Measurement Science and Technology

Related content
- Ultrasonic pulse echography for bubbles
Development of an ultrasonic void fraction profiler traveling in the proximity of a wall
Hyun Jin Park, Yuji Tasaka and Yuichi
Murai
To cite this article: Yuichi Murai et al 2009 Meas. Sci. Technol. 20 114003
- Integration of conductivity probe with
optical and x-ray imaging systems for local
air–water two-phase flow measurement
Dewei Wang, Kyle Song, Yucheng Fu et
al.
View the article online for updates and enhancements.
- Planar shadow image velocimetry for the
analysis of the hydrodynamics in bubbly
flows
D Bröder and M Sommerfeld

Recent citations
- Investigation of the void fraction–quality
correlations for two-phase hydrogen flow
based on the capacitive void fraction
measurement
Yuki Sakamoto et al

- Void fraction measurement using an

imaging and phase isolation method in
horizontal annular flow
Pengman Niu et al

- Bubbly drag reduction investigated by

time-resolved ultrasonic pulse echography
for liquid films creeping inside a turbulent
boundary layer
Hyun Jin Park et al

This content was downloaded from IP address 143.107.225.16 on 04/04/2020 at 20:49

IOP PUBLISHING MEASUREMENT SCIENCE AND TECHNOLOGY
Meas. Sci. Technol. 20 (2009) 114003 (13pp) doi:10.1088/0957-0233/20/11/114003

Development of an ultrasonic void

fraction profiler
Yuichi Murai1 , Shoko Ohta, Akinari Shigetomi, Yuji Tasaka and
Yasushi Takeda
Division of Energy and Environmental System, Graduate School of Engineering, Hokkaido University,
N13W8, Sapporo 060-8628, Japan
E-mail: [email protected]

Received 29 January 2009, in final form 26 July 2009

Published 18 September 2009
Online at stacks.iop.org/MST/20/114003

Abstract
A method for measuring the spatial distribution of the void fraction in bubbly two-phase flow
is presented. The method is referred to as ultrasonic void fraction profiling since it is based on
the signal processing of an ultrasonic pulse scattering on a bubble interface. The method is
established using two processes for the ultrasonic sensing of bubbles. One approach is to
detect the bubble interface along a measurement line, i.e. the path of the ultrasonic pulse in the
liquid. The interface is captured using two types of signal-processing schemes: the echo
intensity method and the Doppler method. The other approach is to reconstruct the void
fraction profile from the number of bubble interfaces. A theoretical formula for the estimation
is proposed by considering ultrasonic reflection in a suspension of bubbles. The validity of the
formula is examined with theoretical and numerical bases. Finally, the method developed here
is applied to four flow configurations for the demonstration, in which the void fraction profile
governs the modulation of liquid flow field.

Keywords: ultrasound, Doppler method, bubble, void fraction

(Some figures in this article are in colour only in the electronic version)

1. Introduction visualization using x-rays (Bieberle et al 2008) and neutrons

(Mishima et al 1997). If the path of fluids is visible, then
The distribution of the void fraction is one of the most the void fraction can be measured using optical imaging and
fundamental and critical aspects of many two-phase flow lasers aided by digital image processing techniques (Murai
systems. In pipe flows, for example, the flow pattern and et al 2001, 2005, 2007, Okawa et al 2007, Schleicher et al
pressure loss are significantly affected by the radial distribution 2008).
of the void fraction. In drag-reducing strategy with bubbles, it In our study, a new and easily used method is developed
is important to assess the position of the peak void fraction in using the physics underlying the ultrasonic pulse around
the direction perpendicular to the boundary layer (e.g. Ferrante the gas–liquid interface. Since ultrasound can penetrate
and Elghobashi 2004). The performances of aeration devices an opaque wall and transmit via a non-transparent fluid,
and fluid machineries are also quite sensitive to the spatial the measurement is performed free of optical and electrical
void distribution, and hence they need to be measured with restrictions. Obviously, the use of weak ultrasound as a way
acceptable precision. There are several void measurement of acquiring information is much easier than the use of x-rays
techniques that have been developed and used to date: electric or impedance tomography. Thus, the measurement target of
conductance probing (Neal and Bankoff 1963, Kim et al the flow does not need to be modified to provide a specific
2000), the impedance method (Yang et al 2003), optical measurement environment. In comparison with optical fiber
fiber probing (e.g. Julia et al 2005) and radiographic void probing, the target flow will not be disturbed. Thanks to
ultrasonic properties, the possible target of measurement is
1 Author to whom any correspondence should be addressed. quite wide such as for liquid metal, liquid food, magnetic

0957-0233/09/114003+13$30.00 1 © 2009 IOP Publishing Ltd Printed in the UK

Meas. Sci. Technol. 20 (2009) 114003 Y Murai et al

liquid, heavy oil and so forth. In fact, the present ultrasonic but it does not involve a macroscopic empirical model such as
void fraction profiler can be applied quickly in the region of that using an attenuation coefficient. The second half of the
interest along working pipelines and in operating two-phase paper deals with the experimental application to four different
flow facilities. Simply mounting the device on the outside two-phase flow configurations, in which the feasibility of the
of a structure enables measurement without time-consuming present technique is well demonstrated.
preparation. Moreover, the present method is a calibration-
free technique, being different from indirect measurements.
2. Measurement principle
In terms of the ultrasonic sensing of gas–liquid two-phase
flow, Xu and Xu (1997) reported on ultrasonic tomography 2.1. Reflection of an ultrasound pulse at an interface
for identifying flow regimes. Use of multiple ultrasonic
frequencies was proposed by Murakawa et al (2008) to The ultrasound used here has an acoustic intensity sufficiently
distinguish bubbles from small solid particles. By introducing weak for the ultrasound to be assumed a linear wave that does
an ultrasonic intensity decay function inside the bubbly not interact with the flow during propagation and reflection.
mixture, Supardan et al (2007) developed ultrasonic computer Under these conditions, when a single ultrasound pulse
tomography, which was successfully applied to measuring is emitted toward a gas–liquid interface, the phase of the
the two-dimensional distribution of void fractions in a bubble ultrasound is reversed during reflection as shown in figure 1.
column. They adopted a continuous logarithmic attenuation The figure represents the spatio-temporal two-dimensional
of the ultrasound intensity in the mixture to estimate the void pressure fields with bright and dark bands indicating positive
fraction profile. Such a methodology has a principle similar to and negative pressure fluctuations. These are obtained by
that of electrical capacitance tomography (Warsito and Fan solving the simplest one-dimensional linear wave equation.
2001). The actual attenuation of ultrasound in a bubbly The value K is the number of waves (or cycles) in a single pulse.
mixture depends on multiple factors such as the bubble size The acoustic intensity of the reflection wave depends on the
relative to the wavelength of the ultrasound, the size relative difference in the acoustic impedance between gas and liquid.
to the ultrasonic beam diameter and the complex reflection of Thus, from the intensity of the echo, we know if the object is a
ultrasound among clustered bubbles. Moreover, the bubble gas bubble (Chen et al 2005). At the same time, the position of
volume oscillates to emit a resonant echo, i.e. there is bubble the bubble is obtained from the time of flight of the ultrasound
pulsation when the bubble is subject to strong ultrasound pulse from the emission to the receiving at the ultrasonic
(Krasovitski and Kimmel 2001). Ultrasound spectroscopy transducer. This type of bubble distribution measurement is
uses this resonance to measure the bubble size distribution. referred to here as the echo intensity technique. This technique
In this study, we developed a new methodology for is very similar to the medical ultrasonic echo method, which
acquiring the spatial distribution of the void fraction in gas– visualizes the internal organs of the human body. Shelled
liquid two-phase flow by considering the reflection of a weak microbubbles have been applied as the ultrasonic contrast
ultrasonic pulse on an individual bubble interface. In general, agent for recent medical echo instruments. However, the
the ultrasonic sensing of the gas–liquid interface is realized echo intensity technique applied in fluid engineering fails to
with the fact that there is a large difference in the acoustic distinguish bubbles and solid particles that produce similar
impedance between gas and liquid phases. There are two echo intensities in liquid. Furthermore, the technique is not
well-known interface detection methods. One method is applicable to bubbles having a diameter larger than the length
an ultrasonic intensity technique in which the interface is of a single ultrasonic pulse. In such a case, the ultrasonic
detected from the amplitude of the ultrasonic wave that is pulse will mirror-reflect at the interface and propagate in a
reflected from the interface (e.g. Murakawa et al 2005). In this straight line away from the transducer. The transducer does
technique, the amplitude depends on the bubble size relative to not capture any information in such a case. The application of
the ultrasonic wavelength as explained by Mie and Rayleigh the technique to a large interface is restricted to the case that
scatterings of waves on dispersion having a finite volume. the interface is accompanied by turbulent waves that induce
The other method is to use information of the ultrasonic a diffused reflection of the ultrasound. The dimension of
frequency in reflected pulses, i.e. the Doppler technique (Murai a large bubble referred to here is larger than approximately
et al 2006). This is suitably applied to an interface that is 4 mm (around 10 wavelengths) when 4 MHz ultrasound in
longer than the ultrasonic beam diameter. In this paper, we water is assumed and is the most common bubble size found
focus on the following problem frequently encountered in the in fluid engineering. To sensitively detect such a bubble,
ultrasonic sensing of interfaces and bubbles. An ultrasonic an alternative method is required to universally detect large
pulse mostly reflects at the interfaces distributed closest to the bubbles in arbitrary flows.
ultrasonic transducer and does not sensitively capture deeper We therefore proposed a technique based on the Doppler
void information. When the closest interface (or bubble) is effect of the ultrasonic pulse in a previous paper (Murai et al
correctly captured either using the echo intensity technique or 2006). This technique requires the seeding of small tracer
the Doppler technique, the next problem is how to reconstruct particles into the target flow field; however, any type of gas
the void fraction. This topic is discussed in the first half bubble is correctly captured using the information hidden in
of the paper. The technique we use is deterministic rather the delta region of the reflection wave (see the triangular
than statistical. Hence the signal processing is formulated on a region near the bubble interface in figure 1). In this delta
discrete domain based on the given spatio-temporal resolution, region, the advancing and returning waves co-exist so that a

2
Meas. Sci. Technol. 20 (2009) 114003 Y Murai et al

Figure 2. Reflection of ultrasound at bubbles closest to transducer

for densely distributed bubbles. Gray dots are the existing bubbles
whereas white dots are the bubbles detected by ultrasound pulse.
Figure 1. Reflection of pulsed ultrasound on the gas–liquid
interface. White and gray stripes stand for positive and negative
local pressure, which forms the local standing wave near the it extremely weakens as it crosses those bubbles on its return
interface as identified by the checker flag pattern. to the transducer. Thus, we need to abandon the idea of using
information of the far distance region. Only the distribution
of the first-reflecting bubbles can be obtained with sufficient
local standing wave forms. Hence the particles existing in signal quality. As shown on the right side in the figure, the
the delta region do not produce a Doppler shift frequency true void fraction profile, α(y), and the profile of detected
because of the cancellation of the two counter waves. The bubbles, β(y), are not identical. We refer to the quantity β(y)
gas–liquid interface can be detected by a null region in the as the detected void fraction in this paper. In the following,
spatio-temporal two-dimensional pattern of the Doppler shift the relationship between these two profiles is explained in
frequency. The ultrasonic velocity profiler (UVP) measures detail.
the Doppler shift frequency in reflected waves to determine Figure 3 illustrates the enlargement to explain the
the velocity of the fluid (Takeda 1986, 1995). Thus, the UVP relationship in the digitalized domain. The vertical direction
is straightforwardly applied to interface detection. In the case indicates the measurement distance from the transducer at a
of a 4 MHz four-cycle ultrasonic pulse, the thickness of the given spatial resolution while the horizontal direction indicates
null Doppler delta region ranges from 0.75 mm to 1.00 mm in the time at a given sampling rate. The gray cells are space–
water at atmospheric pressure. This thickness corresponds,
time cells inside the liquid phase and the dark cells are the
at the same time, to the spatial resolution of the Doppler
first bubbles that reflect the ultrasound. White cells are
technique for interface detection (Suzuki et al 2002). Utilizing
cells unavailable for measurement, where no information is
ultrasound with higher frequency or with a smaller number of
obtained by the ultrasound. The value AC (available cells)
cycles gives a better spatial resolution as shown in figure 1.
is the number of gray and dark cells at each depth from
This technique is established only for an ultrasonic pulse and is
the ultrasonic transducer, the value DB (detected bubbles)
not available for continuous-wave (CW) ultrasound since the
is the number of dark cells at each depth and the value VF
delta region expands infinitely for the CW ultrasound (Longo
(void fraction) is the void fraction at each depth, which is the
2006).
ratio of DB to AC. Once the value of DB is measured, the value
In summary, there are two techniques for capturing the
of AC is also obtained because the cells before the first bubble
gas–liquid interface using an ultrasonic pulse. The echo
should be liquid cells. Therefore, the relationship between
intensity technique can be used for small spherical bubbles and
the true void fraction α and the detected void fraction β in
the Doppler method for large bubbles. The former technique
digitalized form can be expressed by
uses diffused reflection while the latter uses mirror reflection
of the ultrasonic pulse. In this study, we use a UVP (UVP- β2
DUO, MetFlow S.A., Switzerland) as the hardware platform α1 = β1 , α2 = ,
1 − β1
for signal processing. We developed all software used in the (1)
β3 β4
study. α3 = , α4 = ,···.
1 − β1 − β2 1 − β1 − β2 − β3
The first equation means that the two void fractions take the
2.2. Void fraction distribution along the ultrasonic path
same value at the first position from the transducer. The next
In an actual situation, the gas–liquid interface in two-phase equation expresses that the true void fraction at the second
flow is measured by combined use of the echo intensity and position is estimated by the ratio of the detected void fraction,
Doppler techniques. However, this approach only detects β 2 , to 1 − β 1. This denominator is smaller than unity because
the bubbles closest to the ultrasonic transducer as shown of the obstruction due to the bubbles located in the first
by white plots in figure 2 while gray plots indicate existing position. In the same manner, the void fraction at the third
bubbles. Since an ultrasonic pulse cannot propagate regularly and the other deeper layers is expressed as written above. The
beyond bubbles larger than the ultrasonic beam diameter, little description of these relationships between α and β implicitly
information is obtained beyond those bubbles. A certain assumes that the probability of finding a bubble behind a
intensity of ultrasonic pulse penetrates the bubbles; however, previously detected bubble is the same as in the bulk of the

3
Meas. Sci. Technol. 20 (2009) 114003 Y Murai et al

Figure 3. Digitalized representation of void fraction distributed in the space–time two-dimensional domain.

flow. Thus, the relationship at an arbitrary depth coordinate i In summary, equation (6) is used to estimate the detected
is written as void fraction from the true void fraction while equation (3)
⎛ ⎞−1
gives the true void fraction from the detected void fraction
i−1
αi = βi ⎝1 − βj ⎠ . (2) that is measured by the ultrasonic pulse. Obviously, these
j two equations are symmetric to one another, and the solutions
are perfectly reversible as long as the detected void fraction
The conversion of this equation to a continuous system with
coordinate y yields satisfies the following condition:
  −1 
1 y 1 y
α(y) = β(y) 1 − β(y) dy , (3) β(y) dy  1. (7)
δ 0 δ 0
where δ is the interval distance of two data in the y-direction The above condition is always satisfied if the function β(y)
and corresponds to the spatial resolution of the measurement. is defined by the distribution of the first-detected bubbles, i.e.
The equation becomes a function of δ because of the definition the ultrasound detects a single bubble located closest to the
of β. That is, the detected void fraction is defined as the transducer. In other words, the function β(y)/δ is defined as the
ratio of the number of detected bubbles to the total sampling probability density distribution of the first-detected bubbles,
number within a given finite depth δ in the spatial direction. the full-range integral of which is unity.
It is noted that this equation is derived in a deterministic
manner without the introduction of any physical model or 2.3. Analytical relationship between two functions
empirical assumption. For instance, the ultrasonic attenuation
coefficient is not used in the equation. We do not invoke any The meaning of equation (3) is explained with a numerical
probability theory in this derivation. Using equation (3), the simulation using ten types of functions. Figure 4 shows the
true void fraction profile α(y) is calculated from the detected various relationships between the detected void fraction β
void fraction profile β(y). Since the range of integration in (shown by gray profiles) and the true void fraction α (shown
equation (3) is from zero to y, the true void fraction profile is by white profiles), which decrease with the distance from the
explicitly determined by integrating the detected void fraction transducer. The ordinate represents the dimensionless distance
from the origin. This very light computational load allows us y/δ. Five graphs show the difference in the relationship
to realize real-time monitoring of the void fraction profile. The when a peak void fraction of 20% is set at the transducer
distance y is obtained from the time of flight of the ultrasonic (or the starting point of the measurement). A quick look
pulse; y = C × t/2, where C and t are the speed of sound and gives the impression that the function β(y) is insensitive to
the time elapsed since the emission of the ultrasound pulse differences in the function α(y). However, the function α(y)
from the transducer. is exactly reconstructed by integrating the β(y) profile, using
In the same manner, the inverse relationship at each depth equation (3) to show a significant difference.
is derived mathematically as The relationship between these two functions can be
β1 = α1 , β2 = α2 (1 − β1 ) = α2 (1 − α1 ) , analytically solved in the case of a simple function such as
(4) those in table 1. The detected void fraction for a constant void
β3 = α3 (1 − β1 − β2 ) = α3 (1 − α2 ) (1 − α3 ) · · · .
fraction profile exponentially decays, which is conceptually
Therefore, the inverse relationship at an arbitrary depth is similar to logarithmic ultrasonic attenuation. However, it
written as differs in that the void fraction has an inclined distribution.

i−1
  Figure 5 shows the same relationships for non-constant
βi = αi 1 − αj . (5) profiles of the true void fraction. In the case of a Gaussian
j =1
profile, the peak value of β(y) is smaller than the peak value
By conversion to a continuous function, equation (5) is of α(y), and the peak value is at a position shallower than the
rewritten as  y  position of the peak value of α(y). When two local peaks exist
1 within the measurement line, the peak value at the greater depth
β(y) = α(y) exp log {1 − α(y)} dy . (6)
δ 0 is considerably smaller than the first peak value. However, the

4
Meas. Sci. Technol. 20 (2009) 114003 Y Murai et al

(a) uniform (b) parabolic (c) linear (d ) square root (e) delta

Figure 4. Relationship between true void fraction and detected void fraction profiles which decay with depth.

(a) Gaussian (b) Sin (c) two peaks (d ) three peaks (e) random

Figure 5. Relationship between true void fraction and detected void fraction profiles which have wavy peaks.

Table 1. Typical relationships between functions α(y) and β(y). normalized by the total length of the measurement distance.
α(y) β(y) When these two values are sufficiently small, the ratio
  maintains unity like a delta function (see figure 4(e)). The ratio
a a exp − αδ y decreases with the peak void fraction monotonically while it
   −1
b 1 − bδ y b takes a minimum value at around σ = 0.3. For the same set
  y n     y n −1   n 
b 1− l 1 − bδ y 1 − 1
n+1 l
b 1 − yl of data, figure 7 shows the relative error of the peak position
arising in the profile β(y). The error increases with the peak
void fraction and the spatial deviation of the void fraction.
true void fraction is exactly reconstructed from the detected
Concerning the shape of the profile, the coefficient of the
void fraction β(y) as far as β(y) is measured accurately. As
shown in the Gaussian profile, the reduction of the peak value cross correlation between these two functions is calculated
and the displacement of the peak position in β(y) depend on and shown in figure 8. Increases in these parameters result in
the peak void fraction and the integral of α(y). a changing profile. A Gaussian profile for a 10% peak void
Figure 6 shows the ratio of the peak void fractions fraction and 1/4 deviation has a cross correlation coefficient
β max /α max for the Gaussian profile as functions of the peak of 0.9. Thus, the ultrasonic occlusion effect must be corrected
void fraction α max and Gaussian deviation σ , where σ is with the present equation when the local void fraction is 10%

5
Meas. Sci. Technol. 20 (2009) 114003 Y Murai et al

Figure 6. Ratio of two peak void fractions for the Gaussian profile. Figure 8. Cross correlation coefficient of two profiles for the
Gaussian profile.

Figure 7. Relative error of the peak position of the detected void

profile from the true peak position.
Figure 9. Ratio of intergral void profiles for the Gaussian profile.

within the measurement line. Figure 9 shows the ratio of the error. In an actual measurement, the sampling time for
integral defined by obtaining these profiles is finite, and thus the detected void
 y  y −1 fraction involves discrete fluctuations. That is, there is an
R= β(y) dy · α(y) dy . (8) error due to a lack of information.
0 0 Figure 10 demonstrates the reconstruction of the Gaussian
This quantity coincides with the total ratio of the number of profile from limited sampling of the profile β(y). The number
detected bubbles to the number of existing bubbles inside the of sampling profiles (or frames) in the time direction is 1024.
measurement range. When the peak void fraction is 20% and Under such a condition, the void fraction α(y) estimated from
the deviation is 0.3, only 20% of existing bubbles are detected. β(y) has a certain error when compared with the true void
fraction α 0 (y). Here, α 0 (y) is calculated by integrating all
2.4. Estimation of error in the reconstructed void fraction the phase distribution over the sampling time and does not
profile completely match the continuous Gaussian profile because
of a finite number of bubbles. In the range shallower than
The above-mentioned relationships between the two functions the peak position, the reconstructed void fraction matches the
are for ideal continuous functions that completely exclude true void fraction. In contrast, it fluctuates around the value

6
Meas. Sci. Technol. 20 (2009) 114003 Y Murai et al

(a) (b) (c)

Figure 10. Profile transform from the detected void to the true void fraction: (a) α p = 0.05, (b) α p = 0.10 and (c) α p = 0.20.

of true void fraction increasing gradually with the distance assumptions on the attenuation, this and other experimental
after the peak. The comparison among the three cases yields uncertainties would be accounted for by epsilon as
that the magnitude of the error increases for higher peak void
β(y) = β(y) (1 + ε) , (10)
fraction. The increase in the error is explained by the lack of
experimental information of the deep region where only a few where the upper bar indicates the true value and ε the relative
bubbles are detected. The magnitude of the error obeys the error in the profile. We define this type of error considering
inverse root law: the sensitivity of ultrasonic bubble detection. When ultrasonic
√ bubble detection is too sensitive, ε is positively greater.
ε = 1/ N , (9) Substituting equation (10) into equation (3) gives
 −1
where N is the number of detected bubbles at the depth. If only 1 y
ten bubbles are detected, the relative error of the reconstructed α(y) = β(y)(1 + ε) 1 − β(y) dy(1 + ε) . (11)
δ 0
void fraction is 31%. This reduces to 1% when 100 000
bubbles are sampled. The error behaves in the same manner Assuming an ordinary condition for the detected void fraction
as the quantitization error in the resultant void fraction. On expressed as
the other hand, equation (3) can be used for unsteady flow (for |β(y)(1 + ε)|  1, (12)
which void fraction profile changes with time) by reducing the
equation (11) is approximated as
sampling period if the accuracy is neglected. For instance, we
often observe bubble-clustering advection or voidage wave α(y) = β(y) [1 + (1 + ε) B] (1 + ε),
 (13)
in large two-phase flow systems. Such a phenomenon can 1 y
be detected by dividing the total sampling period into small B= β(y) dy.
δ 0
periods. Applying this principle to further minute description
Decomposing equation (13),
of the void distribution, finer structure of the gas phase is
expected to be resolved such as bubble spacing and bubble α(y) = β(y) + β(y)ε + β(y) (1 + ε) B + β(y)B (1 + ε)2 .
size. These are positioned to be the next stage of the study, (14)
which is extension of the development method.
Since the following two equations are available in the dilute
Next, even though the number of sampling times is
condition,
sufficient, another error is introduced when the ultrasonic
bubble detection fails. This failure is due to (1) the bubbles α(y) = β(y)/(1 − B)  β(y)(1 + B), (15)
being too small to be sensed correctly, (2) the pulse width being
too wide to resolve the bubble interface or (3) the ultrasonic α(y) = β(y)/(1 − B)  β(y)(1 + B), (16)
attenuation being irregular or inhomogeneous. Among these
equation (14) is rewritten as
factors, the theory that relates the actual and measured void  y
fractions does not depend on the attenuation. While the 1
α(y) = (1 + ε) α(y) + (1 + ε)2 β(y) β(y) dy. (17)
relationship between alpha and beta was newly derived with δ 0

7
Meas. Sci. Technol. 20 (2009) 114003 Y Murai et al

The above equation involves a nonlinear term for the error. 3. Application to bubbly two-phase flow
This explains why the increase in error results in the
overestimation of the void fraction, reflected as a bias error. In this section, the applications of the present method to
This is consistent with the result shown in figure 10(c); the four different bubbly two-phase flows are presented as a
reconstructed void fraction is slightly higher than the original demonstration. Among four flow configurations, three test
distribution in the region deeper than the peak position. Hence cases include comparison with the optical method.
the ultrasonic sensing of the bubble (closest to the transducer)
must be well tuned before transforming β(y) to α(y). It is worth 3.1. Buoyantly driven two-phase flow in aeration
noting that although the bubble detection is imperfect, two- There is a need in engineering to measure the void fraction
way verification of the profile can be used to correct the result. inside a reactor, such as those used in water treatment,
That is, the profile α(y) reconstructed from β(y) is re-projected sedimentation, fermentation and the refinement of metal. In
to another profile β  (y) using equation (10). The discrepancy such systems, media are highly contaminated with a variety
between β(y) and β  (y) corresponds to the component created of suspensions. The use of ultrasonic sensing of the void
by the bias error. Finding the most adequate value of ε to fraction is hence expected since optical techniques are often
reduce the discrepancy leads to a correction of both α(y) and unavailable. As a basic examination, the bubbling behavior
β(y). However, the application of this proposal is undesirable in a water tank is measured using the present method.
and should only be used when ultrasonic bubble detection has Figure 11(a) shows the flow configuration. Air bubbles are
considerable error. injected through three capillary needles connected to an air
induction tube. Bubbles range from 1.0 to 2.5 mm in diameter.
An ultrasonic transducer is placed on a sidewall of the tank
2.5. On restriction of the reconstruction to measure the bubbles in the horizontal direction. Since the
Weber number of bubbles is around 4, bubbles rise in zigzag
The present reconstruction method is available for bubbly flow,
and spiral modes, resulting in a diffused profile in the upper
i.e. liquid phase is distributed as continuous phase while gas part of the tank. The geometrical relationship between the
phase is as dispersed phase. In contrast, the method cannot be path of ultrasound and the three injection needles is shown on
applied for the other types of gas–liquid two-phase flows such the right. Ultrasound detects only the bubbles that are inside
as for stratified two-phase flows, annular two-phase flows in a the ultrasound beam. Hence, the void fraction is defined
tube and droplet flows. In these flows, the present method still by the volume of these bubbles divided by the cylindrical
detects the first interface closest to the ultrasonic transducer. volume of the beam. As shown in figure 11(b), the first-
However, the void information deeper than the first interface reflecting bubbles are distributed more frequently nearer
will not be obtained since ultrasound propagation is completely the transducer. The void profile, β(x), is thus biased on
blocked. On the other hand, the present reconstruction the left side. The transform using equation (3) gives the
algorithm works regardless of the bubble size as long as the estimated profile of real void fraction, α(x), as shown by
gas phase is dispersed as bubbles in liquid. For example, large solid circles and is flatter in the horizontal direction than
bubbles shield against the ultrasound; however, the ultrasound β(x). Figure 11(c) presents the time-averaged void fraction
still propagates through the liquid between the bubbles to measured by image processing for 60 frames. The image
acquire the void fraction in the deeper layer. processing originally measures the two-dimensional shadow
In the case of small bubbles, there arises another area fraction of bubbles. Hence it has been converted to the
reason for the restriction of the present method. When projection void fraction using the estimated thickness of the
bubbles are smaller than the wavelength of ultrasound, the bubble distribution in the perpendicular direction to the image
sensitivity of the bubble detection decreases because of plane. The comparison with the result indicates that the mean
two factors. One is that local standing wave to detect values of the void fraction roughly match with each other. The
discrepancy is found to be some spiky distribution seen in the
bubble interface will not be formed near the small bubble.
image processing, which is not detected by ultrasound. This
Another is that the acoustic intensity of the reflection
discrepancy originates from the different volume to define the
wave decreases in proportion to bubble diameter squared.
local volume-averaged void fraction. Image processing detects
Microbubbles, so-called, are not to be sensed when ordinary
pixel-by-pixel void fraction so that local chained bubbles in
ultrasound pulse of 0.5–4.0 MHz in basic frequency is used
a few-millimeter row are captured. The ultrasound will not
in water. For such tiny bubbles (e.g. less than 300 μm detect them; however, it captures the profile smoothed in the
in water), the other type of void fraction profiling needs to beam direction. In order to raise the accuracy, one feasible
be replaced. In this context, the present method generally idea is to mount another ultrasonic transducer on the opposite
underestimates the absolute value of void fraction when very side of the tank. This will guarantee the symmetric resolution
small bubbles co-exist in the flow. Alternatively, we need to on the beam direction, too.
know or estimate the bubble size distribution in the target of
flow as external information to be used in the present algorithm. 3.2. Upward jet containing small bubbles
At least the ratio of bubbles that have a diameter smaller than
the ultrasonic wavelength is required to reduce the bias error As another demonstration of upward bubbling flow, we
in the measured void fraction. measured the bubbly two-phase flow ejected from a 13 mm

8
Meas. Sci. Technol. 20 (2009) 114003 Y Murai et al

(a)

(b) (c)

Figure 11. Measurement of bubble distribution in buoyant bubble flow in a quiescent tank: (a) flow and measurement configurations,
(b) void fraction measured by the present method and (c) void fraction measured by image processing for the same condition as (b).

inner diameter nozzle. Bubbles are injected from eight cylindrical structure of the bubbly jet, the projection void
capillary needles of 0.12 mm in inner diameter, which are fraction (i.e. void fraction integrated in the direction of viewing
mounted inside the nozzle. The bubble size observed in axis) is obtained by converting the data of figure 12(b) in
the downstream ranges from 0.5 to 3.0 mm in the case that order to compare the two results on the same frame. The
the initial jet velocity is set as 2 m s−1 . One objective of projection void fraction is derived by integrating the two-
the demonstration is to determine how the small bubbles are dimensional void fraction distribution in the viewing direction.
maintained in the shear layer downstream (Felton and Loth The two-dimensional void fraction is given by projecting the
2002). This is important for estimating the persistency of the on-beam void fraction in the azimuthal direction assuming
jet-modifying effect of bubbles. axisymmetricity. The result shows that there is significant
The void fraction measured by the ultrasonic method is difference from each other. This is caused due the difference
shown in figure 12(b). The individual bubble is measured in spatial resolution as we explained in the previous test case.
using the echo intensity technique. Two cases stand for the However, the following two agreements are confirmed. Case
different gas flow rate supplied at the nozzle: 2.5% in case 1 1 has a higher peak of void fraction near the jet central axis,
and 5.0% in case 2, respectively. We confirm that the bubbles and case 2 has a more diffused profile at a far region from the
clearly tend to accumulate into an annular shear layer of the center. These points are consistent with phenomenological
jet. This accumulation is explained by the distribution of understanding of a two-phase jet. Therefore, we could say
enstrophy, i.e. the roll vortex beside the jet center, which that the functionality of the present ultrasonic technique has
keeps bubbles within the shear layer of the local pressure been confirmed.
gradient in each vortex (Maxey and Riley 1983, Aliseda and
Lasheras 2006). The asymmetric profile of the void fraction 3.3. High-speed two-phase turbulent boundary layer of a
seen along the path of the ultrasound pulse is related to the model ship
flow instability that occurs in the azimuthal direction. The
source of the instability is similar to that seen in two parallel Figure 13 shows measurement data of the bubble distribution
planar jets. inside a turbulent boundary layer of a model ship. The ship
Figures 12(c) and (d) show the comparison of the void is towed at two different speeds in a 400 m long tank. The
fraction profiles between the ultrasonic and the optical length of the ship is 50 m and the ultrasonic transducer for
techniques. Since the image processing is restricted for the measuring the void fraction at its base is located 27 m from

9
Meas. Sci. Technol. 20 (2009) 114003 Y Murai et al

(a)
(b)

(c) (d )

Figure 12. Measurement of the upward bubbly two-phase jet ejected from the concentric nozzle: (a) measurement target and bubble image,
(b) on-beam void fraction measured by the present method, (c) projection void fraction estimated from the ultrasonic data and (d) projection
void fraction measured by image processing.

(a) (b)

Figure 13. Measurement of bubble distribution in the turbulent boundary layer of a towed ship: (a) detected bubbles and (b) void fraction
profiles reconstructed.

the bow of the ship. The purpose of this experiment is to layer, which is difficult to estimate theoretically because of the
confirm and elucidate the frictional drag reduction of a ship by turbulent flow with a very high Reynolds number of around
injecting small bubbles into the boundary layer. The drag is Re = 108 and bubbles with a very high Weber number of
reduced to around 85% (i.e. 15% drag reduction) by injecting We = 103 . In addition, it is impossible to accurately simulate
bubbles, which is determined from the towing force. We such a boundary layer for numerical analyses in any type of
wish to determine the void fraction profile in the boundary two-phase flow model. However, it is known that the drag

10
Meas. Sci. Technol. 20 (2009) 114003 Y Murai et al

(a) (b)

Figure 14. Measurement of horizontal two-phase pipe flow in a slug flow regime: (a) interface detected by the Doppler method and
(b) flowmetering of the liquid phase.

reduction becomes most effective when the void fraction is The upper figure in figure 14(a) shows the gas–liquid
accumulated near the layer of peak Reynolds shear stress interface of slug flow in a horizontal tube. We use air and water
(Kitagawa et al 2005). In short, it is desirable to have a void at room temperature for examination. The tube diameter is
fraction profile similar to the Reynolds shear stress profile to 26 mm and the bulk mean void fraction is 30% at the average
maximize the performance of the drag reduction. However, it total velocity (i.e. total superficial velocity of the two phases)
has not been possible to measure and evaluate the void fraction of 5 m s−1 . A single ultrasonic transducer is set at the bottom of
profile for a bubbly turbulent boundary layer with such high the tube to emit pulses nearly directly upward. The gray curve
Re and We numbers to date. indicates the interface determined using the present ultrasonic
As shown in figure 13(a), the bubbles closest to the method while the black curve is the interface determined using
transducer are detected by ultrasound sensing. In this case, high-speed imaging of the interface at a frame rate of 1000 fps.
the Doppler technique is applied. The time-averaged profile The interface is detected using the Doppler technique. The
of the first-reflecting bubbles, β(y), is then obtained. The two determinations of the interface are conducted at the same
transform to the true void fraction leads to the profiles shown location along the pipe simultaneously by synchronizing the
in figure 13(b). From the results for the two different towing UVP and high-speed video systems. The data show that
speeds, it is confirmed that the peak void fraction is 24 mm ultrasonic detection of the interface is successful for slug
from the ship’s bottom surface for the towing speed U = flow and it matches well the interface obtained by high-speed
4.00 m s−1 and 11 mm for U = 5.14 m s−1 . These results imaging. A small discrepancy between the two interface
clearly indicate that the peak void fraction occurs considerably results is explained by the spatial resolution determined by
deeper than the location of the peak Reynolds shear stress. the thickness of the delta region explained in section 2.1.
This implies that we can further improve the performance By removing the meaningless data for the region above the
of drag reduction by optimizing the bubbling conditions. interface, we obtain a velocity distribution inside the liquid
Moreover, using these new data, we can discuss why the peak phase and the interfacial shape simultaneously. The velocity
approaches the ship hull in the high-speed condition. This is is obtained using the pulse Doppler method in the UVP system.
tightly connected to the following strategy for promoting drag Using these two quantities, the instantaneous volume flow rate
reduction. of liquid is calculated as shown in figure 14(b). The flow rate
is calculated by integrating the velocity profile in the tube from
3.4. Intermittent slugging in gas–liquid two-phase pipe flows the wall to the interface. From the oscillating flow rate, the
slugging force and peak frequency are obtained from power
Not only small bubbles but also large bubbles affect the global
spectrum analysis. The information obtained using the present
flow field significantly. Slug flow in gas–liquid two-phase pipe
technique has great potential in improving the quality of pipe
flow is a typical case. Since the slug flow generates a strong
flow controls and effective safety monitoring of two-phase
vibration both in the flow direction and in the perpendicular
oscillating flows.
direction, the peak frequency and the amplitude should be
investigated. In addition, the origin and development rate of
the vibration must be studied precisely since these phenomena 4. Conclusions
are crucial for mass and energy transportation in pipelines and
nuclear facilities (Murai et al 2007). An optical method is not This study focuses on a problem in the ultrasonic sensing of
feasible for existing or operating pipe flows in these facilities, bubbles in liquid, i.e. the ultrasound not being able to provide
but ultrasonic sensing can contribute to the monitoring and information beyond the bubble closest to the transducer.
online control of two-phase flow as an alternative. Our study addresses this phenomenon and proposes a new

11
Meas. Sci. Technol. 20 (2009) 114003 Y Murai et al

mathematical relationship that enables the reconstruction of Bieberle M, Schleicher E and Hampel U 2008 Simulation study on
the true void fraction along the ultrasonic measurement line. electron beam x-ray CT arrangements for two-phase flow
We assume that the first-reflecting bubbles are correctly measurements Meas. Sci. Technol. 19 094003
Chen Z Q, Hermanson J C, Shear M A and Pedersen P C 2005
measured by the two types of ultrasound sensing: the echo Ultrasonic measurement of interfacial motion of condensing
intensity and Doppler techniques. When this profile, which and non-condensing liquid films Flow Meas. Instrum.
is referred to as the detected void fraction, is obtained by 16 353–64
these techniques, the true void fraction can be estimated Felton K and Loth E 2002 Diffusion of spherical bubbles in a
using equation (3). The difference between the detected void turbulent boundary layer Int. J. Multiph. Flow 28 69–92
Ferrante A and Elghobashi S 2004 On the physical mechanism of
fraction and the true void fraction is explained using several drag reduction in a spatially developing turbulent boundary
analytical functions as well as numerical simulations. We layer laden with microbubbles J. Fluid Mech.
confirm that the difference increases with increases in the 503 345–55
peak void fraction and width of the bubble distribution in Julia J E, Harteveld W K, Muddle R F and Van Den Akker H E A
the measurement direction. While the number of detected 2005 On the accuracy of the void fraction measurements using
optical probes Rev. Sci. Instrum. 76 035103
bubbles decreases with distance in an exponential manner in Kim S, Fu X Y, Wang X and Ishii M 2000 Development of the
the case of a uniform void distribution, the bubble profiles miniaturized four-sensor conductivity probe and the signal
are totally different for other types of distribution. Thus, the processing scheme Int. J. Heat Mass Transfer 43 4101–18
necessity of considering this relationship is explained. The Kitagawa A, Hishida K and Kodama Y 2005 Flow structure of
error produced by the present reconstruction method depends microbubble-laden turbulent channel flow measured by PIV
combined with the shadow image technique Exp. Fluids
only on the original error in the detected void fraction (i.e. the 38 466–75
ultrasonic sensing of bubbles) since the transform equation Krasovitski B and Kimmel E 2001 Gas bubble pulsation in a
of equation (3) is derived deterministically but excludes any semiconfined space subjected to ultrasound J. Acoust. Soc. Am.
physical model or empirical assumption. 109 891–8
To demonstrate and examine the applicability of the Longo S 2006 The effects of air bubbles on ultrasound velocity
measurements Exp. Fluids 41 593–602
present method, four demonstration experiments are carried
Maxey M R and Riley J J 1983 Equation of motion for a small rigid
out under different flow configurations having different bubble sphere in a nonuniform flow Phys. Fluids 26 883–9
sizes. Application to the two-phase turbulent boundary layer Mishima K and Hibiki T 1998 Development of high-frame-rate
of a moving ship, which contains bubbles of around 10 mm neutron radiography and quantitative measurement method for
in diameter, reveals for the first time that the peak of the multiphase flow research Nucl. Eng. Des. 184 183–201
void fraction is located deeper than the position of the peak Mishima K, Hibiki T and Nishihara H 1997 Visualization and
measurement of two-phase flow by using neutron radiography
Reynolds shear stress. Examination of bubbling jet flow gives Nucl. Eng. Des. 175 25–35
phenomenologically valid results consistent with the design Murai Y, Fujii H, Tasaka Y and Takeda Y 2006 Turbulent bubbly
concept, i.e. an annular layer of the void distribution due to channel flow investigated by ultrasound velocity profiler
enstrophy. In the application to slug flow in a tube, which J. Fluid Sci. Technol. 1 12–23
involves elongated bubbles longer than 20 cm, the bubble Murai Y, Inaba K, Takeda Y and Yamamoto F 2007 Backlight
imaging tomography for slug flows in straight and helical tubes
interface is instantaneously and accurately obtained and is a Flow Meas. Instrum. 18 223–9
good match with optical measurements. Murai Y, Matsumoto Y and Yamamoto F 2001 Three dimensional
Since the methodology for one-dimensional spatial measurement of void fraction in a bubble plume using statistic
profiling of the void fraction is established here, the multi- stereoscopic image processing Exp. Fluids 30 11–21
dimensional distribution can be measured in the next stage Murai Y, Oiwa H, Sasaki T, Kondou K, Yoshikawa S
and Yamamoto F 2005 Backlight imaging tomography for
using multiple ultrasonic transducers arranged in different gas–liquid two-phase flow in a helically coiled tube Meas. Sci.
directions. Another possibility is to set two transducers in Technol. 16 1459–68
parallel to simultaneously obtain the advection velocity profile Murakawa H, Kikura H and Aritomi M 2005 Application of
(or phase velocity profile) of the void phase. These expanded ultrasonic Doppler method for bubbly flow measurement using
systems could be used as alternatives to optical methods in the two ultrasonic frequencies Exp. Therm. Fluid Sci.
29 843–50
near future. Murakawa H, Kikura H and Aritomi M 2008 Application of
ultrasonic multi-wave method for two-phase bubbly and slug
Acknowledgments flows Flow Meas. Instrum. 19 205–13
Neal L G and Bankoff S G 1963 A high resolution resistivity probe
for determination of local void properties in gas–liquid flow
The authors are grateful for equipment support from the AIChE J. 9 490–4
National Maritime Research Institute of Japan (NMRI) and Okawa T, Kubota H and Ishida T 2007 Simultaneous measurement
financial support from the Japanese New Energy Development of void fraction and fundamental bubble parameters in
Organization. We also thank Dr Roberto Gonzalez for subcooled flow boiling Nucl. Eng. Des. 237 101–24
technical support with the pipe flow measurement. Schleicher E, da Silva M J, Thiele S, Li A, Wollrab E and Hampel U
2008 Design of an optical tomograph for the investigation of
single- and two-phase pipe flows Meas. Sci. Technol. 19 094006
References Supardan M D, Masuda Y and Uchida S M 2007 The investigation
of gas holdup distribution in a two-phase bubble column using
Aliseda A and Lasheras J C 2006 Effect of buoyancy on the ultrasonic computer tomography Chem. Eng. J. 130 125–33
dynamics of a turbulent boundary layer laden with Suzuki Y, Nakagawa M, Aritomi M, Murakawa H, Kikura H
microbubbles J. Fluid Mech. 559 307–34 and Mori M 2002 Microstructure of the flow field around a

12
Meas. Sci. Technol. 20 (2009) 114003 Y Murai et al

bubble in counter-current bubbly flow Exp. Therm. Fluids Sci. using electrical capacitance tomography Chem. Eng. Sci.
26 221–7 56 6455–62
Takeda Y 1986 Velocity profile measurement by ultrasound Doppler Xu L J and Xu L A 1997 Gas/liquid two-phase flow regime
shift method Int. J. Heat Mass Flow 7 313–8 identification by ultrasonic tomography Flow Meas. Instrum.
Takeda Y 1995 Instantaneous velocity profile measurement by 8 145–55
ultrasonic Doppler method JSME Int. J. B 38 8–16 Yang H C, Kim D K and Kim M H 2003 Void fraction measurement
Warsito W and Fan L S 2001 Measurement of real-time flow using impedance method Flow Meas. Instrum.
structures in gas–liquid and gas–liquid–solid flow systems 14 151–60

13