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

net/publication/303417359

High Speed Imaging of Water Hammer with Column Separation, BHR Group,
Proc. of the 12th Int. Conf. on Pressure Surges, Dublin, Ireland, November
2015, pp. 47-57, ISBN 9781-85598-15...

Conference Paper · November 2015

CITATIONS READS

0 38

3 authors, including:

Tobias Traudt Chiara Manfletti

German Aerospace Center (DLR) European Space Agency
22 PUBLICATIONS   13 CITATIONS    84 PUBLICATIONS   200 CITATIONS   

SEE PROFILE SEE PROFILE

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

LUMEN (Liquid Upper stage deMonstrator ENgine) View project

Laser ignition for liquid propulsion rocket engines View project

All content following this page was uploaded by Tobias Traudt on 05 December 2018.

The user has requested enhancement of the downloaded file.

 High Speed Imaging of Water Hammer with Column
Separation

Tobias Traudt, Cristiano Bombardieri, Chiara Manfletti

DLR Lampoldshausen, Department of Rocket Engines, Germany

ABSTRACT

Previous tests at a test bench in Lampoldshausen revealed a phenomenon in which the

first peak of the pressure surge was not the highest one. Instead the second peak exceeds
the first peak by more than 25%. Other tests showed pressure spikes in the cavitation
area. In an attempt to explain both phenomena, a high speed imaging setup was installed
at the test bench. The resulting images will be presented in this paper, together with the
bubble position over time. The experiments indicate, that the second phenomenon is
related to the flow reversal prior to column reintegration of the cavitating liquid column.

1 INTRODUCTION

Water hammer is of strong interest in many industrial fields, amongst which the space
industry. In rocket engines and thrusters it plays a major role in the designing process of
the feed system and has to be considered especially when the feed lines are primed
during start-up as well as during the rapid closing of valves upon shutdown. In both cases
a pressure peak will occur, leading to a water hammer wave travelling along the pipe.
Water hammer is a well-studied phenomenon because of its importance for the designer
of fluid systems. For CFD tools it is a valuable validation experiment, to check whether
the code is able to correctly simulate steep pressure gradients, absolute pressure peaks
and wave attenuation. When the pressure falls below the saturation pressure and hence
column separation occurs, the complexity of flow simulation increases significantly
because of instantaneous evaporation and condensation. In order to investigate these
phenomena at pressures of up to 100 bar and Reynolds numbers of up to 105 a test bench
was built at the DLR Lampoldshausen. Tests were performed with water to produce a
pressure surge upon valve closing.
In the next chapters we will present the test bench setup and the experimental method.
We will show the reproducibility of the experiments and after that we will summarize
findings from past experiments that show two different effects. One causes the second
water hammer peak to be 25% higher than the first one, while the other one shows small
pressure spikes in the cavitation area of the water hammer oscillation [1].
Since both phenomena appear only in tests with cavitation, it became obvious that optical
access to the test section is mandatory to understand how and where the liquid evaporates
and how the vaporous regions reintegrate; hence a high speed imaging setup was
installed. The high speed imaging setup with the optical segments will be presented. We
will show a time series of images of the evaporating liquid and discuss the results.
Finally the time evolution of the bubble position will be presented in order to give an
estimate for the bubble velocity along the pipe axis.
2 EXPERIMENTAL METHOD

Testing was conducted using the newly constructed test bench M3.5. The test bench was
designed to investigate water hammer and priming phenomena. The main parts of the test
bench are the two tanks and pressurization system, the test section, and the fast acting
valve (Figure 1). The fast acting valve is coaxial and is pneumatically actuated. By
controlling the working pressure of the valve the closing time can be regulated. The
closing time is recorded using the valve’s position encoder. In the tests discussed in this
paper, the measured valve closing time was 22ms.

Figure 1: Schematic of Test bench M3.5

The pressurization system uses nitrogen or helium. Setting a differential pressure

between the tanks controls the initial flow speed in the test section. Automatically
controlled pressure regulators and discharge valves at both tanks ensure that the pressure
level is within ±0.4% prior to valve closing.
The test bench is equipped with a Coriolis flow meter for mass flow measurement. For
the mass flows presented in this paper, the device is precise to <0.1%. The flow meter is
installed downstream of the fast acting valve.
The main dimensions of the test section are shown in Table 1. The longest segment of the
test section is wound into a spiral with a diameter of 1.25m. The material of the test
section is stainless steel of grade 1.4541. To limit its movement, it is mounted onto a
rigid support structure at seven points. The support structure is made of aluminium
profiles. The test section is mounted with an upward slope of about 1°.

Table 1: Dimensions of test bench

Description

Test section length, 𝐿𝐿𝑡𝑡𝑡𝑡 8.34m

Test section inner pipe diameter, 𝑑𝑑𝑖𝑖,𝑡𝑡𝑡𝑡 19mm

Test section wall thickness, 𝑒𝑒𝑡𝑡𝑡𝑡 1.5mm

Sensor distance from valve seat, Position 1 0.3m
Sensor distance from valve seat, Position 2 7.41m
Sensor distance from valve seat, Position 3 8.11m

The segments of the test section are connected by flanges. In between the flanges 20mm
long disks, with the same inner diameter as the test section, are installed in which three
sensors are mounted. Each disk houses one type K thermocouple, one piezo-resistive
static pressure sensor (Kistler 4043-A100 type) and one piezo-electric dynamic pressure
sensor (Kistler 601A). The static pressure sensors and the dynamic pressure sensors are
flush mounted. The sampling rate for the static pressure sensors is set to 10 kHz and for
the dynamic pressure sensors it is set to 150 kHz. To avoid aliasing, Butterworth filters in
the amplifiers are set to 2 kHz and 30 kHz (-3dB) respectively.
Three sensor disks are installed at positions upstream of the fast acting valve. The
distance of each set of sensors to the valve seat of the fast acting valve is listed in Table
1.
The upstream tank has a flanged connection with an inner diameter of 98.3mm (DN100).
The inlet from the side facing flange to the test section has rounded edges with a radius
of 5mm. The length of the DN100 part inside the tank is 218mm.

2.1 High Speed Imaging Setup

A Photron Fastcam SA-X was used for image acquisition. The following settings were
used for all videos presented in this paper:

• Frame rate: 10,000 fps

• Shutter: 1/18604s

Two optical segments were installed in the test section for optical access (see Figure 1).
One is located next to the fast acting valve, while the other is close to the high pressure
tank. Both were of the same type (see Figure 2). The one close to the valve was used for
image recording.
Table 2 shows the positions and the dimensions of the optical segments.

Table 2: Dimensions of quartz glass segment

Description

Distance from valve seat to beginning of glass 0.313m

segment at the valve
Distance from valve seat to beginning of glass 7.15m
segment at the tank
Length of optical segment, 𝐿𝐿𝑔𝑔𝑔𝑔 0.25m

Optical segment outer diameter, 𝑑𝑑𝑜𝑜,𝑔𝑔𝑔𝑔 25mm

Optical segment inner diameter, 𝑑𝑑𝑖𝑖,𝑔𝑔𝑔𝑔 19mm

Diameter of flange 105mm

Thickness of flange 20mm
 Figure 2: Quartz glass segment

The backlight was realized by a flashlight of type LedLenser X21 and an optical diffusor.
Both were placed in the line of sight of the camera behind the quartz glass segment. The
recorded images are 710 x 115 pixels in size. The resolution is 0.28mm/pixel.

2.2 Test Procedure

Before each test day the upstream tank is filled with water and the test section is purged
to remove any entrapped air. This is necessary for several reasons. Air in the test section
lowers the speed of sound, it cushions the column reintegration after a cavitation event
and bubbles of air in the flow will act as nucleation sites during cavitation. Due to the
latter, instantaneous evaporation might happen at different places than without a bubble
of noncondensable gas in the flow.
The water in the test bench is stored overnight at ambient pressure to ensure that there are
no non-dissolved gas bubbles left in the water. For testing, the pressure in the tanks is set
to the desired values, the fast acting valve is opened and the water starts to flow. Before
the valve is closed the water is allowed to flow for a few seconds in order to establish
stationary flow. After each test, gas which might have come out of solution, is removed
from the valve seat of the fast acting valve by a purging valve.

Figure 3: Repeatability WH with cavitation

The tests are repeated at least three times to examine the reproducibility of the results.
Good reproducibility was achieved for all tests presented in this paper. As an example,
Figure 3 shows the pressure evolution of the static pressure sensor at position 1.
The test condition for these tests and all the following tests are summarized in Table 3.

Table 3: Test settings

Test # Mean tank pressure Initial flow velocity
(bar) (m/s)

4, 5, 6 24 4.13
9 1.96 2.08
20 23.78 3.14

In test #9 high speed imaging was used for flow visualization.

3 EXPERIMENTAL FINDINGS IN PRESSURE EVOLUTION

In past research at the M3.5 test bench [1] two unexpected phenomena were observed.
The first phenomenon manifested itself in consecutive pressure spikes in the cavitation
area of a water hammer experiment with column separation (see Figure 4). The pressure
spikes appear in each cavitation area after every water hammer peak and to the
knowledge of the authors, they have not been observed in published experiments before.

1
Figure 4: Short duration spikes in cavitation area (dynamic pressure sensor, position 1)

The pressure spikes are of very short duration. A comparison of the static pressure sensor
readings with the dynamic pressure sensor readings showed that the higher sampling rate

1
The dynamic pressure sensors are not able to do absolute measurements. The zero value in
dynamic sensor readings is actually the mean tank pressure before the test. As a result negative
pressure in the diagram is possible.
of the latter (150 kHz compared to 10 kHz of the former) is necessary to acquire the
correct pressure of the spikes (see Figure 5).

Figure 5: Comparison of static pressure sensor at position 1 (left) and dynamic pressure
sensor at position 1 (right)

The second phenomenon causes the second peak of the water hammer wave to be 25%
higher than the first one (see Figure 6 and [1]). This is especially surprising since the first
peak already reaches the Joukowsky pressure. The effect is reproducible and it is only
visible when the flow is cavitating.

According to Bergant [2] and Gayed [3] the reason for this phenomenon is the additive
interference of the primary water-hammer wave and a secondary water-hammer wave.
The primary wave travels from the valve to the reservoir where it is reflected and moves
to the valve again. At the valve it is reflected as a negative wave which causes the liquid
to evaporate and form a cavity. Now the cavity and the reservoir are the boundary
conditions for the primary water-hammer wave traveling along the pipe.

Figure 6: 2nd peak 25% higher than 1st peak

Each time the wave arrives at the valve the velocity of the liquid column is decreased by
a specific amount. At some point the flow reverses and moves towards the valve again.
This causes the cavity to decrease in size and when the cavity collapses a secondary
water-hammer wave originates at the valve. The pressure peak caused by cavity collapse
is lower than that of the first water-hammer peak (see arrow in Figure 6). If the phase of
the primary and secondary waves is such that they add up, the short duration peak in
Figure 6 will form and a pressure higher than the Joukowsky pressure is possible.

The two phenomena are clearly related to cavitation, so it was decided to install an
optical segment for high speed imaging, in order to visualize them.

4 HIGH SPEED IMAGING OF TIME EVOLUTION OF CAVITATING

BUBBLES

Figure 7 shows the pressure evolution of test #9 with circles indicating the time in
seconds of the individual frames shown in Figure 8.
Clearly visible in Figure 7 are the spikes in the area where the pressure is near the
saturation pressure and the flow is hence cavitating. As pointed out before, these spikes
are caused by the primary water-hammer wave travelling through the liquid column.

Figure 7: Circles indicate times of the frames in Figure 8

In Figure 8 the flow is entering from the right before valve closure and the fast acting
valve is located left of the quartz glass segment. At 0 seconds the command to close the
valve is sent from the control system.
Please note, that the time between the frames in the figure is not constant, in order to
show interesting phenomena in more detail.

With the high speed images it is possible to see that cavitation in the transparent section
starts at 0.1322s and vapour areas grow until 0.1938s (Figure 8).
Figure 8: Time evolution of cavitation in water hammer experiment; numbers indicate
time in seconds, flow direction prior to valve closure was from right to left, arrows
indicate start of visible bubble collapse caused by pressure wave
The first visible bubble collapse is indicated by the small arrows in Figure 8. The three
waves are also visible in the pressure history. The frames at 0.2724s and 0.2806s show
no cavitation, because the second water hammer peak raises the pressure above the
saturation pressure.
The image sequence also shows, that after each bubble collapse the number of bubbles
increases, while the size of the bubbles decreases. This might be related to outgassing of
dissolved gases, while the liquid is cavitating. According to Perko [4] and Bergant, et al.
[2] the adsorption/desorption of gases into liquids occurs on longer time scales than the
almost instantaneous evaporation. Thus it can be assumed that the desorbed gases are not
fully absorbed before the next cavitation round takes place and the small bubbles of
undissolved gas then act as additional evaporation nuclei. The small bubbles of desorbed
gas are not visible in the video because their size is below the resolution of the video
images which is 0.28mm².
It is obvious that there is a delay between time 0.1958 (see Figure 8), which indicates the
first optical appearance of the pressure wave, and the position of the first pressure spike
in the pressure reading. This delay is 8ms and it is easily explained by the location of the
sensor being 19.8cm downstream of the first optical appearance of the pressure wave.

There was no visible cavitation at the second optical segment further upstream.

By comparison with literature (Bergant, et al. [2] and Gayed [3]), it becomes clear that
the small pressure spikes are caused by the primary water hammer wave which travels
through the liquid column several times, as long as cavitation takes place. As such they
are also a visualization of one of the sources for the phenomenon in which the second
water-hammer peak is higher than the first one.

In literature the wave propagation speed 𝑐𝑐 is usually treated as constant when considering
secondary waves in a pipe. This is not correct as the varying time intervals between the
spikes show. The difference in speed of sound is explained by the distributed cavitation
visible in Figure 8. Different amounts of vapour in the pipe result in different wave
propagation speeds.

5 POSITION EVOLUTION OF VAPOUR BUBBLES

The optical segment made it possible to track the positions of individual bubbles and give
an estimate for the velocity at which the bubbles travel close to the valve. The velocities
obtained represent also the liquid velocity when one assumes a homogeneous flow
velocity of liquid and bubbles.

To obtain the bubbles’ velocity, the image was transformed into a black-and-white image
and then the centroid of the resulting bubble area was calculated and tracked. Only the
biggest bubble was considered for the analysis. This choice was made because the other
bubbles visible in the optical segment moved in the same direction, at the same speed.

Figure 9 shows the position of the bubble in axial direction for two consecutive water
hammer oscillations. The position units are relative to the left edge in the video frame.
One can see that the bubble travels away from the valve first. The reason is the kinetic
energy of the liquid column, which causes the liquid column to move away from the
valve. The kinetic energy is large enough to cause cavitation and the increasing demand
in volume of the vapour bubbles, allows a macroscopic movement of the liquid column.
 Figure 9: Position of biggest bubble in axial direction

It is interesting that the velocity of the bubble is almost constant over time, resulting in a
linear slope of the position. The bubbles move away from the valve with a value of about
1.45m/s, but as soon as the pressure wave hits the bubble, it moves in the opposite
direction with a value of about 1.3m/s.
After the second water hammer peak, the velocity of the bubble is smaller with a value of
about 0.5m/s, but this time, the velocity after the change of direction is higher than before
(1.3m/s). The reason for this behaviour is unclear.
When the curve drops to zero, this means that no bubble is recognized anymore by the
algorithm. These points coincide with the bubble collapse due to the pressure waves
travelling through the liquid column.

The high speed images confirm the fact that the flow reversal is an inherent phenomenon
for water hammer with cavitation. It is also visible in experiments done by Adamkowski
[5], but he didn’t record the intermediate pressure spikes. The latter might be related to
the transmitted frequency range and type of sensors used in his work (0-2kHz) or the
different test setup as the one used in this study.

We discussed the reason for the flow reversal before. Each time the primary water
hammer wave arrives at the valve, the velocity of the liquid column is decreased. In our
case the first arrival of the wave reverses the flow.

Looking at the phenomenological behaviour of the liquid column after the first pressure
peak, the flow reversal can be explained as follows: The liquid column moves away from
the valve (at a microscopic length scale), causing the pressure drop after each pressure
peak. When the pressure reaches the saturation pressure the liquid starts to evaporate and
bubbles appear in the liquid. The increasing volume demand by the vapour bubbles allow
for a macroscopic movement of the liquid column away from the valve. This volume
needs to be filled by liquid again, to allow for another pressure peak. Hence the liquid
column has to travel in the direction of the valve again, decreasing the bubble volume. As
soon as the bubble volume is zero, the pressure starts to rise for the second pressure peak
and the process is repeated.
 Our experiments show the acoustic nature of flow reversal. It is not an accelerated
process where the liquid column changes its velocity slowly, instead changes in velocity
happen instantaneously, as indicated by Bergant [2] and Gayed [3].

6 CONCLUSION

Water hammer experiments were performed with an optical segment made of quartz
glass in order to find the explanation for two phenomena which occurred in previous
tests.
The origin of the first phenomenon, which causes the second water hammer peak to be
up to 25% higher than the first one, could not be observed directly in the high speed
images.
The second phenomenon manifested itself in small pressure spikes in an area where only
saturation pressure is expected. The spikes peak values were only correctly resolved by
the 150 kHz sampling rate of dynamic pressure sensors installed at the test bench. By the
high speed imaging it was possible to see how these pressure waves moved through the
liquid column, causing the bubbles to collapse temporarily.
The velocity evolution of the biggest bubble was measured with a bubble tracking
algorithm over two consecutive water hammer oscillations. The propagation velocity of
the bubble is almost constant till the flow is reversed as soon as the first pressure wave
travels through the cavitating liquid.
The pressure spikes are responsible for the instantaneous deceleration of the liquid
column.

7 REFERENCES

[1] T. Traudt, C. Bombardieri and C. Manfletti, “Influences on Water Hammer Wave

Shape an Experimental Study,” in 63. Deutscher Luft- und Raumfahrtkongress,
Augsburg Germany, 2014.
[2] A. Bergant, A. Simpson and A. S. Tijjsseling, “Water Hammer with Column
Separation: A Historical Review,” Journal of Fluids and Structures, vol. 22, pp. 135-
171, 2006.
[3] Y. K. Gayed and M. Y. M. Kamel, “Mechanics of Secondary Water-Hammer
Waves,” Proceedings of the Institution of Mechanical Engineers, vol. 173, no. 1, pp.
675-686, 1959.
[4] H.-D. Perko, “Gasausscheidung in instationärer Rohrströmung,” Ph.D. Dissertation,
Universität Hannover, 1985.
[5] A. Adamkowski and M. Lewandowski, “Improved Numerical Modelling of
Hydraulic Transients in Pipelines with Column Separation,” in 3rd IAHR
International Meeting on Cavitation and Dynamic Problems in Hydraulic Machinery
and Systems, Brno, Czech Republic, 2009.
[6] A. Bergant, A. Tijsseling, J. Vitkovsky, D. Covas, A. Simpson and M. Lambert,
“Parameters Affecting Water-Hammer Wave Attenuation, Shape and Timing,”
Journal of Hydraulic Research, vol. 46, pp. 373-391, 2008.

View publication stats