Sunteți pe pagina 1din 12

Experimental Thermal and Fluid Science 103 (2019) 66–77

Contents lists available at ScienceDirect

Experimental Thermal and Fluid Science


journal homepage: www.elsevier.com/locate/etfs

Bubbly drag reduction investigated by time-resolved ultrasonic pulse T


echography for liquid films creeping inside a turbulent boundary layer
Hyun Jin Park , Yuji Tasaka, Yuichi Murai

Laboratory for Flow Control, Division of Energy and Environmental Systems, Faculty of Engineering, Hokkaido University, N13 W8 Kita-ku, Sapporo 060-8628, Japan

ABSTRACT

Frictional drag reduction due to bubble lubrication was investigated by measuring liquid films creeping along a wall within a two-phase turbulent boundary layer.
We developed noninvasive time-resolved ultrasonic pulse echography for imaging a liquid film at a profiling rate of 3 kHz and a spatial resolution of 50 μm. Various
film patterns were obtained for a 4-m-long flat-bottom model ship towed in a 100-m-long water tank, where a drag reduction rate of 30% was recorded for maximum
air injection. We found that the liquid-film thickness ranged from 50 to 150 wall units of the single-phase turbulent boundary layer; this film was thicker than the
buffer layer depth. With an increase in the air injection flow rate, the average thickness decreased close to the buffer layer limit while the skewness and kurtosis of the
probability density function of the film thickness increased. A sudden transition of the film form was detected according to the kurtosis, allowing the monitoring of
the criteria of the coalescence of dispersed bubbles into large elongated bubbles via ultrasonic pulse echography.

1. Introduction imaging [11], direct-vortex imaging [12], and ultrasound Doppler


monitoring [13,14]. However, most of the required instrumentation can
The injection of bubbles into a wall boundary layer can reduce hardly be introduced to real vessels owing to various restrictions re-
frictional drag, and a variety of drag reduction methods using the lating to bad optical assess, electromagnetic noise, and durability in
bubble injection has been designed and suggested. The air-covered hull high-speed flow environments. Obviously, the seeding of tracer parti-
for the drag reduction has been investigated for > 40 years in the field cles in natural water is prohibited.
of maritime research. Ceccio [1] reviewed various techniques, espe- An important quantity in BDR is the thickness of the liquid film that
cially those for the air-phase replacement of turbulent liquid boundary forms between individual bubbles and the wall. The thickness depends
layers. Ceccio’s group reported how to stabilize the air layer at the on the local wall shear stress. In the laminar boundary layer, the acting
bottom of a ship [2]. The use of small bubbles is called bubbly drag wall shear stress is weaker when the film is thinner. This is because the
reduction (BDR) and is distinguished from air-phase replacement. In the free surface having nearly zero shear approaching the wall reduces the
BDR approach, the modification of turbulence characteristics of the average shear rate inside the film [15–18]. Even in the turbulent
boundary layer affects the comprehensive performance. As reviewed by boundary layer, the same trend is expected as long as the film maintains
Murai [3], several different effects coexist in BDR depending on the laminar flow creeping along the wall. For bubbles larger than the
bubble size. While the underlying physics remains unsolved, the ap- boundary layer thickness, the local wall shear stress strongly correlates
plication of BDR to real marine vessels has advanced in the last decade to the liquid film profile above the bubbles [9,19]. Fig. 1 shows its
[4–6]. A current problem is the gap in the flow conditions between illustrated discovery read from the measurement of local wall shear
fundamental experiments conducted on the laboratory scale and real- stress [9].
ship applications. In the real case, the typical ship speed ranges No past experimental approach has measured the liquid film in the
3–10 m/s, the ship length ranges 10–300 m, and the ship hull has three- assessment of BDR. In the present study, we developed a method of
dimensional curves with roughness that cannot be ignored [6,7]. Re- time-resolved liquid-film measurement employing ultrasonic pulse
presentative dimensionless parameters vary by an order of magnitude echography. The liquid-film thickness is directly obtained because ul-
or more; Reynolds and Froude numbers for real ships are respectively trasound is strongly reflected at gas–liquid interfaces. This point is the
Re > 107 and Fr > 10. Direct numerical simulation is largely un- strongest advantage of the developed method over optical techniques,
available for such flow conditions. To directly extract mechanisms of which mainly visualize the shape parallel to the wall [20,21]. Table 1
BDR, a number of experiments have been conducted employing an lists a number of techniques that have been proposed for the mea-
optical void probe [8], particle image velocimetry [9,10], X-ray surement of liquid films [22–33]. All these techniques are noninvasive


Corresponding author.
E-mail address: park@eng.hokudai.ac.jp (H.J. Park).

https://doi.org/10.1016/j.expthermflusci.2018.12.025
Received 2 July 2018; Received in revised form 13 December 2018; Accepted 21 December 2018
Available online 22 December 2018
0894-1777/ © 2018 Elsevier Inc. All rights reserved.
H.J. Park et al. Experimental Thermal and Fluid Science 103 (2019) 66–77

Nomenclature Re Reynolds number, dimensionless


Rex Reynolds number on a plate, dimensionless
c speed of sound in water, m/s t time, s
Cf frictional coefficient, dimensionless tf time of flight of the ultrasound pulse, s
DR drag reduction rate, dimensionless Umain main flow velocity (equivalent to towing speed), m/s
Frmax supposed maximum Froude number of traveling bubbles, uτ friction velocity, m/s
dimensionless W width of the model ship, m
fs sampling rate for logging the ultrasonic echo, Hz x, y, z Cartesian coordinates of the model ship, m
fu basic frequency of ultrasound, Hz xi value of the ith datum, same dimensions as the data
g acceleration due to gravity, m/s2 x̄ average value of data, same dimensions as the data
H height of the model ship, m Δy spatial resolution of ultrasonic echography, m
h vertical distance between the tip of the transducer and the Δy′ spatial accuracy of ultrasonic echography, m
ship bottom, m αδ void fraction in the turbulent boundary layer, di-
hA superficial air-layer thickness, m mensionless
hL average thickness of the liquid film between bubbles and a αp projection void fraction, dimensionless
horizontal plate, m θ angle of the ultrasonic beam relative to the vertical di-
hmax the maximum height of a bubble injected beneath a hor- rection, degree
izontal plate, m ν kinematic viscosity of water, m2/s
L length of the model ship, m ρ density of water, kg/m3
l wall unit, m σ standard deviation of data, same dimensions as the data
n total number of data, dimensionless τw shear stress acting on the wall, Pa
QA air flow rate for bubble injection, m3/s τw0 shear stress acting on the wall for single-phase flow, Pa
R2 coefficient of determination, dimensionless

in that they do not disturb the film flow, which is sensitive to dis- model ship [35] during the operation of BDR. The final section of this
turbance. Among the listed measurement principles, ultrasound is re- paper discusses the correlation between the liquid film forms and the
latively easy to apply to a ship and it can be used to measure the film drag reduction performance.
thickness through a solid plate; i.e., nothing protrudes into the
boundary layer. Furthermore, in application to sea-sailing ships, elec-
tronic instruments are influenced by minerals and impurities in sea- 2. Experimental facility
water to destabilize the measurement. Other instruments of direct
contact type are often damaged by strong shear force on the hull, re- Fig. 2(a and b) shows schematic diagrams of a flat-bottom model
sulting in a limited duration of measurement. Applying of neutron or ship used in the measurements. The model ship is made of transparent
photon to the ship is infeasibly difficult for permanent installation be- acrylic plates having a length (L) of 4.0 m, height (H) of 0.5 m, and
cause of lack of enough interior spaces. Our previous paper [34] re- width (W) of 0.6 m. x, y, z coordinates are defined as the streamwise
ported the measurement principle of ultrasonic pulse echography for distance from the leading edge of the bow, the vertical depth from the
imaging bubbles in the wall proximity. It used only a single ultrasonic bottom plate, and the spanwise distance from the center of the ship. To
transducer installed at inside surface of ship hull and very simple maintain the one-dimensional spatial development of a turbulent
measurement principle. They give strong durability and allow real-time boundary layer on the plate, a leading edge and two side walls are
monitoring on the measurement. The present study applied the tech- adopted for the ship. The leading edge has a 45° bevel that minimizes
nique to the measurement of a liquid film that flows along a flat-bottom front-edge stagnation (Fig. 2(c)). The two side walls protrude 20 mm
from the bottom plate and are installed with side plates of the ship that
prevent effects of bow-generated splashing waves on flow into the
boundary layer. Air is pumped by a compressor and injected at
x = 0.7 m through 42 holes each having a diameter of 5 mm into the
boundary layer via an isothermal chamber to maintain constant tem-
perature (Fig. 2(d)). There are three measurement ports on the bottom

Table 1
Techniques for measuring the thickness of a film flow.
Signal source Measurement principle Authors

Electricity Electric capacitance Cui et al. [22]


Electric conductance Hagiwara et al. [23]

Neutron Attenuation Zboray and Prasser [24]

Photon Absorption Mendez et al. [25]


Fluorescence Cherdantsev et al. [26]
Interference Butler et al. [27]
Level detection Mendez et al. [25]
Laser displacement gauge Serizawa et al. [28]
Laser focusing distance Hazuku et al. [29]
Total reflection Hurlburt and Newell [30]
Fig. 1. Streamwise profile of the friction coefficient modified by the passage of Shadow Alekseenko et al. [31]
a large single bubble; (a) a typical bubble shape in a quasi-steady state and (b) a
Ultrasound Echo intensity Hunter et al. [32]
local friction profile correlating to the thickness of the liquid film between the
Time of flight Lu et al. [33]
bubble interface and a flat wall [9].

67
H.J. Park et al. Experimental Thermal and Fluid Science 103 (2019) 66–77

Fig. 2. Schematic diagram of the model ship; (a) overall view, (b) side view, (c) details of the bow, (d) details of the bubble injector, and (e) details of the
measurement port.

plate at x = 1.1, 2.3, and 3.5 m. Each port has a shear stress sensor [35] Table 2
and an ultrasonic transducer (see Fig. 2(e)). The transducer is mounted Parameters of the measurement instruments.
in refractive-index-matching material at a tilt angle of θ = 8° approxi- Ultrasonic pulse generator
mately 28 mm from the sensor. For this setting, the propagation line of Ultrasonic basic frequency (fu) 4 MHz
the ultrasonic wave emitted from the transducer crosses the plate–water Number of cycle 4 –
interface at a distance of 25 mm from the sensor in the downstream Pulse repetition frequency 3.2 kHz
Voltage for ultrasonic emission 150 V
direction. The transducer is connected to an ultrasonic velocity profiler
(UVP-DUO MX, MET-FLOW S.A.) that generates an ultrasonic pulse and Ultrasonic transducer
Ultrasonic beam diameter 5 mm
a data logger (DIG-100M1002-PCI, CONTEC) that records ultrasonic
Divergence half-angle 2.2 deg.
echo signals. The index-matching material prevents multiple reflections
Data logger for recording ultrasonic echo
at the plate–water interface by reducing the difference in acoustic im-
Sampling frequency (fs) 50 MHz
pedance against water. Recording time 336 ms
The air flow rate is controlled by a flow controller, consisting of a Range of voltage ±2 V
thermometer, two pressure sensors, a servo valve, and a flowmeter as Resolution of voltage 4 mV
shown in Fig. 3. Controller system linked with a personal computer
maintains a fixed air flow rate even if the air pressure in the isothermal
chamber fluctuates. In terms of sailing the model ship, experiments for approximately 7 s. Information of the measurement system and
were performed using a towing train mounted on a water tank with towing facility is summarized in Tables 2 and 3.
length of 100 m and depth of 4 m at Hiroshima University, Hiroshima,
Japan. The train tows the model ship at a fixed speed (Umain) within the 3. Experimental conditions
range 0–3.0 m/s. The train can maintain the maximum speed, 3.0 m/s,
Fig. 4 shows the relationship between the frictional coefficient (Cf)

Fig. 3. Schematic diagram of the synchronized system for multiple measuring instruments.

68
H.J. Park et al. Experimental Thermal and Fluid Science 103 (2019) 66–77

Table 3 is similar to the length of our model ship. He found that the drag was
Dimensions, performance, and conditions of the towing test facility. almost extinguished at hA > 3 mm, where hA is the apparent air layer
Water tank thickness defined as
Length in towing direction 100 m QA
Depth 3.5 m hA =
Width 8.0 m WUmain (5)
Temperature of water 29 °C
Here W denotes the spanwise width of bubble injection. This quantity is
Properties of water estimated from the temperature often used to summarize the overall drag reduction rate of a ship hull
Density of water (ρ) 996 kg/m3
[38]. We here suppose that drag almost extinguished at hA ∼ 3 mm
Kinematic viscosity of water (ν) 0.847 × 10−6 m2/s
Speed of sound in water (c) 1507 m/s because the air film covered the whole plate. In the present experi-
ments, therefore, hA was set at 0.23 mm < hA < 2.08 mm because the
Towing train
Towing speed (Umain) 0–3.00 m/s air film did not cover the whole ship bottom.
The maximum mileage in the tank 80 m Sample snap images taken near the forward, mid-ship, and aft
measurement ports are shown in Figs. 5–7. The figures show that the
bubble shape depended on the experimental conditions Umain and QA.
When both parameters were set low, bubbles were tiny and spherical.
Under these conditions, bubbles were smaller than 8 mm, which is the
maximum bubble size possible in turbulent shear [39,40]. At higher QA,
the bubbles grew through coalescence owing to their increased mutual
contact. At maximum QA, film-like, large bubbles covered a wide area
of the hull and there were capillary waves on the bubbles. An increase
in Umain tended to elongate the bubbles. Additionally, by comparing
images taken at different locations, we confirm that states of bubbles
traveling in the turbulent boundary layer varied even for the same
experimental conditions. However, it is impossible to estimate the li-
quid-film thickness of individual bubbles from the images.

4. Ultrasonic pulse echography

Fig. 8(a) shows an example of the time series of the raw echo am-
plitude distribution as bubbles are injected. The measurement was
made at the forward measurement port (x ∼ 1.1 m) for the experi-
mental conditions Umain = 2.0 m/s and QA ∼ 1.67 m3/s. Here, both
horizonal and vertical axes represent time. The vertical axis depicts the
echo signal of an individual pulse on the order of microseconds, and the
Fig. 4. Friction coefficient as a function of the Reynolds number [35], where repetition result is blown up on the horizontal axis on the order of
the two curves represent the Blasius friction law for laminar flows and the milliseconds. There was electronic noise generated by ultrasonic pulse
empirical friction coefficient for turbulent flows [36].
emission within 10.00 μs of the emission; i.e., tf < 10.00 μs. To avoid
the erroneous detection of bubbles from an echo with noise, a certain
and Reynolds number (Rex) measured for the model ship, where we distance, h, between the tip of the transducer and the bottom surface
define was secured for space to attenuate the noise. The gas–liquid interfaces
2 w0 allow the total reflection of the ultrasonic wave because of the very
Cf = large gap in acoustic impedance between air and water phases, and the
2
Umain (1)
echo signal should thus vary. Such variation appears around
and tf ∼ 20.00 μs in the example result. By subtracting the average echo
xUmain distribution measured under the single-phase condition from that of
Rex = bubbly flows, it is possible to detect the location of the upper surface of
(2)
the bubbles. The location detected by subtraction is shown in Fig. 8(b),
Here, τw0 is the shear stress in a single-phase flow measured using the where a gray line indicates the interface of the bottom plate.
shear stress sensor; τw0 ranged from 0 to 15 Pa in the test. In the figure, Considering a tilt angle of the transducer θ = 8°, the thickness of the
solid lines indicate the Blasius friction law for laminar boundary layers liquid film between the ship bottom and bubbles is calculated as
and the empirical friction coefficient proposed by Schlichting for tur-
bulent boundary layers [36]; these are respectively expressed as t f c cos
y= h
2 (6)
Cf = 1.328Rex 1/2
(3)
where h is the vertical distance between the tip of the transducer and
and the wall interface as seen in Fig. 2(e). Here, the spatial resolution and
0.558Cf 1/2 spatial accuracy are respectively evaluated as
Cf e = Rex 1 (4)
c / fs
Cf measured for the model ship approximately obeys Schlichting’s curve y= 15 µm
2 (7)
at Rex > 2 × 106. To ensure turbulent conditions, experiments were
performed at Umain > 2.0 m/s, and the corresponding minimum Rex at and
the forward measurement port (x = 1.1 m) was 2.6 × 106.
c/4fu
To generate bubbly flows beneath the ship bottom, the flow rate of y 50 µm
2 (8)
the air volume (QA) supplied to the bubble injector was set in the range
of 0.42–2.50 × 10−3 m3/s. Tokunaga [37] investigated frictional drag The maximum amplitude of the echo from the bubble surface varies
with bubble injection using a flat plate having a length of 3.5 m, which with the shape and size of bubbles [34]. If the maximum echo

69
H.J. Park et al. Experimental Thermal and Fluid Science 103 (2019) 66–77

Fig. 5. Snap images of bubbles taken near the forward measurement port at x ∼ 1.1 m for various air flow rates and main-flow velocities, where hA is the corre-
sponding apparent air layer thickness.

Fig. 6. Snap images of bubbles taken near the mid-ship measurement port at x ∼ 2.2 m for various air flow rates and main-flow velocities, where hA is the corre-
sponding apparent air layer thickness.

70
H.J. Park et al. Experimental Thermal and Fluid Science 103 (2019) 66–77

Fig. 7. Snap images of bubbles taken near the aft at x ∼ 3.3 m for various air flow rates and main-flow velocities, where hA is the corresponding apparent air layer
thickness.

amplitude is sufficiently high, the bubble surface will be detected at the maximum is located at a quarter and three quarters of a single cycle of
rising edge of the echo waveform. In contrast, the surface is barely the ultrasonic pulse. Therefore, the accuracy is determined as ap-
detected when the amplitude reaches the maximum value or cannot be proximately 50 μm for the current specifications. Here, we neglect un-
detected from the waveform. If the maximum amplitude is higher than certainty of the measurement caused by curvature of the upper surface
a threshold value for surface detection, the surface is detected within 1/ of a bubble, because almost bubbles have large sizes comparing with
4fu seconds from appearance of the echo waveform because the the ultrasonic beam diameter excepting the lowest QA condition and

Fig. 8. Example times series of ultrasonic measurement results; (a) raw echo amplitude distribution and (b) detected upper surface of bubbles, where a gray line
indicates the interface of the bottom plate.

71
H.J. Park et al. Experimental Thermal and Fluid Science 103 (2019) 66–77

these large bubbles have sufficiently flat upper surface excepting on


their head and tail.
The thickness of the liquid film can be measured from the time in-
terval between the detected bubble surface and the wall, by multiplying
by the speed of sound in water. An example is shown in Fig. 9. The
lower abscissa shows the streamwise length scale of the interface,
which is estimated by multiplying the time with the average advection
velocity of bubbles (approximately 0.5Umain [35]). In the example, we
find that a large bubble had a flat upper surface in its central part where
the liquid film became thinner than 500 μm. This thickness is com-
parable to 50 wall units; i.e., the buffer layer border of the single-phase
turbulent boundary layer at this flow speed. Here, the wall unit l is
estimated as

2
l= = =
u w Umain Cf (9)

where uτ is the friction velocity. For 2 < Umain < 3 m/s, the wall unit
is from 8 to 10 μm. The turbulent boundary layer obeys the universal
velocity profile law at 104 < Re < 107, a viscous sublayer exists from
0 to 5 wall units, a buffer layer exists from 5 to around 50 wall units,
and a log-law region exists beyond 50 wall units from the wall. The
liquid film interface appearing at the buffer layer border means that the
connecting log-law region is replaced by the air phase with less net
momentum. We also find that the liquid film has an asymmetric form in
the main flow direction in that the front part (left side in the figure) is
thick. A sharp fluctuation is detected by carefully observing the film at
t ∼ 50 ms. This fluctuation is attributed to the capillary wave standing
on the film subject to strong shear. A small bubble is also visible at cf.
3 ms < t < 9 ms, resulting in arch-like plots because of the spherical
interface.

5. Results and discussions

5.1. Projection of the void fraction and liquid-film thickness

We performed an echography measurement four times for each


experimental condition. Each measurement period was 1300 ms, during
which 4000 profiles of the interface were detected at a frequency of
3.2 kHz. Considering the advection velocity of bubbles, the sampling
duration corresponds to 1.2–1.8 m in the streamwise direction. Because
size of bubbles (Figs. 5–7) ranges from a few mm to a few cm, about 50 Fig. 10. Projection void fraction measured by echography at each measurement
or more number of bubbles with different sizes and shapes passed the port versus the gas flow rate QA; (a) Umain = 2.0 m/s, (b) 2.5 m/s and (c) 3.0 m/
measurement ports during the measurements. s.
Fig. 10 shows variations in the projection void faction (αp) with
respect to the air flow rate QA, obtained from echography. The pro-
projection void fraction αp takes an intermediate value upon balance.
jection void fraction αp is defined as the ratio of the time for bubbles
According to the location dependency of αp, the position at which
existing on the ultrasonic measurement line to the total measurement
coalescence began is deduced to be around the forward port at
duration. As a macroscopic trend, αp increased with QA and decreased
Umain = 2.0 m/s, between the forward and mid-ship ports at
with downstream migration. We also find that the downstream decrease
Umain = 2.5 m/s, and between the mid-ship and aft ports at
in αp was appreciable at higher ship speed Umain. This infers that the
Umain = 3.0 m/s. These results correspond consistently to direct ima-
coalescence of bubbles and bubble clustering were dominant in the
ging result of bubbles (Figs. 5–7).
downstream region. In parallel, large bubbles were fragmented by
As αp varied, the liquid-film thickness (hL) above the bubbles was
turbulent shear stress inside the boundary layer. Consequently, the

Fig. 9. Example ultrasonic measurement results, where the streamwise length and thickness in the figure are converted from the time and time of flight in Fig. 8.

72
H.J. Park et al. Experimental Thermal and Fluid Science 103 (2019) 66–77

measured as shown in Fig. 11. Symbols show ensemble-averaged values n (n + 1) xi x¯ 4


3(n 1) 2
Kurtosis =
while bars indicate standard deviations. The average film thickness was (n 1)(n 2)(n 3) (n 2)(n 3)
measured in the range of 500–1200 μm. The range is comparable to (12)
50–150 wall units; i.e., the liquid-film interface located at the buffer-
layer border to the log-law region. Vertical bars indicate the standard where n, xi, x̄ , and σ respectively indicate the number of data, value of
deviation which has large values in any case. The deviation originates the ith datum, average, and standard deviation. Results are presented in
from two factors; one is the case of tiny bubbles floating inside the Fig. 13. Both skewness and kurtosis are zero when the liquid-film
boundary layer, of which instantaneous location directly reflects to the thickness fluctuates in a random manner obeying a normal distribution.
deviation. The other is the case of large bubbles that have curvature in In the present results, skewness and kurtosis have positive values
their front and the rear parts (see Figs. 1 and 9), which also results in throughout the range tested. Positive skewness means an asymmetric
rapid change of the liquid film thickness even during the passage of a probability of the liquid-film thickness and is evidence that each liquid
single bubble. These two factors are categorized to be geometric de- film has a flat region around the center of a bubble. Positive kurtosis
viation dependent on the size and the shape of the bubble. Thus, the indicates the frequent occurrence of a liquid film long in the flow di-
deviation does not mean the error of measurement. The thickness hL has rection. Such trends intensify as the air flow rate QA increases. How-
three different tendencies depending on the air flow rate QA, ship speed ever, the increasing rate is suppressed in the downstream region. Fur-
Umain, and travelling distance. Most importantly, hL became thinner thermore, we find that skewness and kurtosis jump together in a
with an increase in QA. This trend relates to the reduction of wall shear particular range of QA. This indicates a morphological transition of the
stress. Secondly, we find that the thickness further decreased with film flow inside the two-phase turbulent boundary layer.
downstream migration. These two trends are explained by an increase
in bubble size in that bubbles of a higher void fraction coalesce into
large bubbles that have thinner liquid film above them. Additionally, an
increase in ship speed Umain produced thicker liquid films. This is at-
tributed to the dominance of bubble fragmentation under highly tur-
bulent shear stress [40].
Reading from the data of Fig. 11, we constructed a simple fitting
function to comprehensively parameterize the average liquid-film
thickness hL in dimensionless form:

0.18
hL hA Umain QA
= (0.01Frmax + 0.02) , Frmax = , hA =
h max h max gh max WUmain
(10)

where hmax is the maximum thickness of air bubbles that can be kept
beneath a flat plate in quiescent water [41]. Thus, hmax already ac-
counts for the surface tension of the gas–liquid interface. The right-
hand side of Eq. (10) gives the maximum Froude number Frmax and the
apparent air layer thickness hA. These parameters, hmax and Frmax, are
obtained from dimension analysis and are employed in the fitting
equation. Two numerical constants 0.01 and 0.02 are obtained for a
linear function assumed for the factor to hA. The effect of the traveling
distance was negligible compared with the effects of other parameters,
and we thus ignored it in the function. Three constants appearing in Eq.
(10) were determined employing the least-squares approach and all the
data obtained. Fig. 12 shows the relationship between hL and hA, where
symbols are measurements and solid curves are calculated using Eq.
(10). Coefficient of determination R2 between all average data and the
empirical equation is 0.79 and the figure confirms that the fitting
function represents well the average liquid-film thickness, which varies
with hA.

5.2. Higher-order moments of the liquid-film thickness

While the liquid-film thickness strongly fluctuates in time, we


evaluate its higher-order moments to extract morphological informa-
tion. From the probability density function of the liquid-film thickness,
the skewness (third-order moment) and kurtosis (fourth-order moment)
are computed as

3
n xi x¯
Skewness =
(n 1)(n 2) (11) Fig. 11. Thickness of the liquid film between the ship bottom and bubbles
versus the gas flow rate QA, where vertical bars and gray lines indicate standard
and deviations and 1000 μm respectively; (a) Umain = 2.0 m/s, (b) 2.5 m/s and (c)
3.0 m/s.

73
H.J. Park et al. Experimental Thermal and Fluid Science 103 (2019) 66–77

Figs. 5–7, the state of stress can be identified using skewness and kur-
tosis.
A sudden jump in values in the actual data (Fig. 13) is detected in
the forward part at QA > 1.5 × 10−3 m3/s. This indicates that small
spherical bubbles transform into air-cavity-type bubbles owing to
bubble coalescence. This transition further shapely occurs in the high-
speed case that Umean = 3.0 m/s.

5.3. Correlation to drag reduction rates

Fig. 15 shows drag reduction rate DR measured at each measure-


ment location, where DR was measured by the shear stress sensor that
has an effective sensing diameter of 10 mm [19] and averaged over
time. At the forward location, drag rather increases with air flow rate.
At the mid-ship and the aft locations, drag reduction occurs at
αδ > 2.0%. Here, αδ is void fraction inside the turbulent boundary
layer and becomes lower in downstream region as a constant QA is
given because of the downstream expansion of the boundary layer
thickness [35]. In several previous BDR researches [1,14], such a re-
versing trend of DR for the location and the void fraction has been left
unclear. We think that it requires an entrance length for developed
bubbly flow to reduce the drag by injected bubbles. [42].
Fig. 16 presents the relation between hL and DR. The result confirms
that the two parameters have a certain correlation although a large
deviation remains. In the forward region (Fig. 16(a)), almost drag re-
duction rates were negative; i.e., the wall shear stress increased with
bubble injection. The average hL under this condition ranged from 750
to 1250 μm, being much greater than the depth of the buffer layer. The
drag increment is therefore attributed to the turbulent momentum
transfer in the log-law region, which is additionally enhanced by the
bubbles as the bubbles are suspended far from the wall. At the mid-ship
and aft locations (Fig. 16(b)), the drag reduction rate DR ranged from
20% to 75%. At these locations, the average hL ranged from 400 to
650 μm, which is comparable to the thickness of the buffer layer. This
indicates that bubbles do not enter the buffer layer but leave the liquid
film in the laminar state above them while they provide a large drag
reduction. In other words, the bubbles just touching the buffer layer
effectively suppress turbulent momentum transfer.
Correlation coefficient between hL and DR was 0.86 at the forward
port, −0.97 at the mid-ship, and −0.51 at the aft location. The positive
correlation for the forward port supports the idea that bubbles sus-
Fig. 12. Relationship between liquid-film thickness and superficial air layer pended in the log-law region enhance turbulent mixing more as they
thickness, where a solid line indicates the present empirical formula derived approach the border with the buffer layer. This results in an appreciable
from the present experimental results; (a) Umain = 2.0 m/s, (b) 2.5 m/s and (c) increase in drag at −80% < DR < −10%. The negative correlations
3.0 m/s. obtained at the mid-ship and aft locations demonstrate the idea that the
bubble interface co-flowing along the border of the buffer layer can cut
off the turbulent momentum transfer most effectively. Recalling Eq.
To understand the trends of the higher-order moments, we examine
(10), hL is approximately proportional to QA−0.18 while DR increases
how skewness and kurtosis are computed for several different bubbles.
linearly with QA. From these two relations, hL is estimated to be pro-
Top figures in Fig. 14 are outlines of the bubbles sampled from real
portional to DR−0.18, and DR thus increases with hL−5.1. This suggests
experimental data for a horizontal turbulent channel flow [10]. The
that a slight thinning of the liquid film strongly promotes drag reduc-
figure shows (a) a spherical bubble, (b–e) deformed bubbles, and (f–g)
tion. We need to conduct future work to elucidate what physics de-
largely stretched bubbles accompanying capillary waves on the upper
termines the exponential value of −5.1. This will require the con-
surface. In the case of (a) a spherical bubble, skewness is +0.5 and
sideration of three factors: the shape-recovery force of bubbles (i.e., the
kurtosis is −0.2. These values are determined by the geometric nature
We number dependency), bubble buoyancy pushing the film upward
of a sphere. For deformed bubbles (b–f), skewness and kurtosis re-
(i.e., the Fr number dependency), and turbulence in the liquid phase
spectively increase to approximately +2 and +5. The small capillary
around each bubble (i.e., the Re number dependency).
wave in (f) does not affect these quantities appreciably. For bubbles in
(g), the skewness and kurtosis rapidly increase while the bubble size is
6. Conclusions
comparable to the bubble size in (f). This is because the bubbles in (g)
have no slope on the upper surface, which is a result of the yield state in
We investigated the internal structure of BDR by observing liquid
that the fluid stress below the bubble does not transfer to the liquid
films that form above individual bubbles. To measure a liquid film
film, and the bubbles play as the role of local air cavities. While the
flowing at a speed of up to 3 m/s, we employed ultrasonic pulse
bubbles in (g) and (f) cannot be distinguished in images such as those of
echography, which is a technique that we previously developed. The

74
H.J. Park et al. Experimental Thermal and Fluid Science 103 (2019) 66–77

Fig. 13. Skewness and kurtosis of the probability density function of the liquid-film thickness, where zero values indicate a normal distribution.

Fig. 14. Example statistical characteristics of the liquid-film thickness above various types of bubble; (top) shapes of bubbles and (bottom) skewness and kurtosis; (a,
f, g) sketches of bubbles, (b–e) bubbles observed in a channel flow [10].

echography can sense liquid film profiles at a sampling frequency of reduction rate higher than 50%, bubbles left the liquid film with a
3.2 kHz and spatial resolution of 50 μm. This specification allowed us to thickness comparable to the buffer layer border. In contrast, for greater
quantitatively monitor film profiles during the passage of individual drag, bubbles left liquid films protruding into the log-law region of the
bubbles. We targeted the bubbly two-phase turbulent boundary layer boundary layer. The statistical analysis of the probability density
that has a spatial development beneath a 4-m-long flat-bottom model function of the liquid-film thickness revealed a morphological transi-
ship towed in a 100-m-long water tank. Measurements revealed that the tion of the film behavior from a dispersed bubble state to air-cavity-type
liquid film had a thickness in the range from 500 to 1200 μm, which is bubbles via the computation of skewness and kurtosis. These higher-
50 to 150 wall units of the single-phase turbulent boundary layer. As moment values of the liquid film have not be sensed by optical imaging
the air flow rate was increased to reduce drag, the average liquid-film but were quantitatively monitored for the first time with the present
thickness decreased as represented by Eq. (10). In the case of a drag time-resolved ultrasonic pulse echography.

75
H.J. Park et al. Experimental Thermal and Fluid Science 103 (2019) 66–77

Fig. 15. Drag reduction rate, where bars indicate standard deviations; (a) forward and (b) mid-ship and aft measurement locations [35].

Fig. 16. Drag reduction rate versus liquid-film thickness, where bars indicate standard deviations in both directions; (a) forward and (b) mid-ship and aft mea-
surement locations.

Conflict of interest [9] A. Kitagawa, K. Hishida, Y. Kodama, Flow structure of microbubble-laden turbulent
channel flow measured by PIV combined with the shadow image technique, Exp.
Fluids 38 (2005) 466–475.
We wish to confirm that there are no known conflicts of interest [10] Y. Oishi, Y. Murai, Horizontal turbulent channel flow interacted by a single large
associated with this publication. bubble, Exp. Therm. Fluid Sci. 55 (2014) 128–139.
[11] S.A. Mäkiharju, C. Gabillet, B.G. Paik, N.A. Chang, M. Perlin, S.L. Ceccio, Time-
resolved two-dimensional X-ray densitometry of a two-phase flow downstream of a
Acknowledgment ventilated cavity, Exp. Fluid 54 (2013) 1561.
[12] H.J. Park, Y. Tasaka, Y. Oishi, Y. Murai, Vortical structures swept by a bubble
This work was supported by the Fundamental Research Developing swarm in turbulent boundary layers, Chem. Eng. Sci. 116 (2014) 486–496.
[13] Y. Murai, S. Ohta, A. Shigetomi, Y. Tasaka, Y. Takeda, Development of ultrasonic
Association for Shipbuilding and Offshore (REDAS), and JSPS KAKENHI void fraction profiler, Meas. Sci. Tech. 20 (2009) 114003.
under grant numbers 17 K14583 and 17H01245. We acknowledge Dr. [14] H.J. Park, Y. Tasaka, Y. Oishi, Y. Murai, Drag reduction promoted by repetitive
Oishi at Muroran Institute of Technology and Prof. Yasukawa at bubble injection in turbulent channel flows, Int. J. Multiphase Flow 75 (2015)
12–25.
Hiroshima University for their support relating to the experiments.
[15] V. Sobolík, O. Wein, J. Čermák, Simultaneous measurement of film thickness and
wall shear stress in wavy flow of non-Newtonian liquids, Collect. Czechosl. Chem.
References Commun. 52 (1987) 913–928.
[16] P. Tisné, L. Doubliez, F. Aloui, Determination of the slip layer thickness for a wet
foam flow, Colloids Surf. A: Physicochem. Eng. Asp. 246 (2004) 21–29.
[1] S.L. Ceccio, Frictional drag reduction of external flows with bubble and gas injec- [17] A. Saugey, W. Drenckhan, D. Weaire, Wall slip of bubbles in foams, Phys. Fluids 18
tion, Annu. Rev. Fluid Mech. 42 (2010) 183–203. (2006) 053101.
[2] S.A. Mäkiharju, I.H. Lee, G.P. Filip, K.J. Maki, S.L. Ceccio, The topology of gas jets [18] M. Le Merrer, R. Lespiat, R. Höhler, S. Cohen-Addad, Linear and non-linear wall
injected beneath a surface and subject to liquid cross-flow, J. Fluid Mech. 818 friction of wet foams, Soft Matter 11 (2015) 368–381.
(2017) 141–183. [19] Y. Murai, H. Fukuda, Y. Oishi, Y. Kodama, F. Yamamoto, Skin friction reduction by
[3] Y. Murai, Frictional drag reduction by bubble injection, Exp. Fluids 55 (2014) 1773. large air bubbles in a horizontal channel flow, Int. J. Multiphase Flow 33 (2007)
[4] S. Mizokami, M. Kawakado, M. Kawano, T. Hasegawa, I. Hirakawa, Implementation 147–163.
of ship energy-saving operations with Mitsubishi air lubrication system, MHI Tech. [20] J. Huang, Y. Murai, F. Yamamoto, Shallow DOF-based particle tracking velocimetry
Rev. 50 (2013) 44–49. applied to horizonal bubbly wall turbulence, Flow Meas. Inst. 19 (2008) 93–105.
[5] J. Jang, S.H. Choi, S. Ahn, B. Kim, J.S. Seo, Experimental investigation of frictional [21] M.J. Sathe, I.H. Thaker, T.E. Strand, J.B. Joshi, Advanced PIV/LIF and shadow-
resistance reduction with air layer on the hull bottom of a ship, Int. J. Nav. Archit. graphy system to visualize flow structure in two-phase bubbly flows, Chem. Eng.
Ocean Eng. 6 (2014) 363–379. Sci. 65 (2010) 2431–2442.
[6] I. Kumagai, Y. Takahashi, Y. Murai, A new power-saving device for air bubble [22] Z. Cui, C. Yang, B. Sun, H. Wang, Liquid film thickness estimation using electrical
generation using a hydrofoil for reducing ship drag: theory, experiments, and ap- capacitance tomography, Meas. Sci. Rev. 14 (2014) 8–15.
plications to ships, Ocean Eng. 95 (2015) 183–194. [23] Y. Hagiwara, E. Exmaeilzadeh, H. Tsutsui, K. Suzuki, Simultaneous measurement of
[7] J. Johansen, A.M. Castro, P. Carrica, Full-scale two- phase flow measurements on liquid film thickness, wall shear stress and gas flow turbulence of horizontal wavy
Athena research vessel, Int. J. Multiphase Flow 36 (2010) 720–737. two-phase flow, Int. J. Multiphase Flow 15 (1989) 421–431.
[8] H. Kato, T. Iwashina, M. Miyanaga, H. Yamaguchi, Effect of microbubbles on the [24] R. Zboray, H.M. Prasser, Measuring liquid film thickness in annular two-phase flows
structure of turbulence in a turbulent boundary layer, J. Mar. Sci. Tech. 4 (1999) by cold neutron imaging, Exp. Fluids 54 (2013) 1596.
155–162. [25] M.A. Mendez, L. Németh, J.M. Bunhlin, Measurement of liquid film thickness via

76
H.J. Park et al. Experimental Thermal and Fluid Science 103 (2019) 66–77

light absorption and laser tomography, EPJ Web. Conf. 114 (2016) 02072. [34] H.J. Park, Y. Tasaka, Y. Murai, Ultrasonic pulse echography for bubbles traveling in
[26] A.V. Cherdantsev, D.B. Hann, B.J. Azzopardi, Study of gas-sheared liquid film in the proximity of a wall, Meas. Sci. Technol. 26 (2015) 125301.
horizontal rectangular duct using high-speed LIF technique: Three-dimensional [35] H.J. Park, Y. Oishi, Y. Tasaka, Y. Murai, Void waves propagating in the bubbly two-
wavy structure and its relation to liquid entrainment, Int. J. Multiphase Flow 67 phase turbulent boundary layer beneath a flat-bottom model ship during drag re-
(2014) 52–64. duction, Exp. Fluids 57 (2016) 178.
[27] C.S. Butler, Z.L.E. Seeger, T.D.M. Bell, A.I. Biship, R.F. Tabor, Local determination [36] H. Schlichting, Boundary-Layer Theory, seventh ed., McGraw-Hill Higher
of thin liquid film profiles using colour interferometry, Eur. Phys. J. E 39 (2016) 14. Education, New York, 1979.
[28] A. Serizawa, K. Nagane, T. Ebisu, T. Kamei, Z. Kawara, K. Torikoshi, Dynamic [37] K. Tokunaga, Reduction of frictional resistance of a flat plate by microbubbles,
measurement of liquid film thickness in stratified flow by using ultrasonic echo West-Jpn. Soc. Nav. Archit. Ocean Eng. 73 (1987) 79–82.
technique, in: Proc. of the 4th Int. Topical Meeting on Nucl. Thermal-Hydraulics, [38] T. Takahashi, A. Kakugawa, M. Makino, Y. Kodama, Experimental study on scale
Operations and Safety, Taipei, Taiwan, 1994, pp. 42-C. effect of drag reduction by microbubbles using very large flat plate ships, J. Kansai
[29] T. Hazuku, N. Fukamachi, T. Takamasa, T. Hibiki, M. Ishii, Measurement of liquid Soc. Nav. Archit. Jpn. 239 (2003) 11–20 (in Japanese).
film in microchannels using a laser focus displacement meter, Exp. Fluids 38 (2005) [39] J.O. Hinze, Fundamentals of the hydrodynamic mechanism of splitting in dispersion
780–788. processes, AIChE J. 1 (1955) 289–295.
[30] E.T. Hurlburt, T.A. Newell, Optical measurement of liquid film thickness and wave [40] W.C. Sanders, E.S. Winkel, D.R. Dowling, M. Perlin, S.L. Ceccio, Bubble friction drag
velocity in liquid film flows, Exp. Fluids 21 (1996) 357–362. reduction in a high-Reynolds-number flat-plate turbulent boundary layer, J. Fluid
[31] S.V. Alekseenko, V.Y. Nakoryakov, B.G. Pokusaev, Wave formation on a vertical Mech. 552 (2006) 353–380.
falling liquid film, ALChE J. 31 (1985) 1446–1460. [41] S. Das, Y.S. Morsi, G. Brooks, J.J.J. Ghen, W. Yang, Principal characteristics of a
[32] A. Hunter, R. Dwyer-joyce, P. Harper, Calibration and validation of ultrasonic re- bubble formation on a horizontal downward facing surface, Colloids Surf. A:
flection methods for thin-film measurement in tribology, Meas. Sci. Technol. 23 Physicochem. Eng. Asp. 411 (2012) 94–104.
(2012) 105605. [42] H.J. Park, Y. Tasaka, Y. Murai, Bubbly drag reduction accompanied by void wave
[33] Q. Lu, N.V. Suryanarayana, C. Christodoulu, Film thickness measurement with an generation inside turbulent boundary layers, Exp. Fluids 59 (2018) 164.
ultrasonic transducer, Exp. Therm. Fluid Sci. 7 (1993) 354–361.

77

S-ar putea să vă placă și