves at the downstream and upstream ends. This paper
investigates the accidental simultaneous closure of two
automatic control butterfly-type valves during initial testing of
the test rig. Luckily the accidental transient event has been fully
recorded including measurements of pressures along the PVC
pipe test section and flow rates in the system, and photographs
show the damages to the pipeline test rig. The automatic control
valve signals, flow rates at the downstream and upstream ends
of the test PVC pipe section, and pressures at six locations
along the test section during the accident are presented,
analyzed and discussed in detail. Modifications to the test rig
are presented and lessons learned from the accidental event are
addressed.
Service valve
Control valve
Vortex flow meter
Check valve
Control valve
Service valve
Inlet valve
Dead end branch
S
u
p
p
l
y
s
t
e
e
l
p
i
p
e
l
i
n
e
Upstream end
constant head
tank
Steel
PVC
x = 0 m
Transparent
sections
Outlet steel pipeline
Electromagnetic
flow meter
Control valve
Outlet pipe
Sump
Transparent
section
Horizontal PVC pipeline test section:
- internal diameter D
PVC
= 235.4 mm
- wall thickness e
PVC
= 7.3 mm
- length L
PVC
= 257.9 m
Sharp U-bend
M
M
M
M
M
S
t
e
e
l
PVC
Pipe bridge
M
OFF
OFF
Electromagnetic
flow meter
Stand pipe
Large compressed
air reservoir
Figure 1: Dynamic two-phase flow test rig at Deltares, The Netherlands.
3 Copyright © 2010 by ASME
x = 0. m
Horizontal PVC pipeline test section:
- internal diameter D
PVC
= 235.4 mm
- wall thickness e
PVC
= 7.3 mm
- length L
PVC
= 257.9 m
Sharp U-bend
M
M
M
M
x = 111.7 m
Pressure p
5
x = 46.6 m
Pressure p
3
0. m
x = 180.4 m
Pressure p
7
x = 263.7 m ; Sharp bend
x = 257.9 m ; PVC Steel
x = 267. m ; Discharge Q
d
x = 268. m;
Valve position signal y
d
x = 249.4 m ; Pressure p
9
x = 203.5 m ; Pressure p
8
Air at
atmospheric
pressure
x = - 32.8 m
x = - 43.1 m
x = - 31.6 m
+ 1.2 m
x = - 27.2 m
x = - 29.9 m;
Valve position
signal y
u
x = - 30.3 m
x = - 58.2 m
x = - 55.7 m
Large radius bend (R = 5D
PVC
)
x = - 9.8 m
x = - 14. m ; Steel PVC
-4.3 m
Pressure p
u
Discharge Q
u
x = - 19. m
x = 1.6 m ; Pressure p
1
x = - 34.6 m
- 1.1 m
0
.
m
1
.
3
m
- 3.6 m
0. m
x = - 30.8 m
Steel pipeline DN250
Large radius bends
(R = 5D
PVC
)
Large
radius
bends
(R = 5D
PVC
)
x = 124.8 m
Water
Steel
pipeline
DN200
Steel
pipeline
DN500
21.4 m
Figure 2: Layout of dynamic instruments in water supply steel pipe system and PVC pipe test section for
influential quantities during accidental event.
TEST RIG
The dynamic two-phase flow test rig at Deltares, Delft,
The Netherlands has been used for pipeline filling and emptying
experiments. Due to the limited capacities of water tank and
compressed-air reservoir the original industrial size 250 mm
nominal diameter horizontal PVC pipeline test section was
shortened from 600 m to 258 m. The water was supplied from a
25 m constant head tank through a supply steel pipeline and
PVC pipe bridge to the PVC pipe test section (see Fig. 1). The
compressed-air reservoir (volume of about 70 m
3
) provided the
air plug for a controlled rapid emptying of the test section. This
section describes the layout of the test rig during the
commissioning period in which the accident happened. During
trial and error tests the liquid flow in the system was regulated
by automatically operated control valves at the upstream-end
constant-head tank and at the outlet of the PVC pipe test
section. The modified layout after the accident will be described
later on.
The water-supply steel pipeline comprised inlet and service
valves, an automatically operated DN150 butterfly valve, steel
4 Copyright © 2010 by ASME
pipe sections (of 200 mm nominal diameter), and
electromagnetic flow meter connected to the PVC supply pipe
at its downstream end. The stand pipe and the pipe bridge
served for a better control of the inflow conditions. The
horizontal PVC pipe test section consisted of straight pipes,
four large diameter bends (of radius 5×D
PVC
, where D
PVC
= PVC
pipe nominal diameter), one sharp U-bend and three transparent
sections used for flow visualization. It should be noted that due
to space and time limitations the sharp U-bend was mounted to
connect the shortened (from 600 m to 258 m) PVC pipeline.
Because the experiments were planned without any anticipated
large water hammer events, the sharp U-bend was acceptable.
The bend was clamped with two metal supports which partially
(friction) allowed axial movement and five metal axial restraints
(small steel columns) which prevented movement in horizontal
direction. The restraints were calculated to be sufficient for the
planned experiments. The PVC pipeline outlet was connected to
a 250 nominal diameter steel pipeline that diverted water into
the sump. The outlet steel pipeline comprised horizontal and
vertical sections, electromagnetic flow meter and an
automatically-operated DN150 butterfly valve. The
compressed-air supply piping system consisted of 300 mm
nominal diameter steel pipe sections, service and control valves,
vortex flow meter and an undamped swing-type check valve
that prevented backflow of water into the air supply line.
INSTRUMENTATION
The instruments used in the pipeline emptying and filling
measurements have been carefully selected (accuracy,
frequency response) and calibrated prior to and after the
dynamic measurements. The sampling frequency for each
continuously measured quantity during the trial and error tests
was f
s
= 20 Hz. The following quantities were recorded
continuously during the commissioning period:
- valve-position signals to automatically operated control
valves at the upstream-end constant-head tank and at the outlet
of the PVC pipe test section, and to a control valve in the air
supply line (for the upstream valve the voltage supplied by the
computer was recorded; for the downstream valve the current to
the valve was recorded)
- air pressure in the large compressed-air reservoir and at
the control valve
- air temperature at the control valve
- water pressure at the electromagnetic flow meter at the
upstream end of the PVC test section
- water pressures along the PVC pipe test section (at inlet;
at app. 1/5, 2/5, 7/10 and 3/4 of the PVC pipe length measured
from the inlet; at outlet)
- water temperatures along the PVC pipe test section (at
inlet; at app. 1/5 of the PVC pipe length; at outlet)
- air flow rate (vortex flow meter in the compressed-air
supply line)
- water flow rates (2 electromagnetic flow meters: at inlet
and at outlet of the PVC pipe test section)
- void fractions (more accurately: detection of the presence
of water) along the PVC pipe test section (at inlet; at app. 1/2 of
the PVC pipe length; at outlet)
- water levels along the PVC pipe test section (at inlet; at
app. 1/5, 2/5, 7/10 and 3/4 of the PVC pipe length; at outlet)
The accident occurred during steady flow trial and error
tests by unintentional closure of the automatically operated
control valves at the upstream-end constant-head tank and at the
outlet of the PVC pipe test section (see Fig. 1). It has been
found afterwards that void fraction and water level remained
nearly constant during the accidental event (no void, pipe full of
water) and these will not be considered in this paper. The air
supply line was at atmospheric pressure during the accident.
The quantities that significantly changed were the automatically
operated butterfly valve position signals, water pressures and
water flow rates. The layout of the instruments that recorded
these dynamic quantities is depicted in Fig. 2.
Water pressures were measured by strain-gauge type
absolute-pressure transducers (U
x
= ±0.3 %). The uncertainty in
a measurement U
x
is expressed as a root-sum-square
combination of bias and precision error [9]. The uncertainty in
the measured water flow rates was U
x
= ±2 %.
ACCIDENTAL CLOSURE OF CONTROL VALVES
Before the accident
The experiments in the test rig had been ongoing for four
working days when the accident occurred. At the end of the
third working day, after successful completion of the first series
of pipeline filling experiments, it was decided to equip the
motorized upstream butterfly control valve with an automatic
pressure-control function. This computer-supported function
with input from the upstream pressure transducer was available
at that time but not yet in operation. By inclusion of this
function the research group would benefit from the very nice
option to put a constant pressure set-point as upstream boundary
condition. This was one of the important parameters for the
planned second series of pipe emptying experiments.
Figure 3: Pressure history at different positions along the PVC
pipeline during testing and accident.
5 Copyright © 2010 by ASME
The performance of the newly installed upstream-end
automatic pressure-control valve needed to be tested. For this
reason the researchers started with gradually changing the
pressure set-point on the control panel (up and down) thereby
waiting for the initiated transient to damp out into steady state
flow in the PVC pipeline (Figure 3). At all times during the
testing the group focused on stability and reliability of the
automatic upstream-valve performance. Because the
experimental test rig was designed for two-phase flow
experiments, and not for water hammer tests, the pressure jumps
during all tests were moderate and there was no speak of any
significant pressure surge (Fig. 3). As the pressure set-point
testing was not part of the experimental programme, the
sampling rate was reduced to 20 Hz, which was sufficient for its
purpose, and - as it turned out later - it was just high enough to
catch the unintentional water-hammer accident. It turned out
that the performance of the upstream-end automatic control-
valve was reliable and stable. The testing successfully ended
after 6500 seconds and then the team members were gathered in
the control room to discuss new settings for the second series of
experimental measurements (all authors were present except the
last co-author). The second series concerned controlled
emptying of the PVC pipeline; therefore, the flow rate through
the piping system of about 145 litres per second was maintained
and not interrupted. Figure 3 indicates that the accident
happened approximately after 7000 seconds of testing.
PVC U-bend
Metal axial
restraints
damaged
PVC bend
displaced for
about 40 mm
x = 125.5 m
x = 129.3 m
x = 129.6 m
x = 124.1 m
x = 118.8 m
PVC joints
PVC pipes
Metal support
Metal support
PVC pipe and
PVC joint pulled-off
x = 119.1 m
Metal
supports
Figure 4: Overview of damages at the sharp U-bend of the PVC pipeline test section.
a) Damaged PVC pipe joint at distance x = 119.1 m. b) Damaged PVC U-bend at distance x = 125.5 m.
Figure 5: Damages at the sharp U- bend of the PVC pipeline test section.
6 Copyright © 2010 by ASME
The accident event and its consequences
As stated, the validation of the automatic control-valve
took almost two hours, the performance of the upstream-end
automatic valve was satisfactory, the flow rate in the system was
constant, and the team was discussing about how to start with a
new series of experiments when the accident happened. At a
certain moment the team heard a loud noise, but, at the first
instant, it was not clear what was going on. The complete
pipeline was over 300 metres long and only a small part of the
pipeline was under visual inspection of the control room. The
recordings appearing on the display (in the control room) were
a little unclear as they showed that the upstream automatic
control-valve was fully opened and the downstream-end valve
signal was somehow instantaneously interrupted. However, the
measurements clearly showed that the upstream and
downstream discharges were decreasing to zero and this was an
indication that simultaneous closure of the two automatic-
control butterfly type valves could have occurred. The team
members went out of the control room for visual inspection of
the piping system. Then they realized that an accident had
happened, and an immediate decision was taken to manually
close the upstream service valve (Fig. 1). An inspection of the
piping system showed leakages (one very large) and damages
on the piping system at two distinct locations: (1) at the sharp
U-bend (x ~ 125 metres; see Figs. 4 and 5) and (2) at the outlet
side of the horizontal PVC pipeline test section (x ~ 260 metres;
see Figs. 6 and 7).
(1) Description of damages at the sharp U-bend (Figs. 4
and 5): The U-bend was clamped with two metal supports at x
= 124.1 m and x = 125.5 m and five metal axial restraints (steel
hollow cylinders bolted to the concrete floor) which prevented
movement in horizontal direction. During the accident, the
forces on the supports were so large that four metal axial
restraints were pushed away while the fifth axial restraint got
entirely loose and fell away from the U-bend (screws were cut-
off at the floor plane). Photograph Fig. 5b shows a damaged
metal axial restraint (bolting screws were partly pulled-out and
the steel column was inclined). The downstream joint of the
PVC U-bend at x = 125.2 m displaced for about 40 mm and a
small leakage was observed here. The axial movement of the U-
bend resulted also in the partial opening of the PVC pipe in the
joint at the distance x = 119.1 m (Fig. 5a). This was possible
because the metal support at 124.1 m enabled axial
displacement of the PVC pipe for 35 mm. However, no leakage
was detected at this joint. No visible damages were detected at
the metal supports (brackets) at x = 118.8 m and upstream, and
at the metal supports at x = 125.5 m and downstream.
PVC
pipeline
Steel
pipeline
x = 256.4 m
x = 257.2 m
x = 257.9 m
x = 260.6 m
Leakage
Damaged
PVC pipe
joint
Damaged
metal supports
0. m
x = 267. m
x = 268. m
Electromagnetic
flow meter
Control valve
x = 263.5 m
x = 264. m
DN250
DN250
-4.8 m
Figure 6: Overview of damages at the outlet of the horizontal PVC pipeline test section.
7 Copyright © 2010 by ASME
a) Leakage from damaged PVC pipe joint at distance x = 257.2 m. b) Damaged metal anchor at distance x = 260.6 m.
Figure 7: Damages at the outlet side of the horizontal PVC pipeline test section.
(2) Description of damages at the outlet side of the
horizontal PVC pipeline test section (Figs. 6 and 7): The PVC
pipeline was located on the balcony of the laboratory (first
floor) and a steel pipe section with sharp downward 90° elbow
was used to divert water into the sump (ground floor). The
connection between PVC and steel pipeline was at x = 257.9 m,
and the downward elbow was at position x = 263.5 m. The
vertical part of the steel pipeline was supported by a steel
structure and was free to move in the horizontal plane and in
vertical direction (up). The horizontal part of the steel pipeline
was clamped with two metal supports at x = 257.9 m and x =
260.6 m. When the pressure upsurge was travelling through the
downward elbow, the steel pipeline was lifted-up and away
from the PVC pipeline. The steel pipe itself was undamaged but
the metal supports were deformed as it is evident from Figs. 6
and 7. Figure 7 shows that some of the bolts anchoring the
metal supports (brackets at x = 257.9 m and x = 260.6 m) to the
concrete floor were damaged. Then the PVC pipe joint at x =
257.2 m opened and a large water leakage occurred (Fig. 7a).
The pipeline upstream of the undamaged metal support at x =
256.4 m was left intact up to the metal support at x = 125.5 m
(Fig. 4). No damages were detected at the electromagnetic flow
meter and the control valve.
Brief analysis of the event was made at the spot confirming
that the accidental simultaneous closure of two automatic-
control butterfly-type valves had happened. It was found that
the incident had been caused by a fault in an electronic
conversion box due to power failure. Afterwards the
downstream-end automatic valve was modified to a manually
operated valve so to avoid a further unintentional simultaneous
closure of the valves. The damaged piping system was repaired
the very same day and the decision to continue with the
experimental programme was accepted. The next morning, on
the fifth working day, the pipeline test rig was in operation
again and the group was able to proceed with the planned
experiments.
Analysis of the accident
Before the accident, the DN150 upstream-end automatic
valve was opened 75% and the DN150 downstream-end
automatic valve was opened 100% and these settings
corresponded to an initial flow rate in the piping system of Q =
0.145 m
3
/s with an upstream gauge pressure of p
u
= 1.01 bar.
Figure 8 shows the position signals y of the automatically
operated control valves together with the upstream and
downstream flow rates Q during the accidental event. The
position signals were wired from computer via a volt-current
converter to the control-valve actuators. The position signal of
the downstream control valve y
d
was set to a constant value (full
opening) whereas the position signal of the upstream control
valve y
u
was controlled by computer in order to maintain a
desired upstream pressure p
u
set-point. The observed signals in
Fig. 8a were taken as follows: (1) y
u
was captured between the
computer and volt-current converter and (2) y
d
was captured
between the volt-current converter and the valve actuator. A
sudden drop of the signal y
d
to −25% occurred at a test time of
about 7030.2 seconds (see Fig. 8a). Because both control valves
were wired to the same volt-current converter this time can be
considered as the time at which the control valves started to
close. However, the position signal y
u
of the upstream control
valve shows adjustment of the valve opening to its maximal
opening at a time of about 7033 seconds (see Fig. 8a). This
difference is because the position signal y
u
was taken in between
the computer and the volt-current converter.
8 Copyright © 2010 by ASME
7030 7032 7034 7036 7038 7040 7042
-20.0
0.0
20.0
40.0
60.0
80.0
100.0
7030 7032 7034 7036 7038 7040 7042
-0.08
-0.04
0.00
0.04
0.08
0.12
0.16
y
u
y
d
b)
a)
y (%)
Time (s)
Q
u
Q
d
Q (m
3
s
-1
)
Time (s)
Figure 8: Automatically-operated control valve position signals (y) and water flow rates (Q) during the accidental event.
The consequent change of upstream and downstream flow
rates is even more peculiar (Fig. 8b). The downstream flow
meter detects the first change of Q
d
at time of 7030.7 seconds
and the flow is stopped at time of 7033 seconds. The upstream
flow meter detects the first change of Q
u
at the time of 7031.5
seconds and the flow is stopped at the time of 7041 seconds.
The different timing of flow changes may be attributed to
different responses of the valve actuators to the accidental loss
of the signal and to different flow conditions (boundary
conditions). It should be noted that the distance between the
upstream control valve and the upstream flow meter was about
16 metres with a pipe branch in between, whereas the distance
between the downstream control valve and the downstream flow
meter was only one metre (Figs. 1 and 2).
Figure 9 shows pressure histories at different positions
along the PVC pipeline. From these graphs one can deduce that
there were two major pressure waves in the accident: the
pressure upsurge that was travelling in the upstream direction
from the control valve at the outlet of the pipeline and the
pressure downsurge that was travelling in the downstream
direction from the control valve at the upstream end tank. Based
on these recordings one may conclude that such pressures are
possible only if both valves were closed almost simultaneously.
The pressure upsurge travelling upstream caused damage to the
piping system at various locations. Upsurge and downsurge
were superimposed when they met and this is evident from the
figures. Figures 9a to 9c show that pressure peak was cut-off at
5 bar. This is because experiments with expected maximal
pressures well below 5 bar were performed (the set maximum
for the recorded pressure signal). Extrapolation of the gradient
of the rising and dropping pressure gives a rough estimate of the
maximal pressure p
9
close to the outlet of the PVC pipe of
about 7.5 bar (Fig. 2); however, the Joukowsky pressure rise
[10] in the PVC pipeline based on the initial flow rate would be
approximately 10 bar. The difference between estimated actual
maximal pressure and the rough theoretical estimation is
attributed to the large deformations and leakages and not to
viscoelastic damping. It is worth to mention here that the PVC
part of the experimental test rig was designed to withstand a
maximal pressure of 7.5 bar. Figures 9d and 9e indicate intense
transient vaporous cavitation zones along an extended length of
the PVC pipeline.
We may conclude that the incident was caused by a fault in
an electronic conversion box due to power failure and this led
to the actual closure of both automatic control valves as it is
clear from the flow rate (Fig. 8b) and pressure traces (Fig. 9).
The short power failure damaged the volt-current converter
which sent signals to the valve actuators. Both valve actuators
were motorized and 1 mA converter output corresponded to
closed positions of the valves (the current converter box should
always supply a current in the range 1 mA to 5 mA). Immediately
after the accident a technician measured an output at the
converter of 0 mA, indicating that the current converter box had
broken down. This confirmed that in the case of loss of signal
the valves close.
CONCLUSIONS
A large-scale pipeline test rig at Deltares, Delft, The
Netherlands has been used for filling and emptying
experiments. The test rig is a horizontal 250 mm diameter PVC
pipe of 258 m length with control valves at its downstream and
upstream ends. An accidental simultaneous closure of two
automatic-control butterfly-type valves occurred during initial
testing of the test rig. Fortunately, the unintentional transient
event has been fully recorded including measurements of
pressures along the PVC pipe test section, flow rates and valve
positions. Photographs of damages to the pipeline tell their own
story. The accidental closure of both control valves induced a
large upsurge at the downstream end and a downsurge at the
upstream end. The large pressure rise due to the closure of the
downstream-end control valve caused failure of pipe supports
and leakage at pipe joints. The incident was caused by a fault in
an electronic conversion box due to power failure. Afterwards
the downstream-end automatic valve has been modified to a
manually operated valve so to avoid future accidental
simultaneous closures of valves.
9 Copyright © 2010 by ASME
7030 7032 7034 7036 7038 7040 7042
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
7030 7032 7034 7036 7038 7040 7042
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
7030 7032 7034 7036 7038 7040 7042
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
7030 7032 7034 7036 7038 7040 7042
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
7030 7032 7034 7036 7038 7040 7042
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
7030 7032 7034 7036 7038 7040 7042
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
p
9
p
8
f)
e)
d)
c)
b)
a)
p (bar)
Time (s)
p
8
p
7
p (bar)
Time (s)
p
7
p
5
p (bar)
Time (s)
p
5
p
3
p (bar)
Time (s)
p
3
p
1
p (bar)
Time (s)
p
1
p
u
p (bar)
Time (s)
Figure 9: Gauge pressures in water pipeline during the accidental event.
What do we learn from the accident?
1) Even under controlled laboratory circumstances
accidents may happen (here: power failure and
consequent closure of two automatic control valves).
2) Large axial pipe motion can damage the pipe joints.
3) In the case of a water hammer accident it is virtually
impossible to anchor pipes and bends sufficiently rigid.
4) PVC pipes themselves can withstand high transient
pressures, but the joints will fail.
5) Full records of accidents (i.e. measured data,
photographic records and many witnesses) are rare and
reported here as an illustration to pipeline engineers.
6) Assessment of the maximum anchor forces from the
measured pressure histories (not presented in this
paper).
7) Assessment of the strength of PVC pipes and joints (not
presented in this paper).
10 Copyright © 2010 by ASME
ACKNOWLEDGMENTS
The project Transient vaporous and gaseous cavitation in
pipelines carried out at Deltares, Delft, The Netherlands, was
partially funded through EC-HYDRALAB III Contract 022441
(R113) by the European Union and their support is gratefully
acknowledged. The authors would especially like to thank the
research and technical staff of Deltares for their efforts in
constructing the test rig.
NOMENCLATURE
D = pipe diameter
e = pipe wall thickness
f
s
= sampling frequency
L = length
p = pressure
Q = discharge (flow rate)
R = radius
U
x
= uncertainty in a measurement
x = axial distance
y = valve position signal
Subscripts
d = downstream end
PVC = PVC pipeline
S = steel pipeline
u = upstream end
Abbreviations
DN = nominal diameter
REFERENCES
[1] Safwat, H. H., Arastu, A. H., and Husaini, S. M., 1986,
"Generalized applications of the method of characteristics
for the analysis of hydraulic transients involving empty
sections", Paper F3, Proceedings of the 5th International
Conference on Pressure Surges, M. Papworth, ed., BHRA,
Cranfield, UK, pp. 157-167.
[2] Liou, C. P., and Hunt, W. A., 1996, "Filling of pipelines
with undulating elevation profiles", ASCE Journal of
Hydraulic Engineering, 122(10), pp. 534-539.
[3] Zhou, F., Hicks, F. E., and Steffler, P. M., 2002,
"Transient flow in rapidly filling horizontal pipe
containing trapped air", ASCE Journal of Hydraulic
Engineering, 128(6), pp. 625-634.
[4] Vasconcelos, J. G., Wright, S. J., and Guizani, M., 2005,
"Experimental investigations on rapid filling of empty
pipelines", Report UMCEE-05-01, University of
Michigan, Department of Civil and Environmental
Engineering, Ann Arbor, USA.
[5] Zhang, Y., and Vairavamoorthy, K., 2006, "Transient flow
in rapidly filling air-entrapped pipelines with moving
boundaries", Tsinghua Science and Technology, 11(3), pp.
313-323.
[6] León, A. S., Ghidaoui, M. S., Schmidt, A. R., and García,
M. H., 2009, "Application of Godunov-type schemes to
transient mixed flows", IAHR Journal of Hydraulic
Research, 47(2), pp. 147-156.
[7] Andrews, J. S., 1970, "Water hammer generated during
pipeline filling", M.Sc. thesis, Colorado State University,
Department of Civil Engineering, Fort Collins, USA.
[8] Laanearu, J., Bergant, A., Annus, I., Koppel, T., van ’t
Westende, J. M. C., 2009, "Some aspects of fluid elasticity
related to filling and emptying of large-scale pipeline",
Paper H4, Proceedings of the 3rd IAHR International
Meeting of the Workgroup on Cavitation and Dynamic
Problems in Hydraulic Machinery and Systems, P. Rudolf,
ed., Brno University of Technology, Brno, Czech
Republic, II, pp. 465-474.
[9] Coleman, H. W., and Steele, W. G., 1989,
Experimentation and uncertainty analysis for engineers,
John Wiley & Sons, New York, USA.
[10] Wylie, E. B., Streeter, V. L., 1993, Fluid transients in
systems, Prentice-Hall, Inc., Englewood Cliffs, USA.
